ChatGPT

Manufacturing executives are wagering nearly half their modernisation budgets on AI, betting these systems will boost profit within two years. This aggressive capital allocation marks a definitive pivot. AI is now seen as the primary engine for financial performance. According to the Future-Ready Manufacturing Study 2025 by ***** Consultancy Services (TCS) and AWS, 88 percent of manufacturers anticipate AI will capture at least five percent of operating margin. One in four expect returns exceeding 10 percent. The money is there. The ambition is there. The plumbing, unfortunately, is not. A disparity exists between financial forecasts and the reality of the factory floor. While spending on intelligent systems accelerates, the underlying data infrastructure remains brittle, and risk management strategies still rely on expensive manual buffers. Pressure to extract value from AI for manufacturing The pressure to extract cash value from tech stacks has never been higher. 75 percent of respondents expect AI to rank as a top-three contributor to operating margins by 2026. Consequently, organisations are funneling 51 percent of their transformation spending toward AI and autonomous systems over the next two years. This spending eclipses other vital areas. Allocations for AI outpace workforce reskilling (19%) and cloud infrastructure modernisation (16%) by a wide margin. For CIOs, this imbalance signals a looming crisis: attempting to deploy advanced algorithms on shaky legacy foundations. Anupam Singhal, President of Manufacturing at TCS, said: “Manufacturing is an industry defined by precision, reliability, and the relentless pursuit of performance. Today, that strength of foundation becomes multifold with AI in orchestrating decisions—delivering transformational business outcomes through greater predictability, stability, and control. “At TCS, we see this as a defining opportunity to help manufacturers build resilient, adaptive, and future-ready enterprise ecosystems that can thrive in an era of intelligent autonomy.” Analogue hedges in a digital era Despite the heavy investment in predictive capabilities, operational behaviour betrays a lack of trust. When disruption hits, manufacturers aren’t leaning on the agility of their digital systems; they are reverting to physical safeguards. Following recent disruptions, 61 percent of organisations increased their safety stock. Half opted for multisourcing logistics. Only 26 percent utilised scenario planning via digital twins to navigate volatility. This is the disconnect. While AI promises dynamic inventory optimisation, a benefit cited by 49 percent of respondents, the prevailing instinct is to hoard inventory. Supply chain leaders are buying Ferraris but driving them like tractors. Bridging this gap requires moving from reactive safety measures to proactive and system-led responses. Ozgur Tohumcu, General Manager of Automotive and Manufacturing at AWS, commented: “Manufacturers today are facing unprecedented pressure—from tight margins to volatile supply chains and workforce gaps. At AWS, we are revolutionising manufacturing through AI-powered autonomous operations, shifting from manual, reactive processes to intelligent, self-optimising systems that operate at scale. “By embedding artificial intelligence into every layer of the operation and leveraging cloud-native architecture, manufacturers can move beyond simple automation to true autonomous decision-making where systems predict, adapt, and act independently with minimal human intervention. This enables not just faster response times, but fundamentally transforms operations with AI-driven predictability, resilience, and agility.” Infrastructure debt The primary obstacle to these financial returns isn’t the AI models; it’s the data they feed on. Only 21 percent of manufacturers claim to be “fully AI-ready” with clean, contextual, and unified data. The majority (61%) operate with partial readiness, struggling with inconsistent quality across different plants. This fragmentation creates data silos that prevent algorithms from accessing the enterprise-wide inputs necessary for accurate decision-making. Integration with legacy systems stands as the primary hurdle, cited by 54 percent of respondents. This “technical debt,” accumulated over decades of digitisation, makes it difficult to overlay modern autonomous agents on older operational technology. Security also bites. Security and governance concerns top the list of plant-level obstacles at 52 percent. In an environment where a cyber-physical breach can halt production or cause physical harm, the risk appetite for autonomous intervention remains low. The shift towards agentic AI in manufacturing Despite the headwinds, the industry is charging toward agentic AI (i.e. systems capable of making decisions with limited human oversight.) Seventy-four percent of manufacturers expect AI agents to manage up to half of routine production decisions by 2028. More immediately, 66 percent of organisations already allow – or plan to allow within 12 months – AI agents to approve routine work orders without human sign-off. This progression from “copilots” to independent agents capable of completing entire tasks fundamentally alters the workforce. While 89 percent of manufacturers expect AI-guided robotics to impact the workforce, the focus is on augmentation rather than displacement. Productivity gains are currently concentrated in knowledge-intensive roles. Quality inspectors (49%) and IT support staff (44%) are seeing the fastest gains. Traditional production roles like maintenance technicians (29%) lag behind. Adoption is following a pattern of cognitive augmentation before addressing physical coordination. As AI agents embed themselves across platforms, enterprise architects face a choice regarding orchestration. The market shows a strong aversion to vendor lock-in. 63 percent of manufacturers favour hybrid or multi-platform strategies over single-vendor solutions. Specifically, 33 percent plan to coordinate through multiple platform-native agents, while 30 percent prefer a hybrid model blending platform-native and custom orchestration. Only 13 percent are willing to anchor on a single foundational platform. Converting the manufacturing industry’s AI investment to profit To convert this massive capital outlay into actual profit, the C-suite needs to look past the hype. First, fix the data. With only 21 percent of firms fully ready, the immediate priority must be modernisation rather than algorithm development. Without clean, unified data, high-value use cases in sustainability and predictive maintenance will fail to scale. Second, leaders must bridge the AI trust gap. The reliance on safety stock indicates a lack of faith in digital signals. Staged autonomy is the answer—starting with administrative tasks like work orders, where 66 percent are already heading, before handing over complex supply chain decisions. Finally, avoid the monolithic trap. The data supports a multi-platform approach to maintain leverage and agility. Manufacturers are betting their future on AI, but realising those returns requires less focus on the “intelligence” of the models and more on the mundane work of cleaning data, integrating legacy equipment, and building workforce trust. See also: Frontier AI research lab tackles enterprise deployment challenges Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events including the Cyber Security Expo. Click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post AI in manufacturing set to unleash new era of profit appeared first on AI News. View the full article

The cybersecurity training provider Hack The Box (HTB) has launched the HTB AI Range, designed to let organisations test autonomous AI security agents under realistic conditions, albeit with oversight from human cybersecurity professionals. Its goal is to help users assess how well AI, and mixed human–AI teams might defend infrastructure. Vulnerabilities in AI models add to those already present in traditional IT, so before agentic or AI-based cybersecurity tools can be deployed in anger, HTB is proposing a testing environment where AI agents and human defenders can work together under realistic pressure to measure their cybersecurity prowess. How HTB AI Range works HTB describes the AI Range as a simulation of enterprise complexity with thousands of offensive and defensive targets that are continuously updated. The platform supports mapping to established cyber frameworks, including MITRE ATT&CK, the NIST/NICE guidelines, and the Open Worldwide Application Security Project (OWASP) Top 10. HTB says in a recent AI vs. human capture the flag (CTF) exercise, autonomous AI agents solved 19 out of 20 basic challenges. But in multi-step challenges in more complex environments, human teams outperformed the AI agents. The company suggests AI struggles with complexity and multi-stage operations, and this points to the continuing value of human expertise, especially in high-stakes or complex work. Testing, and closing the skills gap Enterprises can use the AI Range to validate whether existing security measures work under AI-powered attacks, give their cybersecurity teams experience of AI-powered threats, and develop more resilient cybersecurity tools based on agentic AI. Such exercises could be used to justify cybersecurity investment to financial decision-makers, Hack The Box suggests. HTB’s AI Range can be used for continuous testing and validation of cybersecurity defences, which the company states is more effective in the long-term than static audits or pen-testing exercises, and thus is closer to a CTEM model (continuous threat exposure management). HTB is launching a AI Red Teamer Certification early next year in an attempt quantify the skills necessary to harden AI defences. At present it seems wise to regard AI cyber-ranges as part of a layered security and resilience offering. As AI matures and frameworks like MITRE ATLAS gain traction, tools like HTB’s AI Range may become standard components in enterprise security programmes. “Hack The Box is where AI agents and humans learn to operate under real pressure together,” said Gerasimos Marketos, chief product officer at Hack The Box. “We’re addressing the urgent need to continuously validate AI systems in realistic operational contexts where stakes are high and human oversight remains vital. HTB AI Range makes that possible.” Haris Pylarinos, CEO and founder of Hack The Box said, “For over two years, we’ve been advancing AI-driven learning paths, labs, and research where machines and humans compete, collaborate, and co-evolve. With HTB AI Range, we’re not reacting to AI’s rise in cyber; we’re defining how defence evolves alongside it. This is how cybersecurity advances: not through fear, but through mastery.” (Image source: “The main cast” by Tim Dorr is licensed under CC BY-SA 2.0.) See also: New Nvidia Blackwell chip for China may outpace H20 model Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events, click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post HTB AI Range offers experiments in cyber-resilience training appeared first on AI News. View the full article

AI is moving deeper into the physical world, and EY is laying out a more structured way for companies to work with robots, drones, and other smart devices. The organisation is introducing a physical AI platform built with NVIDIA tools, opening a new EY.ai Lab in Georgia, and adding new leadership to guide its work in this field. The platform uses NVIDIA Omniverse libraries, NVIDIA Isaac, and NVIDIA AI Enterprise software. EY says the setup gives organisations a clearer way to plan, test, and manage AI systems that operate in real environments, from factory robots to drones and edge devices. Omniverse libraries support the creation of digital twins so firms can model and test systems before deployment. NVIDIA Isaac tools offer open models and simulation frameworks to design and validate AI-driven robots in detailed 3D settings. NVIDIA AI Enterprise provides the computing base needed to run heavier AI workloads. EY describes the platform as built around three main areas: AI-ready data: Synthetic data to mirror a wide range of physical scenarios. Digital twins and robotics training: Tools that connect digital and physical systems, monitor performance in real time, and support operational continuity. Responsible physical AI: Governance and controls that address safety, ethics, and compliance. The platform is meant to support everything from early planning to long-term maintenance in sectors like industrials, energy, consumer, and health. Raj Sharma, EY Global Managing Partner – Growth & Innovation, says physical AI is already “transforming how businesses in sectors operate and help create value,” saying that it brings more automation and can help lower operating costs. He says the combination of EY’s industry experience and NVIDIA’s infrastructure is expected to speed up how companies move “from experimentation to enterprise-scale deployment.” NVIDIA’s John Fanelli notes that more enterprises are bringing robots and automation into real settings to address workforce changes and improve safety. He says the EY.ai Lab, supported by NVIDIA AI infrastructure, helps organisations “simulate, optimise and safely deploy robotics applications at enterprise scale,” which he views as part of the next phase of industrial AI. New leadership and a dedicated physical AI lab EY has also appointed Dr. Youngjun Choi as its Global Physical AI Leader. He will oversee robotics and physical AI work and help shape EY’s role as an advisor in this area. Choi, who has nearly 20 years’ experience in robotics and AI, previously led the UPS Robotics AI Lab, where he worked on digital twins, robotics projects, and AI tools to modernise its network. Before that, he served as research faculty in Aerospace Engineering at the Georgia Institute of Technology, contributing to aerial robotics and autonomous systems. A key part of his role is directing the newly opened EY.ai Lab in Alpharetta, Georgia – the first EY site focused on physical AI. The Lab includes robotics systems, sensors, and simulation tools so organisations can test ideas and build prototypes before deploying them at scale. Joe Depa, EY Global Chief Innovation Officer, says his clients want better ways to use technology for decision-making and performance. He adds that physical AI requires strong data foundations and trust from the start. With Choi leading the Lab, Depa says EY teams are beginning to “get beyond the surface of what is possible” and set up the base for scalable operations. At the Lab, organisations can: Design and test physical AI systems in a virtual testbed, Build solutions for humanoids, quadrupeds, and other next-generation robots, Improve logistics, manufacturing, and maintenance with digital twins. The new platform and Lab build on earlier collaboration between EY and NVIDIA, including an AI agent platform launched earlier this year. Both organisations plan to expand their physical AI work to areas like energy, health, and smart cities. They also aim to support automation projects that cut waste and help reduce environmental impact. See also: Microsoft, NVIDIA, and Anthropic forge AI compute alliance Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events, click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post EY and NVIDIA to help companies test and deploy physical AI appeared first on AI News. View the full article

For years, cybersecurity experts debated when—not if—artificial intelligence would cross the threshold from advisor to autonomous attacker. That theoretical milestone has arrived. Anthropic’s recent investigation into a ******** state-sponsored operation has documented the first case of AI-orchestrated cyberattacks executing at scale with minimal human oversight, fundamentally altering what enterprises must prepare for in the threat landscape ahead. The campaign, attributed to a group Anthropic designates as GTG-1002, represents what security researchers have long warned about but never actually witnessed in the wild: an AI system autonomously conducting nearly every phase of cyber intrusion—from initial reconnaissance to data exfiltration—while human operators merely supervised strategic checkpoints. This isn’t incremental evolution. It’s a categorical shift in offensive capabilities that compresses what would take skilled hacking teams weeks into operations measured in hours, executed at machine speed across dozens of targets simultaneously. The numbers tell the story. Anthropic’s forensic analysis revealed that 80 to 90% of GTG-1002’s tactical operations ran autonomously, with humans intervening at just four to six critical decision points per campaign. The operation targeted approximately 30 entities—major technology corporations, financial institutions, chemical manufacturers, and government agencies—achieving confirmed breaches of several high-value targets. At peak activity, the AI system generated thousands of requests at rates of multiple operations per second, a tempo physically impossible for human teams to sustain. Anatomy of an autonomous breach The technical architecture behind these AI-orchestrated cyberattacks reveals a sophisticated understanding of both AI capabilities and safety bypass techniques. GTG-1002 built an autonomous attack framework around Claude Code, Anthropic’s coding assistance tool, integrated with Model Context Protocol (MCP) servers that provided interfaces to standard penetration testing utilities—network scanners, database exploitation frameworks, password crackers, and binary analysis suites. The breakthrough wasn’t in novel malware development but in orchestration. The attackers manipulated Claude through carefully constructed social engineering, convincing the AI it was conducting legitimate defensive security testing for a cybersecurity firm. They decomposed complex multi-stage attacks into discrete, seemingly innocuous tasks—vulnerability scanning, credential validation, data extraction—each appearing legitimate when evaluated in isolation, preventing Claude from recognising the broader malicious context. Once operational, the framework demonstrated remarkable autonomy. In one documented compromise, Claude independently discovered internal services within a target network, mapped complete network topology across multiple IP ranges, identified high-value systems including databases and workflow orchestration platforms, researched and wrote custom exploit code, validated vulnerabilities through callback communication systems, harvested credentials, tested them systematically across discovered infrastructure, and analyzedstolen data to categorize findings by intelligence value—all without step-by-step human direction. The AI maintained a persistent operational context across sessions spanning days, enabling campaigns to resume seamlessly after interruptions. It made autonomous targeting decisions based on discovered infrastructure, adapted exploitation techniques when initial approaches failed, and generated comprehensive documentation throughout all phases—structured markdown files tracking discovered services, harvested credentials, extracted data, and complete attack progression. What this means for enterprise security The GTG-1002 campaign dismantles several foundational assumptions that have shaped enterprise security strategies. Traditional defences calibrated around human attacker limitations—rate limiting, behavioural anomaly detection, operational tempo baselines—face an adversary operating at machine speed with machine endurance. The economics of cyberattacks have shifted dramatically, as 80-90% of tactical work can be automated, potentially bringing nation-state-level capabilities within reach of less sophisticated threat actors. Yet AI-orchestrated cyberattacks face inherent limitations that enterprise defenders should understand. Anthropic’s investigation documented frequent AI hallucinations during operations—Claude claiming to have obtained credentials that didn’t function, identifying “critical discoveries” that proved to be publicly available information, and overstating findings that required human validation. These reliability issues remain a significant friction point for fully autonomous operations, though assuming they’ll persist indefinitely would be dangerously naive as AI capabilities continue advancing. The defensive imperative The dual-use reality of advanced AI presents both challenge and opportunity. The same capabilities enabling GTG-1002’s operation proved essential for defence—Anthropic’s Threat Intelligence team relied heavily on Claude to analyse the massive data volumes generated during their investigation, demonstrating how AI augments human analysts in detecting and responding to sophisticated threats. For enterprise security leaders, the strategic priority is clear: active experimentation with AI-powered defence tools across SOC automation, threat detection, vulnerability assessment, and incident response. Building organisational experience with what works in specific environments—understanding AI’s strengths and limitations in defensive contexts—becomes critical before the next wave of more sophisticated autonomous attacks arrives. Anthropic’s disclosure signals an inflexion point. As AI models advance and threat actors refine autonomous attack frameworks, the question isn’t whether AI-orchestrated cyberattacks will proliferate across the threat landscape—it’s whether enterprise defences can evolve rapidly enough to counter them. The window for preparation, while still open, is narrowing faster than many security leaders may realise. The post Anthropic just revealed how AI-orchestrated cyberattacks actually work—Here’s what enterprises need to know appeared first on AI News. View the full article

As AI adoption has continued to surge over the last couple of months, one thing has become abundantly clear, i.e. there isn’t enough computational horsepower to go around (something that has become painfully obvious as cloud providers have accrued months-long waitlists for high-end GPU instances). And, unlike the brief crypto-mining GPU craze from just a few years ago, today’s crunch is being driven by real demand from AI research and deployments. For perspective sake, Amazon Web Services has been charging about $98 per hour for an 8-GPU server loaded with Nvidia’s top-tier H100 chips, while some decentralized GPU platforms offer comparable hardware for as little as $3 an hour. Amidst this stark 30× price gap, Singularity Compute, the infrastructure arm of decentralized AI pioneer SingularityNET, has announced the phase I deployment of its first enterprise-grade NVIDIA GPU cluster at a state-of-the-art data center in Sweden. Under a partnership with Swedish operator Conapto, Singularity’s cluster is using cutting-edge NVIDIA hardware (including the next-generation H200 and L40S GPUs) in a Stockholm facility powered entirely by renewable energy. What’s on offer exactly? The cluster, which has been made to be high density by design, serves as the foundation for both traditional enterprise workloads and the projects of the Artificial Superintelligence (ASI) Alliance, a decentralized AI ecosystem spearheaded by SingularityNET. It offers flexible access modes that mirror the needs of modern AI developers wherein companies can rent whole machines on bare metal, spin up GPU-powered virtual machines, or even tap into dedicated API endpoints for AI inference. In real world terms what this means is that an organization can potentially train entire large machine learning models from scratch, fine-tune existing models on custom datasets, or run heavy-duty inference for applications like generative AI, all using Singularity’s infrastructure. On the operational front, it bears mentioning that the partnership is set to be managed by popular cloud provider and NVIDIA partner Cudo Compute, with the latter ensuring the cluster’s timely delivery of enterprise-grade reliability and support that mission-critical AI projects demand. On the entire development, Dr. Ben Goertzel, founder of SingularityNET and co-chair of the ASI Alliance, opined: “As AI accelerates toward AGI and beyond, access to high-performance, ethically aligned compute is becoming a defining factor in who shapes the future. We need powerful compute that is configured for interoperation with decentralized networks running a rich variety of AI algorithms carrying out tasks for diverse populations. The new GPU deployment in Sweden is a meaningful milestone on the road to a truly open, global Artificial Superintelligence.” A similar sentiment was echoed by Singularity Compute CEO Joe Honan who believes the launch is about more than just extra compute capacity but rather a step toward a new paradigm in AI infrastructure, emphasizing that the cluster’s NVIDIA GPUs will deliver the performance and reliability modern AI demands, while upholding principles of openness, security, and sovereignty in how the compute is provisioned. In this broader context, it also bears mentioning that the Swedish cluster is set to serve as the backbone for ASI:Cloud, Singularity’s new AI model inference service developed in collaboration with Cudo. To elaborate, ASI:Cloud provides developers with wallet-based access to an OpenAI-compatible API for model inference, offering a smooth path to scale from serverless functions up to dedicated GPU servers. Early customers are already being onboarded to the Swedish cluster, with the team hinting that this is only the beginning of additional hardware and new geographic locations entering the fray. Thus, for a community that has often been at the bleeding edge of the ongoing AI and blockchain revolution, this deployment seems to be a tangible step toward the long-held goal of a decentralized, globally distributed AI infrastructure. The race for AI compute is underway and heating up fast Since the turn of the decade, the tech sector has poured major investments into AI infrastructure, with 2025 alone having witnessed over $1 trillion in new AI-focused data center projects. Even nation-states seem to be wading in with France, for example, having unveiled a surprise €100+ billion plan to boost AI infrastructure. Yet not everyone can spend billions to solve the current compute shortage, resulting in emergence of alternate approaches like decentralized or distributed GPU networks (that can tap into hardware spread across many locations and operators). In other words, if the 2010s rewarded those who accumulated data, the 2020s will seemingly reward those who control compute power. Within that future, efforts like Singularity Compute’s new GPU cluster embody a growing determination to democratize who gets to shape AI’s next chapter (primarily by broadening where the compute behind it is coming from). Interesting times ahead. The post Amidst the Ongoing AI Infrastructure Crunch, Singularity Compute Launches Swedish GPU Cluster appeared first on AI News. View the full article

Chip stacking strategy is emerging as China’s innovative response to US semiconductor restrictions, but can this approach truly close the performance gap with Nvidia’s advanced GPUs? As Washington tightens export controls on cutting-edge chipmaking technology, ******** researchers are proposing a bold workaround: stack older, domestically-producible chips together to match the performance of chips they can no longer access. The Core Concept: Building upward instead of forward The chip stacking strategy centres on a deceptively simple premise—if you can’t make more advanced chips, make smarter systems with the chips you can produce. Wei Shaojun, vice-president of the China Semiconductor Industry Association and a professor at Tsinghua University, recently outlined to the South China Morning Post an architecture that combines 14-nanometer logic chips with 18-nanometer DRAM using three-dimensional hybrid bonding. This matters because US export controls specifically target the production of logic chips at 14nm and below, and DRAM at 18nm and below. Wei’s proposal works precisely at these technological boundaries, using processes that remain accessible to ******** manufacturers. The technical approach involves what’s called “software-defined near-memory computing.” Instead of shuffling data back and forth between processors and memory—a major bottleneck in AI workloads—this chip stacking strategy places them in intimate proximity through vertical stacking. The 3D hybrid bonding technique creates direct copper-to-copper connections at sub-10 micrometre pitches, essentially eliminating the physical distance that slows down conventional chip architectures. The performance claims and reality check Wei claims this configuration could rival Nvidia’s 4nm GPUs while significantly reducing costs and power consumption. He’s cited performance figures of 2 TFLOPS per watt and a total of 120 TFLOPS. There’s just one problem: Nvidia’s A100 GPU, which Wei positions as the comparison point, actually delivers up to 312 TFLOPS—more than 2.5 times the claimed performance. This discrepancy highlights a critical question about the chip stacking strategy’s feasibility. While the architectural innovation is real, the performance gaps remain substantial. Stacking older chips doesn’t magically erase the advantages of advanced process nodes, which deliver superior power efficiency, higher transistor density, and better thermal characteristics. Why China is betting on this approach The strategic logic behind the chip stacking strategy extends beyond pure performance metrics. Huawei founder Ren Zhengfei has articulated a philosophy of achieving “state-of-the-art performance by stacking and clustering chips rather than competing node for node.” This represents a fundamental shift in how China approaches the semiconductor challenge. Consider the alternatives. TSMC and Samsung are pushing toward 3nm and 2nm processes that remain completely out of reach for ******** manufacturers. Rather than fighting an unwinnable battle for process node leadership, the chip stacking strategy proposes competing on system architecture and software optimisation instead. There’s also the CUDA problem. Nvidia’s dominance in AI computing rests not just on hardware but on its CUDA software ecosystem. Wei describes this as a “triple dependence” spanning models, architectures, and ecosystems. ******** chip designers pursuing traditional GPU architectures would need to either replicate CUDA’s functionality or convince developers to abandon a mature, widely adopted platform. The chip stacking strategy, by proposing an entirely different computing paradigm, offers a path to sidestep this dependency. The feasibility question Can the chip stacking strategy actually work? The technical foundations are sound—3D chip stacking is already used in high-bandwidth memory and advanced packaging solutions worldwide. The innovation lies in applying these techniques to create entirely new computing architectures rather than simply improving existing designs. However, several challenges loom large. First, thermal management becomes exponentially more difficult when stacking multiple active processing dies. The heat generated by 14nm chips is considerably higher than modern 4nm or 5nm processes, and stacking intensifies this problem. Second, yield rates in 3D stacking are notoriously difficult to optimise—a defect in any layer can compromise the entire stack. Third, the software ecosystem required to efficiently utilise such architectures doesn’t exist yet and would take years to mature. The most realistic assessment is that the chip stacking strategy represents a valid approach for specific workloads where memory bandwidth matters more than raw computational speed. AI inference tasks, certain data analytics operations, and specialised applications could potentially benefit. But matching Nvidia’s performance across the full spectrum of AI training and inference tasks remains a distant goal. What this means for the AI chip wars The emergence of the chip stacking strategy as a focal point for ******** semiconductor development signals a strategic pivot. Rather than attempting to replicate Western chip designs with inferior process nodes, China is exploring architectural alternatives that play to available manufacturing strengths. Whether this chip stacking strategy succeeds in closing the performance gap with Nvidia remains uncertain. What’s clear is that China’s semiconductor industry is adapting to restrictions by pursuing innovation in areas where export controls have less impact—system design, packaging technology, and software-hardware co-optimisation. For the global AI industry, this means the competitive landscape is becoming more complex. Nvidia’s current dominance faces challenges not just from traditional competitors like AMD and Intel, but from entirely new architectural approaches that may redefine what an “AI chip” looks like. The chip stacking strategy, whatever its current limitations, represents exactly this kind of architectural disruption—and that makes it worth watching closely. See also: New Nvidia Blackwell chip for China may outpace H20 model Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events, click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post Can China’s chip stacking strategy really challenge Nvidia’s AI dominance? appeared first on AI News. View the full article

Thomson Reuters and Imperial College London have established a frontier AI research lab to overcome historic deployment challenges. Speed and scale have defined the current AI *****. But for enterprises, the primary obstacles to deployment are different: trust, accuracy, and lineage. Addressing these barriers, Thomson Reuters and Imperial College London have announced a five-year partnership to establish a joint ‘Frontier AI Research Lab’. With the involvement of both a corporate and academic leader, the initiative appears built to target the disconnect between high-level computer science and the pragmatic requirements of professional services. The lab will pursue academic research in AI, focusing on safety, reliability, and the development of frontier capabilities. It offers enterprise leaders a preview of how future systems might advance beyond generative text to perform reliable work in high-stakes environments. Improving reliability with practical frontier AI research Current Large Language Models (LLMs) often struggle with the precision required in sectors such as law, tax, and compliance. To counter this, the lab plans to train large-scale foundation models jointly. This is an opportunity typically restricted to a handful of industrial technology giants. Researchers will experiment with data-centric machine learning and retrieval-augmented generation using Thomson Reuters’ substantial repository of content. By grounding AI models in verified and domain-specific data, the initiative aims to greatly improve the algorithms used to drive positive impact in the wider world and address challenges prior to real-world deployment. Dr Jonathan Richard Schwarz, Head of AI Research at Thomson Reuters, said: “We are only beginning to understand the transformative impact this technology will have on all aspects of society. “Our vision is a unique research space where foundational algorithms are developed and made available to world experts, advancing the transparency, verifiability, and trustworthiness in which these changes are driving impact in the world.” Data provenance is the central theme here. As Dr Schwarz suggests, the value lies not merely in the model architecture but in the quality of the information it processes. The partnership creates an avenue for researchers to access high-quality data spanning complex and knowledge-intensive domains. Making enterprise AI deployment challenges history The lab’s frontier AI research agenda indicates where enterprise technology is heading. Beyond simple content generation, the facility will investigate agentic AI systems, reasoning, planning, and human-in-the-loop workflows. These areas are essential for organisations looking to automate multi-step processes rather than just discrete tasks. Professor Alessandra Russo, who will co-lead the lab alongside Dr Schwarz and Cambridge’s Professor Felix Steffek, believes the dedicated infrastructure will empower researchers to deliver scientific advances that have practical relevance. “With dedicated space, a focused PhD cohort, and high-quality computing infrastructure and support, our researchers will be empowered to push the boundaries of AI and deliver scientific advances that truly matter,” Professor Russo stated. “Our collaboration with Thomson Reuters anchors that work in real-world use cases, ensuring that breakthroughs translate into meaningful societal benefit. There is huge potential to unlock creative approaches to a wide range of roles and sectors, enabling AI to strengthen society, energise traditional industries, and create new roles and opportunities across the economy.” Operations leaders should note that future AI implementations will likely require robust “reasoning” capabilities (i.e. the ability for a system to plan a series of actions and verify its own outputs) before they can be trusted with autonomous decision-making in regulated industries. Boosting infrastructure and talent pipelines to advance frontier AI research Running these experiments requires substantial compute power, a resource often lacking in purely academic settings. The partnership addresses this by providing researchers access to Imperial’s high-performance computing cluster. This enables AI experiments at a meaningful scale to uncover any challenges that need to be overcome prior to real-world deployment. The setup creates a feedback loop between research and practice. The lab is planned to host over a dozen PhD students who will work alongside Thomson Reuters foundational research scientists. This structure accelerates the translation of research into practice and establishes a direct pipeline for talent development and real-world validation. Professor Mary Ryan, Vice Provost for Research and Enterprise at Imperial, commented: “This collaboration gives our researchers the space and support to explore fundamental questions about how AI can and should work for society. “Progress in this area depends on rigorous science, open inquiry, and strong partnerships—ideals exemplified by the approach this lab will take.” Overcoming legal and economic challenges for successful enterprise AI deployments The risks associated with AI are as much legal and economic as they are technical. Recognising this, the lab’s steering committee includes Professor Felix Steffek, a Professor of Law at the University of Cambridge. “AI has great potential to improve access to justice,” said Professor Steffek. “However, there are significant challenges that foundational research needs to address in order to make legal AI applications safe and ethically responsible. “The lab will bring together bright minds from multiple disciplines – including law, ethics, and AI – to advance the potential and address the risks of legal AI.” The scope of research extends to the technology’s broader economic impact and the future of work. The lab aims to produce insights on how AI can energise traditional industries and create new roles across the economy. Overall, the Frontier AI Research Lab represents a model for de-risking enterprise AI strategies and overcoming challenges that have historically held back deployments. Coupling industrial data and compute resources with academic rigour helps organisations understand the “****** box” nature of these systems and overcome the challenges to ensure the success of any deployment. Activities at the lab will commence upon formal launch, starting with the recruitment of the initial PhD cohort. Business leaders should track the joint publications coming out of this unit as these findings will likely serve as valuable benchmarks for evaluating the safety and efficacy of internal AI deployments. See also: Agentic AI autonomy grows in North American enterprises Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events including the Cyber Security Expo. Click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post Frontier AI research lab tackles enterprise deployment challenges appeared first on AI News. View the full article

Enterprise leaders are entering 2026 with an uncomfortable mix of volatility, optimism, and pressure to move faster on AI and quantum computing, according to a paper published by the IBM Institute for Business Value. Its findings are based on more than 1,000 C-suite executives and 8,500 employees and consumers. While only around a third of executives are optimistic about the global economy, more than four in five are confident about their own organisation’s performance in the year ahead. Executives expect to make faster decisions and are willing to redesign operating models, while employees are broadly positive about AI in their working lives. Customers, in turn, are ready to reward (or punish) brands based on how companies use their data. Trend 1: agentic AI a strategic asset Agentic AI is emerging as one of the main tools leaders expect to use in the coming year, and most execs say AI agents are already helping them. However, for agentic AI to succeed, the expressed opinions state: Data architecture needs to support near real-time insight, not periodic reporting. AI agents’ success will depend on access to core systems (ERP, CRM, supply chain platforms). Agentic AI shifts from experimental to operational. Leaders feel they must decide which decisions can be delegated to AI agents, which require human review, and should must remain human-led. Trend 2: employees will ask for more training and AI is okay Most employees say the pace of technology change in their roles is sustainable, and that they’re confident about keeping up with new tools. Twice as many employees say they would embrace, not resist, greater use of AI in the workplace, seeing the technology as a way to remove repetitive tasks and learn new skills. This aligns with findings in research by KPMG. Executives expect a significant re-skilling requirement from their employees, so leaders should anticipate that at least half their workforce will need some form of re-skilling by the end of 2026, thanks to AI automation. Other surveys concur with IBM, and state the skills needed most are problem-solving, creativity, and innovation. Employees say they are willing to change employers to access better training opportunities, meaning skills development now plays a direct role reducing employee churn. Trend 3: customers will hold data policies to account The executives surveyed agreed that consumer trust in a brand’s use of AI will define the success of new products and services. Consumers are willing to tolerate occasional errors, but not opacity. Customers want explanations of how their data is used, knowledge of when AI is involved in interactions with them, and simple ways to opt in or out. The studies by Deloitte and KPMG (see above) reinforce this picture. Implications for leaders include treating transparency as a product feature and selecting models that support explainability. Trend 4: AI and cloud will need local provision AI sovereignty—an organisation’s ability to control and govern its AI systems, data, and infrastructure—has moved to the centre of resilience planning. Almost all executives surveyed said they will factor AI sovereignty into their 2026 strategy. In the light of concerns about data residency and cloud jurisdiction, leaders are rethinking where models run and where data lives. Studies from *** and European IT leaders show rising concern about over-reliance on foreign (read, ‘US-based’ cloud services in the latter case). Advisory firm Accenture also urges leaders [PDF] to develop sovereign AI strategies that prioritise control, transparency, and choice. Key takeaways include the need for portable AI platforms, monitoring for data compliance, and a heavy emphasis on the physical location of data. AI resilience is ultimately about continuity and transparency. It requires ensuring the organisation can adapt and operate openly, even when the global technological and geopolitical landscapes shift. Trend 5: planning on quantum advantage The report’s findings say quantum is moving towards experimentation in the near term. IBM’s own research on quantum readiness (in line with its monetisation of quantum services) suggests that early quantum advantage is likely in targeted domains such as optimisation and materials science. The report urges the identification of small numbers of high-impact quantum uses in the enterprise, and the joining of ecosystems early. “Identify big bets to win with emerging technologies, including quantum, and partner on innovation to share costs,” the report states. (Image source: “California Perfect” by moonjazz is licensed under CC BY-SA 2.0.) See also: How the MCP spec update boosts security as infrastructure scales Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events including the Cyber Security Expo. Click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post IBM cites agentic AI, data policies, and quantum as 2026 trends appeared first on AI News. View the full article

While tech giants pour billions into computational power to train frontier AI models, China’s DeepSeek has achieved comparable results by working smarter, not harder. The DeepSeek V3.2 AI model matches OpenAI’s GPT-5 in reasoning benchmarks despite using ‘fewer total training FLOPs’ – a breakthrough that could reshape how the industry thinks about building advanced artificial intelligence. For enterprises, the release demonstrates that frontier AI capabilities need not require frontier-scale computing budgets. The open-source availability of DeepSeek V3.2 lets organisations evaluate advanced reasoning and agentic capabilities while maintaining control over deployment architecture – a practical consideration as cost-efficiency becomes increasingly central to AI adoption strategies. The Hangzhou-based laboratory released two versions on Monday: the base DeepSeek V3.2 and DeepSeek-V3.2-Speciale, with the latter achieving gold-medal performance on the 2025 International Mathematical Olympiad and International Olympiad in Informatics – benchmarks previously reached only by unreleased internal models from leading US AI companies. The accomplishment is particularly significant given DeepSeek’s limited access to advanced semiconductor chips due to export restrictions. Resource efficiency as a competitive advantage DeepSeek’s achievement contradicts the prevailing industry assumption that frontier AI performance requires greatly scaling computational resources. The company attributes this efficiency to architectural innovations, particularly DeepSeek Sparse Attention (DSA), which substantially reduces computational complexity while preserving model performance. The base DeepSeek V3.2 AI model achieved 93.1% accuracy on AIME 2025 mathematics problems and a Codeforces rating of 2386, placing it alongside GPT-5 in reasoning benchmarks. The Speciale variant was even more successful, scoring 96.0% on the American Invitational Mathematics Examination (AIME) 2025, 99.2% on the Harvard-MIT Mathematics Tournament (HMMT) February 2025, and achieving gold-medal performance on both the 2025 International Mathematical Olympiad and International Olympiad in Informatics. The results are particularly significant given DeepSeek’s limited access to the raft of tariffs and export restrictions affecting China. The technical report reveals that the company allocated a post-training computational budget exceeding 10% of pre-training costs – a substantial investment that enabled advanced abilities through reinforcement learning optimisation rather than brute-force scaling. Technical innovation driving efficiency The DSA mechanism represents a departure from traditional attention architectures. Instead of processing all tokens with equal computational intensity, DSA employs a “lightning indexer” and a fine-grained token selection mechanism that identifies and processes only the most relevant information for each query. The approach reduces core attention complexity from O(L²) to O(Lk), where k represents the number of selected tokens – a fraction of the total sequence length L. During continued pre-training from the DeepSeek-V3.1-Terminus checkpoint, the company trained DSA in 943.7 billion tokens using 480 sequences of 128K tokens per training step. The architecture also introduces context management tailored for tool-calling scenarios. Unlike previous reasoning models that discarded thinking content after each user message, the DeepSeek V3.2 AI model retains reasoning traces when only tool-related messages are appended, improving token efficiency in multi-turn agent workflows by eliminating redundant re-reasoning. Enterprise applications and practical performance For organisations evaluating AI implementation, DeepSeek’s approach offers concrete advantages beyond benchmark scores. On Terminal Bench 2.0, which evaluates coding workflow capabilities, DeepSeek V3.2 achieved 46.4% accuracy. The model scored 73.1% on SWE-Verified, a software engineering problem-solving benchmark, and 70.2% on SWE Multilingual, demonstrating practical utility in development environments. In agentic tasks requiring autonomous tool use and multi-step reasoning, the model showed significant improvements over previous open-source systems. The company developed a large-scale agentic task synthesis pipeline that generated over 1,800 distinct environments and 85,000 complex prompts, enabling the model to generalise reasoning strategies to unfamiliar tool-use scenarios. DeepSeek has open-sourced the base V3.2 model on Hugging Face, letting enterprises implement and customise it without vendor dependencies. The Speciale variant remains accessible only through API due to higher token use requirements – a trade-off between maximum performance and deployment efficiency. Industry implications and acknowledgement The release has generated substantial discussion in the AI research community. Susan Zhang, principal research engineer at Google DeepMind, praised DeepSeek’s detailed technical documentation, specifically highlighting the company’s work stabilising models post-training and enhancing agentic capabilities. The timing ahead of the Conference on Neural Information Processing Systems has amplified attention. Florian Brand, an expert on China’s open-source AI ecosystem attending NeurIPS in San Diego, noted the immediate reaction: “All the group chats today were full after DeepSeek’s announcement.” Acknowledged limitations and development path DeepSeek’s technical report addresses current gaps compared to frontier models. Token efficiency remains challenging – the DeepSeek V3.2 AI model typically requires longer generation trajectories to match the output quality of systems like Gemini 3 Pro. The company also acknowledges that the breadth of world knowledge lags behind leading proprietary models due to lower total training compute. Future development priorities include scaling pre-training computational resources to expand world knowledge, optimising reasoning chain efficiency to improve token use, and refining the foundation architecture for complex problem-solving tasks. See also: AI business reality – what enterprise leaders need to know Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events, click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post China’s DeepSeek V3.2 AI model achieves frontier performance on a fraction of the computing budget appeared first on AI News. View the full article

Thrive Holdings’ push to modernise accounting and IT services is entering a new stage, as OpenAI prepares to take an ownership stake in the company and place its own specialists inside Thrive’s businesses. In doing so, OpenAI is testing an AI-driven model that pairs capital, sector expertise, and embedded technical teams. Thrive started its holding company earlier this year to buy and manage firms in day-to-day service industries. Its aim has been to rebuild these companies with more efficient processes, new data practices, and practical uses of AI. OpenAI’s deeper involvement now turns that idea into a real-time experiment in how traditional providers can update their work without relying only on off-the-shelf tools. A test case for bringing AI into core operational work While most enterprise discussions about AI tend to revolve around pilots and proof-of-concepts, Thrive is taking a different approach: buying companies outright and redesigning how they run. Its two current businesses – Crete Professionals Alliance (accounting) and Shield Technology Partners (IT services) – employ more than 1,000 people. Thrive has committed $500 million to Crete and, together with ZBS Partners, more than $100 million to Shield. For companies watching from the outside, the appeal is clear. These industries carry heavy workloads, manual tasks, and tight margins. They also handle sensitive data and operate under strict deadlines. Any AI system introduced into that environment needs domain context, training, and adjustments that fit local processes – not generic automation. Crete has already begun using AI to cut down routine tasks like data entry and early-stage tax workflows. Shield is on track to complete 10 acquisitions by the end of the year, giving Thrive a base of IT operations which it intends to redesign with new tools and methods. What OpenAI gains OpenAI is under pressure to find real, enterprise-scale use cases for its models. Investors value the company at roughly $500 billion, and its long-term commitments include about $1.4 trillion in infrastructure spending through 2033. To justify those figures, it is betting that businesses will spend heavily on tools that help them work faster and handle complex tasks at volume. By taking a stake in Thrive Holdings, OpenAI gains something it cannot produce on its own: access to companies where it can experience models in day-to-day working, and training specialists on real operations. The more Thrive’s companies grow, the more OpenAI’s stake may expand, according to a person familiar with the deal. Joshua Kushner, founder of both Thrive Capital and Thrive Holdings, said, “We are excited to extend our partnership with OpenAI to embed their frontier models, products, and services into sectors we believe have tremendous potential to benefit from technological innovation and adoption.” The partnership also gives OpenAI a path to collect value from the engineering support it provides. Its team will develop custom models for Thrive’s companies and embed researchers and engineers on site, according to partner Anuj Mehndiratta, who oversees product and technology strategy at Thrive Holdings. What enterprises can learn from this approach For many companies, the hardest part of using AI is not the model but the redesign of existing work. Thrive’s strategy reflects a shift toward deeper integration, where AI teams sit inside the business units they support rather than acting as external advisers. The model lets companies: Build tools shaped around real workflows, not abstract use cases Train models on controlled, high-quality data Reduce the gap between engineering teams and front-line employees Test changes faster, with direct feedback from staff It also surfaces the real cost of AI adoption. Custom work requires engineering time, domain knowledge, and long-term alignment between owners and model developers. Thrive’s partnership with OpenAI formalises that alignment in a way that may become more common as enterprises look for results rather than demonstrations. Brad Lightcap, OpenAI’s COO, said, “The partnership with Thrive Holdings is about demonstrating what’s possible when frontier AI research and deployment are rapidly deployed in entire organisations to revolutionise how businesses work and engage with customers.” The wider competitive landscape The deal lands at a time when AI companies are trying to anchor themselves inside major enterprise accounts. Anthropic is reaching more businesses through Microsoft partnerships, and. Google is drawing interest with its latest model and has seen its market value rise as companies explore new AI options. OpenAI, meanwhile, has taken stakes in partners like AMD and CoreWeave to support its long-term infrastructure needs. OpenAI also expanded its reach on Monday this week, announcing a separate agreement with Accenture. Its ChatGPT Enterprise product will be rolled out to “tens of thousands” of Accenture employees, giving OpenAI another route into large-scale corporate use. A possible blueprint If Thrive’s companies show meaningful improvement in how they operate, the model could influence how other enterprises think about AI transformation. Rather than layering tools on top of old processes, some may move toward deeper restructuring, guided by technical teams that understand both the model and the business. For now, Thrive Holdings serves as a live case study of what that approach looks like when applied to industries that rarely make tech headlines but form the backbone of day-to-day business operations. See also: AI business reality – what enterprise leaders need to know Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events, click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post How OpenAI and Thrive are testing a new enterprise AI model appeared first on AI News. View the full article

North American enterprises are now actively deploying agentic AI systems intended to reason, adapt, and act with complete autonomy. Data from Digitate’s three-year global programme indicates that, while adoption is universal across the board, regional maturity paths are diverging. North American firms are scaling toward full autonomy, whereas their European counterparts are prioritising governance frameworks and data stewardship to build long-term resilience. From utility to profitability The story of enterprise automation has changed. In 2023, the primary objective for most IT leaders was cost reduction and the streamlining of routine tasks. By 2025, the focus has expanded. AI is no longer viewed solely as an operational utility but as a capability enabling profit. Data supports this change in perspective. The report indicates that North American organisations are seeing a median return on investment (ROI) of $175 million from their implementations. Interestingly, this financial validation is not unique to the fast-moving North American market. European enterprises, despite a more measured and governance-heavy approach, report a comparable median ROI of approximately $170 million. This consistency suggests that while deployment strategies differ, with Europe focusing on risk management and North America on speed, the financial outcomes are similar. Every organisation surveyed confirmed implementing AI within the last two years, utilising an average of five distinct tools. While generative AI remains the most widely deployed at 74 percent, there is a notable rise in “agentic” capabilities. Over 40 percent of enterprises have introduced agentic or agent-based AI, advancing beyond static automation toward systems that can manage goal-oriented workflows. IT operations autonomy becomes the proving ground for agentic AI While marketing and customer service often dominate public discourse regarding AI, the IT function itself has emerged as the primary laboratory for these deployments. IT environments are inherently data-rich and structured, creating ideal conditions for models to learn, yet they remain dynamic enough to require the adaptive reasoning that agentic AI systems promise. This explains why 78 percent of respondents have deployed AI within IT operations, the highest rate of any business function. Cloud visibility and cost optimisation lead the adoption curve at 52 percent, followed closely by event management at 48 percent. In these scenarios, the technology is not alerting humans to problems so much as actively interpreting telemetry data to provide a unified view of spending across hybrid environments. Teams leveraging these tools report improvements in decision accuracy (44%) and efficiency (43%), allowing them to handle higher workloads without a corresponding increase in escalations. The cost-human conundrum Despite the optimism surrounding ROI, the report highlights a “cost-human conundrum” that threatens to stall progress. The paradox is straightforward: enterprises deploy AI to reduce reliance on human labour and operational costs, yet those exact factors act as the primary inhibitors to growth. 47 percent of respondents cite the continued need for human intervention as a major drawback. Far from achieving the complete autonomy of “set and forget” solutions, these agentic AI systems require ongoing oversight, tuning, and exception management. Simultaneously, the cost of implementation ranks as the second-highest concern at 42 percent, driven by the expenses associated with model retraining, integration, and cloud infrastructure. The talent required to manage these costs is in short supply. A lack of technical skills remains the primary obstacle to further adoption for 33 percent of organisations. Demand for professionals capable of developing, monitoring, and governing these complex systems exceeds current supply, creating a self-reinforcing loop where investment increases operational capacity but simultaneously raises human and financial dependencies. Trust and perception gap A divergence in perspective exists between executive leadership and operational practitioners. While 94 percent of total respondents express trust in AI, this confidence is not distributed evenly. C-suite leaders are markedly more optimistic, with 61 percent classifying AI as “very trustworthy” and viewing it primarily as a financial lever. Only 46 percent of non-C-suite practitioners share this high level of trust. Those closer to the daily operation of these models are more acutely aware of reliability issues, transparency deficits, and the necessity for human oversight. This gap suggests that while leadership focuses on long-term overhaul and autonomy, teams on the ground are grappling with pragmatic delivery and governance challenges. There is also a mixed view on how these agents will function. 61 percent of IT leaders view agentic systems not as replacements, but as collaborators that augment human capability. However, the expectation of automation varies by industry. In retail and transport, 67 percent believe agentic AI will alter the essential tasks of their roles, while in manufacturing, the same percentage views these agents primarily as personal assistants. Complete agentic AI autonomy is rapidly approaching The industry anticipates a rapid progression toward reduced human involvement in routine processes. Currently, 45 percent of organisations operate as semi- to fully-autonomous enterprises. Projections indicate this figure will rise to 74 percent by 2030. This evolution implies a change in the role of IT. As capabilities mature, IT departments are expected to transition from being operational enablers to acting as orchestrators. In this model, the IT function manages the “system of systems,” ensuring that various intelligent agents interact correctly while humans focus on creativity, interpretation, and governance rather than execution. “Agentic AI is the bridge between human ingenuity and autonomous intelligence that marks the dawn of IT as a profit-driving, strategic capability,” notes Avi Bhagtani, CMO at Digitate. “Enterprises have moved from experimenting with automation to scaling AI for measurable impact.” The transition to agentic AI requires more than just software procurement; it demands an organisational philosophy that balances automation with human augmentation. Policies alone are insufficient; governance must be integrated directly into system design to ensure transparency and ethical oversight in every decision loop. European organisations are currently leading in this area, prioritising ethical deployment and strong oversight frameworks as a foundation for resilience. Furthermore, the shortage of technical talent cannot be solved by hiring alone. Organisations must invest in upskilling existing teams, combining operations expertise with data science and compliance literacy. Finally, reliable autonomy depends on high-quality data. Investments in data integration and observability platforms are necessary to provide agents with the context required to act independently. The era of experimental AI has passed. The current phase is defined by the pursuit of autonomy, where value is derived not from novelty, but from the ability to scale agentic AI sustainably across the enterprise. “As organisations balance autonomy with accountability, those that embed trust, transparency, and human engagement into their AI strategy will shape the future of digital business,” Bhagtani concludes. See also: How the MCP spec update boosts security as infrastructure scales Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events including the Cyber Security Expo. Click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post Agentic AI autonomy grows in North American enterprises appeared first on AI News. View the full article

When JPMorgan Asset Management reported that AI spending accounted for two-thirds of US GDP growth in the first half of 2025, it wasn’t just a statistic – it was a signal. Enterprise leaders are making trillion-dollar bets on AI transformation, even as market observers debate whether we might be witnessing bubble-era exuberance. The conversation reached a turning point recently when OpenAI CEO Sam Altman, Amazon’s Jeff Bezos, and Goldman Sachs CEO David Solomon each acknowledged market froth within days of each other. But here’s what matters for enterprise decision-makers: acknowledging overheated markets isn’t the same as dismissing AI’s enterprise value. Corporate AI investment reached US$252.3 billion in 2024, with private investment climbing 44.5%, according to Stanford University. The question isn’t whether to invest in AI – it’s how to invest strategically while others – specifically, an organisation’s competitors – overspend on infrastructure and solutions that may never deliver returns. What separates AI winners from the 95% who fail An MIT study found that 95% of businesses invested in AI have failed to make money off the technology, according to ABC News. But that statistic masks a more important truth: 5% succeed – and they’re doing things fundamentally differently. High-performing organisations are investing more in AI capabilities, with more than one-third committing over 20% of their digital budgets to AI technologies, a McKinsey report shows. But they’re not just spending more – they’re spending smarter. The McKinsey research reveals what separates winners from the pack. About three-quarters of high performers say their organisations are scaling or have scaled AI, compared with one-third of other organisations. The leaders share common characteristics: they push for transformative innovation rather than incremental improvements, redesign workflows around AI capabilities, and implement rigorous governance frameworks. The infrastructure investment dilemma Enterprise leaders face a genuine dilemma. Google’s Gemini Ultra cost US$191 million to train, while OpenAI’s GPT-4 required US$78 million in hardware costs alone. For most enterprises, building proprietary large language models isn’t viable – and that makes vendor selection and partnership strategy important. Despite surging demand, CoreWeave slashed its 2025 capital expenditure guidance by up to 40%, citing delayed power infrastructure delivery. Oracle is “still waving off customers” due to capacity shortages, CEO Safra Catz confirmed, as per a Euronews report. This creates risk and opportunity. Enterprises that diversify their AI infrastructure strategies – building relationships with multiple providers, validating alternative architectures, and stress-testing for supply constraints – position themselves better than those betting everything on a single hyperscaler. Strategic AI investment in a frothy market Goldman Sachs equity analyst Peter Oppenheimer points out that “unlike speculative companies of the early 2000s, today’s AI giants are delivering real profits. While AI stock prices have appreciated strongly, this has been matched by sustained earnings growth.” The enterprise takeaway isn’t to avoid AI investment – it’s to avoid the mistakes that plague the 95% who see no returns: Focus on specific use cases with measurable ROI: High performers are more than three times more likely than others to say their organisation intends to use AI to bring about transformative change to their businesses, data from McKinsey shows. They’re not deploying AI for AI’s sake – they’re targeting specific business problems where AI delivers quantifiable value. Invest in organisational readiness, not just technology: Having an agile product delivery organisation is strongly correlated with achieving value. Establishing robust talent strategies and implementing technology and data infrastructure show meaningful contributions to AI success. Build governance frameworks now: The share of respondents reporting mitigation efforts for risks like personal and individual privacy, explainability, organisational reputation, and regulatory compliance has grown since 2022. As regulations tighten globally, early governance investment becomes a competitive advantage. Learning from market concentration In late 2025, 30% of the US S&P 500 was held up by just five companies – the greatest concentration in half a century. For enterprises, this concentration creates dependencies worth managing. The successful five percent diversify their AI vendors and their strategic approaches. They’re combining cloud-based AI services with edge computing, partnering with multiple model providers, and building internal capabilities for the workflows most important to competitive advantage. The real AI investment strategy Google’s Sundar Pichai captured the nuance enterprises must navigate: “We can look back at the internet right now. There was clearly a lot of excess investment, but none of us would question whether the internet was profound. I expect AI to be the same.” OpenAI’s ChatGPT has about 700 million weekly users, making it one of the fastest-growing consumer products in history. The enterprise challenge is deploying it effectively, leaving others waste billions on vanity projects. The enterprises winning at AI share a common approach: they treat AI as a business transformation initiative, not a technology project. They establish clear success metrics before deployment. They invest in change management as much as infrastructure. And they maintain healthy scepticism about vendor promises and remain committed to the technology’s potential. What this means for enterprise strategy Whether we’re in an AI bubble matters less to enterprise leaders than building sustainable AI capabilities. The market will correct itself – it always does. But businesses that develop genuine AI competencies during this investment surge will emerge stronger regardless of market dynamics. In 2024, the proportion of survey respondents reporting AI use by their organisations jumped to 78% from 55% in 2023, as per the Stanford data. AI adoption is accelerating, and enterprises that wait for perfect market conditions risk falling behind competitors building capabilities today. The strategic imperative isn’t to predict when the bubble bursts – it’s to ensure your AI investments deliver measurable business value regardless of market sentiment. Focus on practical deployments, measurable outcomes, and organisational readiness. Let others chase inflated valuations while you build sustainable competitive advantage. (Image source:Jasper Campbell) Want to experience the full spectrum of enterprise technology innovation? Join TechEx in Amsterdam, California, and London. Covering AI, Big Data, Cyber Security, IoT, Digital Transformation, Intelligent Automation, Edge Computing, and Data Centres, TechEx brings together global leaders to share real-world use cases and in-depth insights. Click here for more information. TechHQ is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post AI business reality – what enterprise leaders need to know appeared first on AI News. View the full article

If you asked most enterprise leaders which AI tools are delivering ROI, many would point to front-end chatbots or customer support automation. That’s the wrong door. The most value-generating AI systems today aren’t loud, customer-facing marvels. They’re tucked away in backend operations. They work silently, flagging irregularities in real-time, automating risk reviews, mapping data lineage, or helping compliance teams detect anomalies before regulators do. The tools don’t ask for credit, but are saving millions. Operational resilience no longer comes from having the loudest AI tool. It comes from having the smartest one, placed where it quietly does the work of five teams before lunch. The machines that spot what humans don’t Take the case of a global logistics company that integrated a background AI system for monitoring procurement contracts. The tool scanned thousands of PDFs, email chains, and invoice patterns per hour. No flashy dashboard. No alerts that interrupt workflow. Just continuous monitoring. In the first six months, it flagged multiple vendor inconsistencies that, if left unchecked, would have resulted in regulatory audits. The system didn’t just detect anomalies. It interpreted patterns. It noticed a vendor whose delivery timelines were always one day off compared to logged timestamps. Humans had seen those reports for months. But the AI noticed that the error always occurred near quarter-end. The conclusion? Inventory padding. That insight led to a contract renegotiation that saved millions. This isn’t hypothetical. One similar real-world use case reported a seven-figure operational loss prevented through a near-identical approach. That’s the kind of ROI that doesn’t need a flashy pitch deck. Why advanced education still matters in the age of AI It’s easy to fall into the trap of thinking AI tools are replacing human expertise. But smart organisations aren’t replacing but reinforcing. People with advanced academic backgrounds are helping enterprises integrate AI with strategic precision. Specifically, those with a doctorate of business administration in business intelligence bring an irreplaceable level of systems thinking and contextual insight. The professionals understand the complexity behind data ecosystems, from governance models to algorithmic biases, and can assess which tools serve long-term resilience versus short-term automation hype. When AI models are trained on historical data, it takes educated leadership to spot where historical bias may become a future liability. And when AI starts making high-stakes decisions, you need someone who can ask better questions about risk exposure, model explainability, and ethics in decision-making. This is where doctorates aren’t just nice to have – they’re essential. Invisible doesn’t mean simple Too often, companies install AI as if it were antivirus software. Set it, forget it, hope it works. That’s how you get ******-box risk. Invisible tools must still be transparent internally. It’s not enough to say, “AI flagged it.” The teams relying on these tools – risk officers, auditors, operations leads – must understand the decision-making logic or at least the signals that drive the alert. The requires not just technical documentation, but collaboration between engineers and business units. Enterprises that win with background AI systems build what could be called “decision-ready infrastructure.” The are workflows where data ingestion, validation, risk detection, and notification are all stitched together. Not in silos. Not in parallel systems. But in one loop that feeds actionable insight straight to the team responsible. That’s resilience. Where operational AI works best Here’s where invisible AI is already proving its worth in industries: Compliance Monitoring: Automatically detecting early signs of non-compliance in internal logs, transactional data, and communication channels without triggering false positives. Data Integrity: Identifying stale, duplicate, or inconsistent data in business units to prevent decision errors and reporting flaws. Fraud Detection: Recognising pattern shifts in transactions before losses occur. Not reactive alerts after the fact. Supply Chain Optimisation: Mapping supplier dependencies and predicting bottlenecks based on third-party risk signals or external disruptions. In all these cases, the key isn’t automation for automation’s sake. It’s precision. AI models that are well-calibrated, integrated with domain knowledge, and fine-tuned by experts – not simply deployed off the shelf. What makes the systems resilient? Operational resilience isn’t built in a sprint. It’s the result of smart layering. One layer catches data inconsistencies. Another tracks compliance drift. Another layer analyses behavioural signals in departments. And yet another feeds all of that into a risk model trained on historical issues. The resilience depends on: Human supervision with domain expertise, especially from those trained in business intelligence. Cross-functional transparency, so that audit, tech, and business teams are aligned. The ability to adapt models over time as the business evolves, not just retrain when performance dips. Systems that get this wrong often create alert fatigue or over-correct with rigid rule-based models. That’s not AI. That’s bureaucracy in disguise. Real ROI doesn’t scream Most ROI-focused teams chase visibility. Dashboards, reports, charts. But the most valuable AI tools don’t scream. They tap a shoulder. They point out a loose thread. They suggest a second look. That’s where the money is. Quiet detection. Small interventions. Avoided disasters. The companies that treat AI as a quiet partner – not a front-row magician – are already ahead. They’re using it to build internal resilience, not just customer-facing shine. They’re integrating it with human intelligence, not replacing it. And most of all, they’re measuring ROI not by how cool the tech looks, but by how quietly it works. That’s the future. Invisible AI agents and assistants. Visible outcomes. Real, measurable resilience. The post How background AI builds operational resilience & visible ROI appeared first on AI News. View the full article

SAP is moving its sovereignty plans forward with EU AI Cloud, a setup meant to bring its past efforts under one approach. The goal is simple: give organisations in Europe more choice and more control over how they run AI and cloud services. Some may prefer SAP’s own data centres, some may use trusted European providers, and others may want everything managed on-site. EU AI Cloud is built to support those different needs while keeping data inside the region and in line with EU rules. Strengthening AI sovereignty across Europe SAP is also working with Cohere to bring new agent-style and multimodal AI tools to customers through Cohere North. These models will be available through SAP Business Technology Platform (SAP BTP), giving industries with strict data residency needs a way to build production-ready AI into everyday operations. The two companies say the goal is to help enterprises find better insights, improve decision support, and automate complex tasks without giving up control over compliance or performance. As Cohere’s team put it, their work with SAP is meant to keep advanced AI accessible to organisations that cannot move data outside Europe. SAP is building EU AI Cloud with help from a range of European and global partners. Models and applications from Cohere, Mistral AI, OpenAI, and others are integrated directly into SAP BTP, giving customers a clearer path to build, deploy, and scale AI applications. Companies can access partner tools as SaaS, PaaS, or IaaS and choose where to run them: on SAP infrastructure or on approved European partners. The aim is to give enterprises and public sector groups access to modern AI tools while staying within European standards for security, data protection, and sovereignty. Deployment choices tied to different security needs EU AI Cloud works through SAP Sovereign Cloud, which lets customers pick the level of control they want across the stack—from infrastructure to applications. AI models run on SAP’s cloud infrastructure and SAP BTP in European data centres, which keeps operations separate from US hyperscalers. Here are the deployment options: SAP Sovereign Cloud on SAP Cloud Infrastructure (EU) SAP’s IaaS is based on open-source tools and runs inside SAP’s European data centre network. Data stays within the EU to support compliance with regional data protection rules. SAP Sovereign Cloud On-Site Infrastructure is managed by SAP but housed in a customer’s chosen data centre. This setup offers the highest level of control over data, operations, and legal requirements while keeping access to SAP’s cloud architecture. Selected hyperscalers by market Some customers may still run SAP commercial SaaS on global cloud providers. When they do, they can add sovereignty features based on regional needs. Delos Cloud A sovereign cloud service in Germany designed for the public sector. It supports local rules and is built to help government organisations modernise their digital systems. EU AI Cloud gives organisations in Europe more choice in how they run AI and cloud workloads while keeping control of their data. The mix of deployment options, partner models, and sovereign design aims to support companies that face strict rules around privacy, storage, and operational oversight. For enterprises and public bodies that need AI systems built around local requirements, SAP’s approach offers a way to use advanced tools without giving up the safeguards they rely on. (Photo by Antoine Schibler) See also: Adversarial learning breakthrough enables real-time AI security Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events, click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post SAP outlines new approach to European AI and cloud sovereignty appeared first on AI News. View the full article

The latest MCP spec update fortifies enterprise infrastructure with tighter security, moving AI agents from pilot to production. Marking its first year, the Anthropic-created open-source project released a revised spec this week aimed at the operational headaches keeping generative AI agents stuck in pilot mode. Backed by Amazon Web Services (AWS), Microsoft, and Google Cloud, the update adds support for long-running workflows and tighter security controls. The market is drifting away from fragile, bespoke integrations. For enterprises, this is a chance to deploy agentic AI that can read and write to corporate data stores without incurring massive technical debt. MCP advances from ‘developer curiosity’ to practical infrastructure The narrative has shifted from experimental chatbots to structural integration. Since September, the registry has expanded by 407 percent, now housing nearly two thousand servers. “A year on from Anthropic’s launch of the Model Context Protocol, MCP has gone from a developer curiosity to a practical way to connect AI to the systems where work and data live,” says Satyajith Mundakkal, Global CTO at Hexaware, following this latest spec update. Microsoft has already “signaled the shift by adding native MCP support to Windows 11,” effectively moving the standard directly into the operating system layer. This software standardisation arrives alongside an aggressive hardware scale-up. Mundakkal highlights the “unprecedented infrastructure build-out,” citing OpenAI’s multi-gigawatt ‘Stargate’ programme. “These are clear signals that AI capabilities, and the data they depend on, are scaling fast,” he says. MCP is the plumbing feeding these massive compute resources. As Mundakkal puts it: “AI is only as good as the data it can reach safely.” Until now, hooking an LLM into a database was mostly synchronous. That works for a chatbot checking the weather, but it fails when migrating a codebase or analysing healthcare records. The new ‘Tasks’ feature changes this (SEP-1686). It gives servers a standard way to track work, allowing clients to poll for status or cancel jobs if things go sideways. Ops teams automating infrastructure migration need agents that can run for hours without timing out. Supporting states like working or input_required finally brings resilience to agentic workflows. MCP spec update improves security For CISOs especially, AI agents often look like a massive and uncontrolled attack surface. The risks are already visible; “security researchers even found approximately 1,800 MCP servers exposed on the public internet by mid-2025,” implying that private infrastructure adoption is significantly wider. “Done poorly,” Mundakkal warns, “[MCP] becomes integration sprawl and a ******* attack surface.” To address this, the maintainers tackled the friction of Dynamic Client Registration (DCR). The fix is URL-based client registration (SEP-991), where clients provide a unique ID pointing to a self-managed metadata document to cut the admin bottleneck. Then there’s ‘URL Mode Elicitation’ (SEP-1036). It allows a server – handling payments, for instance – to bounce a user to a secure browser window for credentials. The agent never sees the password; it just gets the token. It keeps the core credentials isolated, a non-negotiable for PCI compliance. Harish Peri, SVP at Okta, believes this brings the “necessary oversight and access control to build a secure and open AI ecosystem.” One feature as part of the spec update for MCP infrastructure has somewhat flown under the radar: ‘Sampling with Tools’ (SEP-1577). Servers used to be passive data fetchers; now they can run their own loops using the client’s tokens. Imagine a “research server” spawning sub-agents to scour documents and synthesise a report. No custom client code required—it simply moves the reasoning closer to the data. However, wiring these connections is only step one. Mayur Upadhyaya, CEO at APIContext, argues that “the first year of MCP adoption has shown that enterprise AI doesn’t begin with rewrites, it begins with exposure.” But visibility is the next hurdle. “The next wave will be about visibility: enterprises will need to monitor MCP uptime and validate authentication flows just as rigorously as they monitor APIs today,” Upadhyaya explains. MCP’s roadmap reflects this, with updates targeting better “reliability and observability” for debugging. If you treat MCP servers as “set and forget,” you’re asking for trouble. Mundakkal agrees, noting the lesson from year one is to “pair MCP with strong identity, RBAC, and observability from day one.” Star-studded industry line-up adopting MCP for infrastructure A protocol is only as good as who uses it. In a year since the original spec’s release, MCP hit nearly two thousand servers. Microsoft is using it to bridge GitHub, Azure, and M365. AWS is baking it into Bedrock. Google Cloud supports it across Gemini. This reduces vendor lock-in. A Postgres connector built for MCP should theoretically work across Gemini, ChatGPT, or an internal Anthropic agent without a rewrite. The “plumbing” phase of Generative AI is settling down, and open standards are winning the debate on connectivity. Technology leaders should look to audit internal APIs for MCP readiness – focusing on exposure rather than rewrites – and verify that the new URL-based registration fits current IAM frameworks. Monitoring protocols must also be established immediately. While the latest MCP spec update is backward compatible with existing infrastructure; the new features are the only way to bring agents into regulated, mission-relevant workflows and ensure security. See also: Adversarial learning breakthrough enables real-time AI security Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events including the Cyber Security Expo. Click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post How the MCP spec update boosts security as infrastructure scales appeared first on AI News. View the full article

The next frontier for edge AI medical devices isn’t wearables or bedside monitors—it’s inside the human body itself. Cochlear’s newly launched Nucleus Nexa System represents the first cochlear implant capable of running machine learning algorithms while managing extreme power constraints, storing personalised data on-device, and receiving over-the-air firmware updates to improve its AI models over time. For AI practitioners, the technical challenge is staggering: build a decision-tree model that classifies five distinct auditory environments in real time, optimise it to run on a device with a minimal power budget that must last decades, and do it all while directly interfacing with human neural tissue. Decision trees meet ultra-low power computing At the core of the system’s intelligence lies SCAN 2, an environmental classifier that analyses incoming audio and categorises it as Speech, Speech in Noise, Noise, Music, or Quiet. “These classifications are then input to a decision tree, which is a type of machine learning model,” explains Jan Janssen, Cochlear’s Global CTO, in an exclusive interview with AI News. “This decision is used to adjust sound processing settings for that situation, which adapts the electrical signals sent to the implant.” The model runs on the external sound processor, but here’s where it gets interesting: the implant itself participates in the intelligence through Dynamic Power Management. Data and power are interleaved between the processor and implant via an enhanced RF link, allowing the chipset to optimise power efficiency based on the ML model’s environmental classifications. This isn’t just smart power management—it’s edge AI medical devices solving one of the hardest problems in implantable computing: how do you keep a device operational for 40+ years when you can’t replace its battery? The spatial intelligence layer Beyond environmental classification, the system employs ForwardFocus, a spatial noise algorithm that uses inputs from two omnidirectional microphones to create target and noise spatial patterns. The algorithm assumes target signals originate from the front while noise comes from the sides or behind, then applies spatial filtering to attenuate background interference. What makes this noteworthy from an AI perspective is the automation layer. ForwardFocus can operate autonomously, removing cognitive load from users navigating complex auditory scenes. The decision to activate spatial filtering happens algorithmically based on environmental analysis—no user intervention required. Upgradeability: The medical device AI paradigm shift Here’s the breakthrough that separates this from previous-generation implants: upgradeable firmware in the implanted device itself. Historically, once a cochlear implant was surgically placed, its capabilities were frozen. New signal processing algorithms, improved ML models, better noise reduction—none of it could benefit existing patients. Jan Janssen, Chief Technology Officer, Cochlear Limited The Nucleus Nexa Implant changes that equation. Using Cochlear’s proprietary short-range RF link, audiologists can deliver firmware updates through the external processor to the implant. Security relies on physical constraints—the limited transmission range and low power output require proximity during updates—combined with protocol-level safeguards. “With the smart implants, we actually keep a copy [of the user’s personalised hearing map] on the implant,” Janssen explained. “So you lose this [external processor], we can send you a blank processor and put it on—it retrieves the map from the implant.” The implant stores up to four unique maps in its internal memory. From an AI deployment perspective, this solves a critical challenge: how do you maintain personalised model parameters when hardware components fail or get replaced? From decision trees to deep neural networks Cochlear’s current implementation uses decision tree models for environmental classification—a pragmatic choice given power constraints and interpretability requirements for medical devices. But Janssen outlined where the technology is headed: “Artificial intelligence through deep neural networks—a complex form of machine learning—in the future may provide further improvement in hearing in noisy situations.” The company is also exploring AI applications beyond signal processing. “Cochlear is investigating the use of artificial intelligence and connectivity to automate routine check-ups and reduce lifetime care costs,” Janssen noted. This points to a broader trajectory for edge AI medical devices: from reactive signal processing to predictive health monitoring, from manual clinical adjustments to autonomous optimisation. The Edge AI constraint problem What makes this deployment fascinating from an ML engineering standpoint is the constraint stack: Power: The device must run for decades on minimal energy, with battery life measured in full days despite continuous audio processing and wireless transmission. Latency: Audio processing happens in real-time with imperceptible delay—users can’t tolerate lag between speech and neural stimulation. Safety: This is a life-critical medical device directly stimulating neural tissue. Model failures aren’t just inconvenient—they impact quality of life. Upgradeability: The implant must support model improvements over 40+ years without hardware replacement. Privacy: Health data processing happens on-device, with Cochlear applying rigorous de-identification before any data enters their Real-World Evidence program for model training across their 500,000+ patient dataset. These constraints force architectural decisions you don’t face when deploying ML models in the cloud or even on smartphones. Every milliwatt matters. Every algorithm must be validated for medical safety. Every firmware update must be bulletproof. Beyond Bluetooth: The connected implant future Looking ahead, Cochlear is implementing Bluetooth LE Audio and Auracast broadcast audio capabilities—both requiring future firmware updates to the implant. These protocols offer better audio quality than traditional Bluetooth while reducing power consumption, but more importantly, they position the implant as a node in broader assistive listening networks. Auracast broadcast audio allows direct connection to audio streams in public venues, airports, and gyms—transforming the implant from an isolated medical device into a connected edge AI medical device participating in ambient computing environments. The longer-term vision includes totally implantable devices with integrated microphones and batteries, eliminating external components entirely. At that point, you’re talking about fully autonomous AI systems operating inside the human body—adjusting to environments, optimising power, streaming connectivity, all without user interaction. The medical device AI blueprint Cochlear’s deployment offers a blueprint for edge AI medical devices facing similar constraints: start with interpretable models like decision trees, optimise aggressively for power, build in upgradeability from day one, and architect for the 40-year horizon rather than the typical 2-3 year consumer device cycle. As Janssen noted, the smart implant launching today “is actually the first step to an even smarter implant.” For an industry built on rapid iteration and continuous deployment, adapting to decade-long product lifecycles while maintaining AI advancement represents a fascinating engineering challenge. The question isn’t whether AI will transform medical devices—Cochlear’s deployment proves it already has. The question is how quickly other manufacturers can solve the constraint problem and bring similarly intelligent systems to market. For 546 million people with hearing loss in the Western Pacific Region alone, the pace of that innovation will determine whether AI in medicine remains a prototype story or becomes standard of care. (Photo by Cochlear) See also: FDA AI deployment: Innovation vs oversight in drug regulation Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events, click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post Edge AI inside the human body: Cochlear’s machine learning implant breakthrough appeared first on AI News. View the full article

Indonesia’s push into AI-led growth is gaining momentum as more local organisations look for ways to build their own applications, update their systems, and strengthen data oversight. The country now has broader access to cloud and AI tools after Microsoft expanded the services available in the Indonesia Central cloud region, which first went live six months ago. The expansion gives businesses, public bodies, and developers more options to run AI workloads inside the country instead of relying on overseas data centres. The update was shared at the Cloud & AI Innovation Summit in Jakarta, where business and government leaders met to discuss how Indonesia can advance its AI ambitions. Speakers included Mike Chan, who leads Azure AI Apps & Agents in Asia, and Dharma Simorangkir, President Director of Microsoft Indonesia. Their message was consistent: local capacity is only useful if organisations put it to work. During the event, Dharma said the new services “open the door for every organisation to innovate in Indonesia, for Indonesia,” calling on teams across sectors to build solutions that tackle national needs. A shift toward building, not just adopting Many Indonesian enterprises are moving beyond basic AI trials and are now designing tools that solve problems unique to their operations. Microsoft describes these kinds of organisations as Frontier Firms — teams that treat AI as a core part of how they work rather than an optional add-on. These firms tend to focus on building applications that make tasks easier for customers, improve internal processes, or modernise old workflows. To support this shift, the Indonesia Central region now hosts a range of Azure services that help teams design and deploy software. These include tools for building data-connected applications, services for storing and managing structured data, and a set of AI-ready virtual machines that can train and run advanced models. The machines, built for heavy computing work, allow teams to keep data inside the country while working with complex AI workloads. The region now supports Microsoft 365 Copilot as well, bringing AI features to common work tools. Developers also have access to GitHub Copilot, which suggests code and speeds up software development. These services form a connected stack that helps teams move past small pilots and into production, where reliability and cost control matter more. Early Microsoft cloud projects emerging across Indonesia The expansion of the region follows steady demand since its launch in May 2025. Companies across mining, travel, and digital services are already using local cloud infrastructure to refresh legacy systems and meet stricter data governance needs. Petrosea and Vale Indonesia are among the firms using the region to support technical upgrades and secure local data storage. Digital-first players are also experimenting with more direct AI engagement. One example is tiket.com, which built its own AI travel assistant using the Azure OpenAI Service. The assistant lets customers interact with the platform in everyday language, from checking flight updates to adding extra services after a booking. “Our advancements in artificial intelligence are designed to deliver the best possible experience for our customers,” said Irvan Bastian Arief, PhD, Vice President of Technology GRAND, Data & AI at tiket.com. The company sees conversational AI as a way to make travel planning simpler while reducing friction in customer support. Bringing scattered data into one system A major theme at the Summit was the need to get data in order before adopting AI at scale. To support this, Microsoft introduced Microsoft Fabric to the Indonesian market. Fabric is a single environment that brings together data engineering, integration, warehousing, analytics, and business intelligence. It includes Copilot features that help teams prepare data and build insights without juggling multiple tools. For many organisations, data sits across different internal systems and cloud providers. Fabric gives teams one place to bring these sources together, which may help improve governance, speed up reporting, and control costs. The platform is designed for teams that want structure without building their own data foundation from scratch. Preparing Indonesia’s workforce for practical AI with Microsoft tools The day’s focus was not limited to infrastructure. Microsoft also highlighted its AI training program, Microsoft Elevate, which is now entering its second year. The program has already reached more than 1.2 million learners and aims to certify 500,000 people in AI skills by 2026. The next phase will focus on hands-on use, encouraging participants to apply AI in real settings rather than only learning concepts in theory. Training covers a wide range of groups — teachers, nonprofit workers, community leaders, and people looking to improve their digital skills. Participants learn through tools such as Microsoft Copilot, Learning Accelerator, Minecraft Education, and modules designed to explain how AI can support practical tasks. During the Summit, Dharma said that cloud and AI “are the backbone of national competitiveness” and stressed that infrastructure only matters if people are prepared to use it. Building a long-term ecosystem These efforts sit within a broader commitment of US$1.7 billion that Microsoft has pledged for Indonesia from 2024 to 2028. The investment spans infrastructure, partner support, and talent development. The company is also preparing to host GitHub Universe Jakarta on 3 December 2025, a developer-focused gathering meant to encourage collaboration among software teams, startups, and researchers. Indonesia is aiming to position itself as a centre for secure and inclusive AI development in the region. With the expansion of the Indonesia Central cloud region, new data and AI tools, and growing attention on workforce training, the country is taking steps to build the foundations needed for long-term digital growth. Companies now have the option to build AI systems closer to home, developers have more resources, and workers have more pathways to gain practical skills. The coming years will show how these pieces fit together as organisations move from experimentation to long-term use. (Photo by Simon Ray) See also: Microsoft, NVIDIA, and Anthropic forge AI compute alliance Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events, click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post New Microsoft cloud updates support Indonesia’s long-term AI goals appeared first on AI News. View the full article

Manufacturers today are working against rising input costs, labour shortages, supply-chain fragility, and pressure to offer more customised products. AI is becoming an important part of a response to those pressures. When enterprise strategy depends on AI Most manufacturers seek to reduce cost while improving throughput and quality. AI supports these aims by predicting equipment failures, adjusting production schedules, and analysing supply-chain signals. A Google Cloud survey found that more than half of manufacturing executives are using AI agents in back-office areas like planning and quality. ([Hidden Content]) The shift matters because the use of AI links directly to measurable business outcomes. Reduced downtime, lower scrap, better OEE (overall equipment effectiveness), and improved customer responsiveness all contribute to positive enterprise strategy and overall competitiveness in the market. What recent industry experience reveals Motherson Technology Services reported major gains – 25-30% maintenance-cost reduction, 35-45% downtime reduction, and 20-35% higher production efficiency after adopting agent-based AI, data-platform consolidation, and workforce-enablement initiatives. ServiceNow has described how manufacturers unify workflows, data, and AI on common platforms. It reported that just over half of advanced manufacturers have formal data-governance programmes in support of their AI initiatives. These instances show the direction of travel: AI is being deployed inside operations – not in pilots, but in workflows. What cloud and IT leaders should consider Data architecture Manufacturing systems depend on low-latency decisions, especially for maintenance and quality. Leaders must work out how to combine edge devices (often OT systems with supporting IT infrastructure) with cloud services. Microsoft’s maturity-path guidance highlights that data silos and legacy equipment remain a barrier, so standardising how data is collected, stored, and shared is often the first step for many future-facing manufacturing and engineering businesses. Use-case sequencing ServiceNow advises starting small and scaling AI roll-outs gradually. Focusing on two or three high-value use-cases helps teams avoid the “pilot trap”. Predictive maintenance, energy optimisation, and quality inspection are strong starting points because benefits are relatively easy to measure. Governance and security Connecting operational technology equipment with IT and cloud systems increases cyber-risk, as some OT systems were not designed to be exposed to the wider internet. Leaders should define data-access rules and monitoring requirements carefully. In general, AI governance should not wait until later phases, but begin in the first pilot. Workforce and skills The human factor remains important. Operators’ trust AI-supported systems goes without saying and there needs to be confidence using systems underpinned by AI. According to Automation.com, manufacturing faces persistent skilled-labour shortages, making upskilling programmes an integral part of modern deployments. Vendor-ecosystem neutrality The ecosystem of many manufacturing environments includes IoT sensors, industrial networks, cloud platforms, and workflow tools operating in the back office and on the facility floor. Leaders should prioritise interoperability and avoid lock-in to any one provider. The aim is not to adopt a single vendor’s approach but to build an architecture that supports long-term flexibility, honed to the individual organisation’s workflows. Measuring impact Manufacturers should define metrics, which may include downtime hours, maintenance-cost reduction, throughput, yield, and these metrics should be monitored continuously. The Motherson results provide realistic benchmarks and show the outcomes possible from careful measurement. The realities: beyond the hype Despite rapid progress, challenges remain. Skills shortages slow deployment, legacy machinery produces fragmented data, and costs are sometimes difficult to forecast. Sensors, connectivity, integration work, and data-platform upgrades all add up. Additionally, security issues grow as production systems become more connected. Finally, AI should coexist with human expertise; operators, engineers, and data scientists behind the scenes need to work together, not in parallel. However, recent publications show these challenges are manageable with the right management and operational structures. Clear governance, cross-functional teams, and scalable architectures make AI easier to deploy and sustain. Strategic recommendations for leaders Tie AI initiatives to business goals. Link work to KPIs like downtime, scrap, and cost per unit. Adopt a careful hybrid edge-cloud mix. Keep real-time inference close to machines while using cloud platforms for training and analytics. Invest in people. Mixed teams of domain experts and data scientists are important, and training should be offered for operators and management. Embed security early. Treat OT and IT as a unified environment, assuming zero-trust. Scale gradually. Prove value in one plant, then expand. Choose open ecosystem components. Open standards allow a company to remain flexible and avoid vendor lock-in. Monitor performance. Adjust models and workflows as conditions change, according to results measured against pre-defined metrics. Conclusion Internal AI deployment is now an important part of manufacturing strategy. Recent blog posts from Motherson, Microsoft, and ServiceNow show that manufacturers are gaining measurable benefits by combining data, people, workflows, and technology. The path is not simple, but with clear governance, the right architecture, an eye to security, business-focussed projects, and a strong focus on people, AI becomes a practical lever for competitiveness. (Image source: “Jelly Belly Factory Floor” by el frijole is licensed under CC BY-NC-SA 2.0. ) Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and co-located with other leading technology events. Click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post Manufacturing’s pivot: AI as a strategic driver appeared first on AI News. View the full article

The ability to execute adversarial learning for real-time AI security offers a decisive advantage over static defence mechanisms. The emergence of AI-driven attacks – utilising reinforcement learning (RL) and Large Language Model (LLM) capabilities – has created a class of “vibe hacking” and adaptive threats that mutate faster than human teams can respond. This represents a governance and operational risk for enterprise leaders that policy alone cannot mitigate. Attackers now employ multi-step reasoning and automated code generation to bypass established defences. Consequently, the industry is observing a necessary migration toward “autonomic defence” (i.e. systems capable of learning, anticipating, and responding intelligently without human intervention.) Transitioning to these sophisticated defence models, though, has historically hit a hard operational ceiling: latency. Applying adversarial learning, where threat and defence models are trained continuously against one another, offers a method for countering malicious AI security threats. Yet, deploying the necessary transformer-based architectures into a live production environment creates a bottleneck. Abe Starosta, Principal Applied Research Manager at Microsoft NEXT.ai, said: “Adversarial learning only works in production when latency, throughput, and accuracy move together. Computational costs associated with running these dense models previously forced leaders to choose between high-accuracy detection (which is slow) and high-throughput heuristics (which are less accurate). Engineering collaboration between Microsoft and NVIDIA shows how hardware acceleration and kernel-level optimisation remove this barrier, making real-time adversarial defence viable at enterprise scale. Operationalising transformer models for live traffic required the engineering teams to target the inherent limitations of CPU-based inference. Standard processing units struggle to handle the volume and velocity of production workloads when burdened with complex neural networks. In baseline tests conducted by the research teams, a CPU-based setup yielded an end-to-end latency of 1239.67ms with a throughput of just 0.81req/s. For a financial institution or global e-commerce platform, a one-second delay on every request is operationally untenable. By transitioning to a GPU-accelerated architecture (specifically utilising NVIDIA H100 units), the baseline latency dropped to 17.8ms. Hardware upgrades alone, though, proved insufficient to meet the strict requirements of real-time AI security. Through further optimisation of the inference engine and tokenisation processes, the teams achieved a final end-to-end latency of 7.67ms—a 160x performance speedup compared to the CPU baseline. Such a reduction brings the system well within the acceptable thresholds for inline traffic analysis, enabling the deployment of detection models with greater than 95 percent accuracy on adversarial learning benchmarks. One operational hurdle identified during this project offers valuable insight for CTOs overseeing AI integration. While the classifier model itself is computationally heavy, the data pre-processing pipeline – specifically tokenisation – emerged as a secondary bottleneck. Standard tokenisation techniques, often relying on whitespace segmentation, are designed for natural language processing (e.g. articles and documentation). They prove inadequate for cybersecurity data, which consists of densely packed request strings and machine-generated payloads that lack natural breaks. To address this, the engineering teams developed a domain-specific tokeniser. By integrating security-specific segmentation points tailored to the structural nuances of machine data, they enabled finer-grained parallelism. This bespoke approach for security delivered a 3.5x reduction in tokenisation latency, highlighting that off-the-shelf AI components often require domain-specific re-engineering to function effectively in niche environments. Achieving these results required a cohesive inference stack rather than isolated upgrades. The architecture utilised NVIDIA Dynamo and Triton Inference Server for serving, coupled with a TensorRT implementation of Microsoft’s threat classifier. The optimisation process involved fusing key operations – such as normalisation, embedding, and activation functions – into single custom CUDA kernels. This fusion minimises memory traffic and launch overhead, which are frequent silent killers of performance in high-frequency trading or security applications. TensorRT automatically fused normalisation operations into preceding kernels, while developers built custom kernels for sliding window attention. The result of these specific inference optimisations was a reduction in forward-pass latency from 9.45ms to 3.39ms, a 2.8x speedup that contributed the majority of the latency reduction seen in the final metrics. Rachel Allen, Cybersecurity Manager at NVIDIA, explained: “Securing enterprises means matching the volume and velocity of cybersecurity data and adapting to the innovation speed of adversaries. “Defensive models need the ultra-low latency to run at line-rate and the adaptability to protect against the latest threats. The combination of adversarial learning with NVIDIA TensorRT accelerated transformer-based detection models does just that.” Success here points to a broader requirement for enterprise infrastructure. As threat actors leverage AI to mutate attacks in real-time, security mechanisms must possess the computational headroom to run complex inference models without introducing latency. Reliance on CPU compute for advanced threat detection is becoming a liability. Just as graphics rendering moved to GPUs, real-time security inference requires specialised hardware to maintain throughput >130 req/s while ensuring robust coverage. Furthermore, generic AI models and tokenisers often fail on specialised data. The “vibe hacking” and complex payloads of modern threats require models trained specifically on malicious patterns and input segmentations that reflect the reality of machine data. Looking ahead, the roadmap for future security involves training models and architectures specifically for adversarial robustness, potentially using techniques like quantisation to further enhance speed. By continuously training threat and defence models in tandem, organisations can build a foundation for real-time AI protection that scales with the complexity of evolving security threats. The adversarial learning breakthrough demonstrates the technology to achieve this – balancing latency, throughput, and accuracy – is now capable of being deployed today. See also: ZAYA1: AI model using AMD GPUs for training hits milestone Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events including the Cyber Security Expo. Click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post Adversarial learning breakthrough enables real-time AI security appeared first on AI News. View the full article

Malaysia has captured 32% of Southeast Asia’s total AI funding—equivalent to US$759 million—between H2 2024 and H1 2025, establishing itself as the region’s dominant destination for artificial intelligence investment as massive infrastructure expansion and high consumer adoption converge to reshape the country’s technology landscape, according to the e-Conomy SEA 2025 report released by Google, Temasek, and Bain & Company. The Malaysia AI investment surge is underpinned by a dramatic expansion in physical infrastructure that sets the country apart from regional competitors. Data centre capacity exploded from 120 megawatts in 2024 to 690 MW in the first half of 2025, with plans reported to further increase capacity by 350%—representing half of all planned regional capacity. This infrastructure-first approach appears to be working. Google has committed US$2 billion in investment, including the development of its first Google data centre and Google Cloud region in Malaysia, specifically to meet growing demand for AI-ready cloud services both locally and globally. The funding reality: concentration and opportunity While the headline US$759 million figure positions Malaysia as a regional leader in Malaysia AI investment, the composition reveals both strengths and vulnerabilities. The funding was supported primarily by major digital financial services deals, particularly a significant private equity transaction in H2 2024 that elevated the overall numbers. Private funding across Malaysia’s broader digital economy tells a more nuanced story. The deal count in H1 2025 stood at just 23 deals, significantly below the 2021 peak of 236 deals, indicating that while individual transaction sizes have increased, the breadth of investment activity has narrowed considerably. Digital financial services accounted for 84% of H1 2024 funding, raising questions about whether Malaysia’s AI investment ecosystem has sufficient diversification to sustain momentum if fintech consolidation slows or regulatory headwinds emerge. However, investor sentiment remains optimistic. Nearly two-thirds (64%) of surveyed investors expect funding activity in Malaysia to rise through 2030, particularly in software, services, AI and deep tech—categories that extend beyond the current fintech concentration. Malaysia also led Southeast Asia in IPO activity over the past 12 months, contributing roughly half of the region’s total listings. This exit activity signals that investors see viable pathways to liquidity, a critical factor for sustaining long-term AI investment flows. Consumer adoption: rapid uptake with emerging commercial validation If infrastructure investment represents Malaysia’s strategic bet on AI, consumer behaviour suggests the market is responding. Some 74% of Malaysian digital consumers report interacting with AI tools and features daily—a penetration rate that positions the country among the region’s most engaged AI user bases. The nature of engagement extends beyond passive consumption. According to the report, 68% of consumers have conversations with and ask questions of AI chatbots, indicating comfort with conversational AI interfaces that go beyond simple task automation. More significantly for commercial AI development, 55% of Malaysian consumers expect AI to make decisions faster and with less mental effort. This trust signal suggests readiness for agentic AI applications that operate with greater autonomy. This consumer readiness is translating into measurable commercial outcomes. Revenue growth for apps with marketed AI features surged 103% in H1 2025 compared to H1 2024, providing concrete evidence that AI functionality drives monetisation beyond experimentation or novelty value. “With three in four Malaysian digital consumers having used GenAI tools, this strong daily engagement is laying a solid foundation for the next phase of AI-powered growth,” said Ben King, Managing Director of Google Malaysia & Singapore. “In line with the nation’s goal of becoming a regional digital leader by 2030, Google remains fully committed to supporting Malaysia’s ambition to build an inclusive, innovative, and AI-ready digital economy.” The trust equation: data sharing versus privacy concerns One of the most striking findings in Malaysia’s AI adoption profile is consumer willingness to share data access with AI agents. Some 92% of respondents indicated they would share data such as shopping and viewing history, and social connections with AI systems—a figure that significantly exceeds comfort levels seen in more privacy-conscious markets. For context, privacy and data security concerns around agentic AI in Malaysia stand at 60%, which is actually 10 percentage points higher than the ASEAN-10 average of 50%. This apparent contradiction—high willingness to share data coupled with elevated privacy concerns—suggests Malaysian consumers recognise both the utility and the risks of AI systems, rather than exhibiting naive enthusiasm. This nuanced trust profile creates both opportunities and responsibilities for AI developers. The willingness to share data enables more sophisticated personalisation and AI agent capabilities, but the parallel privacy concerns indicate that consumers expect robust data governance in return. Top motivations for using or paying for AI features reveal a pragmatic consumer base. Saving time on research and comparisons ranks highest at 51%, followed by saving money through better deals or price tracking at 39%, and exclusive access to products and 24/7 customer support at 30%. These priorities suggest AI adoption in Malaysia is driven by functional value rather than technological curiosity. Infrastructure scale meets strategic questions The planned 350% increase in data centre capacity positions Malaysia to host not just domestic AI workloads but regional and potentially global operations. Half of all planned Southeast Asian data centre capacity being located in Malaysia represents a concentration that could drive network effects and talent clustering. However, several strategic questions remain unresolved. Can Malaysia move beyond hosting infrastructure to developing proprietary AI capabilities? The emergence of ILMU, Malaysia’s first home-grown large language model now being deployed by digital banks, suggests domestic AI development is beginning, but scale remains limited. Will the infrastructure investments translate into high-value job creation, or will Malaysia primarily provide the physical substrate while control and value accrue elsewhere? The country’s 80% AI awareness rate—indicating most users have learned about AI through various approaches—suggests potential for workforce development, but awareness alone doesn’t guarantee technical capability. The regulatory environment also faces testing. The new Consumer Credit Act, requiring buy-now-pay-later providers and non-bank lenders to be licensed, indicates authorities are introducing structure to previously loosely governed digital sectors. How regulators approach AI governance—balancing innovation enablement with consumer protection—will significantly impact whether Malaysia’s AI investment sustains its current trajectory. Regional implications and competitive dynamics Malaysia’s infrastructure and funding concentration create both collaboration and competition dynamics across Southeast Asia. The interoperability of the DuitNow QR standard across an increasing number of regional markets, now including Cambodia, demonstrates Malaysia’s capacity for cross-border digital integration that could extend to AI services. However, as neighbouring countries observe Malaysia’s AI momentum, competitive infrastructure buildouts are likely. The sustainability of Malaysia’s current leadership position depends on translating first-mover advantages into durable capabilities—technical talent, regulatory frameworks, and commercial ecosystems that compound rather than commoditise. “The real opportunity now lies in how businesses harness AI as a catalyst for impact while building on Malaysia’s strong digital foundations,” said Amanda Chin, Partner at Bain & Company. This framing acknowledges that infrastructure and funding, while necessary, are insufficient without execution. As Malaysia’s AI investment reaches significant scale, the critical test shifts from capital attraction to value creation—whether the US$759 million in funding and massive infrastructure expansion generate genuinely innovative AI applications or primarily replicate capabilities developed elsewhere. The data confirms Malaysia has secured a leadership position in Southeast Asia’s AI landscape. Converting that position into sustained technological advantage requires moving beyond infrastructure provision into invention, a transition that remains very much in progress. (Photo by Luiz Cent) See also: Huawei commits to training 30,000 Malaysian AI professionals as local tech ecosystem expands Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events including the Cyber Security Expo. Click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post e-Conomy SEA 2025: Malaysia takes 32% of regional AI funding appeared first on AI News. View the full article

Zyphra, AMD, and IBM spent a year testing whether AMD’s GPUs and platform can support large-scale AI model training, and the result is ZAYA1. In partnership, the three companies trained ZAYA1 – described as the first major Mixture-of-Experts foundation model built entirely on AMD GPUs and networking – which they see as proof that the market doesn’t have to depend on NVIDIA to scale AI. The model was trained on AMD’s Instinct MI300X chips, Pensando networking, and ROCm software, all running across IBM Cloud’s infrastructure. What’s notable is how conventional the setup looks. Instead of experimental hardware or obscure configurations, Zyphra built the system much like any enterprise cluster—just without NVIDIA’s components. Zyphra says ZAYA1 performs on par with, and in some areas ahead of, well-established open models in reasoning, maths, and code. For businesses frustrated by supply constraints or spiralling GPU pricing, it amounts to something rare: a second option that doesn’t require compromising on capability. How Zyphra used AMD GPUs to cut costs without gutting AI training performance Most organisations follow the same logic when planning training budgets: memory capacity, communication speed, and predictable iteration times matter more than raw theoretical throughput. MI300X’s 192GB of high-bandwidth memory per GPU gives engineers some breathing room, allowing early training runs without immediately resorting to heavy parallelism. That tends to simplify projects that are otherwise fragile and time-consuming to tune. Zyphra built each node with eight MI300X GPUs connected over InfinityFabric and paired each one with its own Pollara network card. A separate network handles dataset reads and checkpointing. It’s an unfussy design, but that seems to be the point; the simpler the wiring and network layout, the lower the switch costs and the easier it is to keep iteration times steady. ZAYA1: An AI model that punches above its weight ZAYA1-base activates 760 million parameters out of a total 8.3 billion and was trained on 12 trillion tokens in three stages. The architecture leans on compressed attention, a refined routing system to steer tokens to the right experts, and lighter-touch residual scaling to keep deeper layers stable. The model uses a mix of Muon and AdamW. To make Muon efficient on AMD hardware, Zyphra fused kernels and trimmed unnecessary memory traffic so the optimiser wouldn’t dominate each iteration. Batch sizes were increased over time, but that depends heavily on having storage pipelines that can deliver tokens quickly enough. All of this leads to an AI model trained on AMD hardware that competes with larger peers such as Qwen3-4B, Gemma3-12B, Llama-3-8B, and OLMoE. One advantage of the MoE structure is that only a sliver of the model runs at once, which helps manage inference memory and reduces serving cost. A bank, for example, could train a domain-specific model for investigations without needing convoluted parallelism early on. The MI300X’s memory headroom gives engineers space to iterate, while ZAYA1’s compressed attention cuts prefill time during evaluation. Making ROCm behave with AMD GPUs Zyphra didn’t hide the fact that moving a mature NVIDIA-based workflow onto ROCm took work. Instead of porting components blindly, the team spent time measuring how AMD hardware behaved and reshaping model dimensions, GEMM patterns, and microbatch sizes to suit MI300X’s preferred compute ranges. InfinityFabric operates best when all eight GPUs in a node participate in collectives, and Pollara tends to reach peak throughput with larger messages, so Zyphra sized fusion buffers accordingly. Long-context training, from 4k up to 32k tokens, relied on ring attention for sharded sequences and tree attention during decoding to avoid bottlenecks. Storage considerations were equally practical. Smaller models hammer IOPS; larger ones need sustained bandwidth. Zyphra bundled dataset shards to reduce scattered reads and increased per-node page caches to speed checkpoint recovery, which is vital during long runs where rewinds are inevitable. Keeping clusters on their feet Training jobs that run for weeks rarely behave perfectly. Zyphra’s Aegis service monitors logs and system metrics, identifies failures such as NIC glitches or ECC blips, and takes straightforward corrective actions automatically. The team also increased RCCL timeouts to keep short network interruptions from killing entire jobs. Checkpointing is distributed across all GPUs rather than forced through a single chokepoint. Zyphra reports more than ten-fold faster saves compared with naïve approaches, which directly improves uptime and cuts operator workload. What the ZAYA1 AMD training milestone means for AI procurement The report draws a clean line between NVIDIA’s ecosystem and AMD’s equivalents: NVLINK vs InfinityFabric, NCCL vs RCCL, cuBLASLt vs hipBLASLt, and so on. The authors argue the AMD stack is now mature enough for serious large-scale model development. None of this suggests enterprises should tear out existing NVIDIA clusters. A more realistic path is to keep NVIDIA for production while using AMD for stages that benefit from the memory capacity of MI300X GPUs and ROCm’s openness. It spreads supplier risk and increases total training volume without major disruption. This all leads us to a set of recommendations: treat model shape as adjustable, not fixed; design networks around the collective operations your training will actually use; build fault tolerance that protects GPU hours rather than merely logging failures; and modernise checkpointing so it no longer derails training rhythm. It’s not a manifesto, just our practical takeaway from what Zyphra, AMD, and IBM learned by training a large MoE AI model on AMD GPUs. For organisations looking to expand AI capacity without relying solely on one vendor, it’s a potentially useful blueprint. See also: Google commits to 1000x more AI infrastructure in next 4-5 years Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events including the Cyber Security Expo. Click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post ZAYA1: AI model using AMD GPUs for training hits milestone appeared first on AI News. View the full article

In order to meet the massive demand for AI, Google wants to double the overall size of its servers every six months, a growth rate that would create a 1000x greater capacity in the next four or five years. The statement came from the head of Google’s AI infrastructure, Amin Vahdat, during an all-hands meeting on November 6, according to CNBC. Alphabet, Google’s parent company is certainly performing well, so such a requirement may be within its financial capabilities. It reported good Q3 figures at the end of October, and has raised its capital expenditure forecast to $93 billion, up from $91 billion. Vahdat addressed one employee’s question about the company’s future amid talk of an ‘AI bubble’ by re-stating the risks of not investing aggressively enough. In its cloud operations, such investment in infrastructure has paid off. “The risk of under-investing is pretty high […] the cloud numbers would have been much better if we had more compute.” Google’s cloud business continues to grow at around a 33% per year, creating an income stream that enables the company to be “better positioned to withstand misses than other companies,” he said. With better infrastructure running more efficient hardware such as the seventh-gen Tensor Processing Unit and more efficient LLM models, Google is confident that it can continue to create value for its enterprise users’ increased implementation of AI technologies. According to Markus Nispel of Extreme Networks, writing on techradar.com in September, it’s IT infrastructure that’s making companies’ AI vision falter. He places the blame for any failure of AI projects on the high demands AI workloads place on legacy systems, the need for real-time and edge facilities (often lacking in current enterprises), and the continuing presence of data silos. “Even when projects do launch, they’re often hampered by delays caused by poor data availability or fragmented systems. If clean, real-time data can’t flow freely across the organisation, AI models can’t operate effectively, and the insights they produce arrive too late or lack impact,” he said. “With 80% of AI projects struggling to deliver on expectations globally, primarily due to infrastructure limitations rather than the AI technology itself, what matters now is how we respond.” His views are shared by decision-makers at the large technology providers: Capital expenditure by Google, Microsoft, Amazon, and Meta is expected to top $380 billion this year, the majority of which is focused on AI infrastructure. The message from the hyperscalers is clear: If we build it, they will come. Addressing the infrastructure challenges that organisations experience is the key component to successful implementation of AI-based projects. Agile infrastructure as close as possible to the point of compute and data sets that are unified are seen as important parts of the recipe for getting full value from next-generation AI projects. Although some market realignment is expected across the AI sector in the next six months, companies like Google are among those expected to be able to consolidate on the market and continue to offer game-changing technologies based on AI as it evolves. (Image source: “Construction site” by tomavim is licensed under CC BY-NC 2.0.) Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and co-located with other leading technology events. Click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post Google commits to 1000x more AI infrastructure in next 4-5 years appeared first on AI News. View the full article

A €1.2 trillion AI prize sits on the table for Europe’s economy, and the region has the talent and raw ingredients to claim it. While the global narrative often focuses on competition with the US and China, the view from the ground in Europe is a region of untapped potential, world-class talent, and deep infrastructure investment. Debbie Weinstein, President of Google EMEA, sees a “new generation of visionary founders” ready to drive the region’s future. The opportunity is built on a foundation of scientific excellence and a workforce that is “as bright as anywhere else in the world.” The task now is to leverage Europe’s strengths to close the AI adoption gap and accelerate growth. A foundation of innovation Europe is already a powerhouse of scientific breakthrough. The Google DeepMind team – which includes Nobel prize winners – drives discovery from London, while nearly one million researchers across EMEA use AlphaFold to solve biological problems. Europe isn’t starting from scratch; it is a hub of high-level R&D. That intellectual capital is being matched by hard investment. Just last week, Google announced a €5.5 billion investment in Germany to support connectivity and infrastructure. The choice to base ‘Security Operations Centres’ in Munich, Dublin, and Malaga also highlights Europe’s specific strength: a deep, culturally ingrained commitment to privacy and security. For businesses, this signals that Europe offers a stable and secure environment for building long-term digital strategies. The potential of AI in Europe Currently, only 14 percent of European businesses use AI. While some see this as a lag, optimists see it as massive headroom for growth. The businesses that do adopt these tools are seeing powerful results. Weinstein points to Spanish startup Idoven as a prime example of Europe’s potential. They are using AI to help doctors detect heart disease earlier, proving that when European founders get access to the right tools, they build world-changing solutions. The operational gains are equally tangible in traditional sectors. In automotive, upgrading from basic voice assistants to AI co-pilots can prevent accidents by detecting driver fatigue. In cybersecurity, modern tools allow teams to stay ahead of sophisticated threats. The technology acts as a force multiplier, giving businesses the “most powerful toolbox they’ve ever had.” To fully realise this €1.2 trillion potential, Europe’s businesses need access to the same high-performance AI models as their global peers. The latest models are 300 times more powerful than those from two years ago, offering a massive productivity boost to those who can deploy them. There is positive momentum on the regulatory front. Weinstein notes that the release of the Commission’s Digital Omnibus is a “step in the right direction” to help businesses compete globally. The goal now is harmonisation; creating a clearer and simpler regime that allows companies to train models responsibly and launch products faster. A unified market with clear and sensible rules will be the catalyst that turns potential into GDP. Investing in the workforce The final piece of the puzzle is people. Seizing this moment requires a workforce confident in using it. Weinstein stresses that we need leaders who can identify opportunities and managers who are AI-literate. This is happening through partnership. Google has already helped over 15 million Europeans learn digital skills and is now rolling out a €15 million AI Opportunity Fund to support vulnerable workers. For enterprise leaders, the message is clear: investing in skills today builds the confidence to take risks and grow tomorrow. Europe has the talent, the values, and the infrastructure. With the right focus on skills and a push for harmonised access to tools, Europe is well-positioned to lead the way and capture the full value of the AI era. See also: How the Royal Navy is using AI to cut its recruitment workload Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events including the Cyber Security Expo. Click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post How Europe’s talent can secure a trillion-euro AI economic injection appeared first on AI News. View the full article

AI spending in Asia Pacific continues to rise, yet many companies still struggle to get value from their AI projects. Much of this comes down to the infrastructure that supports AI, as most systems are not built to run inference at the speed or scale real applications need. Industry studies show many projects miss their ROI goals even after heavy investment in GenAI tools because of the issue. The gap shows how much AI infrastructure influences performance, cost, and the ability to scale real-world deployments in the region. Akamai is trying to address this challenge with Inference Cloud, built with NVIDIA and powered by the latest Blackwell GPUs. The idea is simple: if most AI applications need to make decisions in real time, then those decisions should be made close to users rather than in distant data centres. That shift, Akamai claims, can help companies manage cost, reduce delays, and support AI services that depend on split-second responses. Jay Jenkins, CTO of Cloud Computing at Akamai, explained to AI News why this moment is forcing enterprises to rethink how they deploy AI and why inference, not training, has become the real bottleneck. Why AI projects struggle without the right infrastructure Jenkins says the gap between experimentation and full-scale deployment is much wider than many organisations expect. “Many AI initiatives fail to deliver on expected business value because enterprises often underestimate the gap between experimentation and production,” he says. Even with strong interest in GenAI, large infrastructure bills, high latency, and the difficulty of running models at scale often block progress. Jay Jenkins, CTO of Cloud Computing at Akamai. Most companies still rely on centralised clouds and large GPU clusters. But as use grows, these setups become too expensive, especially in regions far from major cloud zones. Latency also becomes a major issue when models have to run multiple steps of inference over long distances. “AI is only as powerful as the infrastructure and architecture it runs on,” Jenkins says, adding that latency often weakens the user experience and the value the business hoped to deliver. He also points to multi-cloud setups, complex data rules, and growing compliance needs as common hurdles that slow the move from pilot projects to production. Why inference now demands more attention than training Across Asia Pacific, AI adoption is shifting from small pilots to real deployments in apps and services. Jenkins notes that as this happens, day-to-day inference – not the occasional training cycle – is what consumes most computing power. With many organisations rolling out language, vision, and multimodal models in multiple markets, the demand for fast and reliable inference is rising faster than expected. This is why inference has become the main constraint in the region. Models now need to operate in different languages, regulations, and data environments, often in real time. That puts enormous pressure on centralised systems that were never designed for this level of responsiveness. How edge infrastructure improves AI performance and cost Jenkins says moving inference closer to users, devices, or agents can reshape the cost equation. Doing so shortens the distance data must travel and allows models to respond faster. It also avoids the cost of routing huge volumes of data between major cloud hubs. Physical AI systems – robots, autonomous machines, or smart city tools – depend on decisions made in milliseconds. When inference runs distantly, these systems don’t work as expected. The savings from more localised deployments can also be substantial. Jenkins says Akamai analysis shows enterprises in India and Vietnam see large reductions in the cost of running image-generation models when workloads are placed at the edge, rather than centralised clouds. Better GPU use and lower egress fees played a major role in those savings. Where edge-based AI is gaining traction Early demand for edge inference is strongest from industries where even small delays can affect revenue, safety, or user engagement. Retail and e-commerce are among the first adopters because shoppers often abandon slow experiences. Personalised recommendations, search, and multimodal shopping tools all perform better when inference is local and fast. Finance is another area where latency directly affects value. Jenkins says workloads like fraud checks, payment approval, and transaction scoring rely on chains of AI decisions that should happen in milliseconds. Running inference closer to where data is created helps financial firms move faster and keeps data inside regulatory borders. Why cloud and GPU partnerships matter more now As AI workloads grow, companies need infrastructure that can keep up. Jenkins says this has pushed cloud providers and GPU makers into closer collaboration. Akamai’s work with NVIDIA is one example, with GPUs, DPUs, and AI software deployed in thousands of edge locations. The idea is to build an “AI delivery network” that spreads inference across many sites instead of concentrating everything in a few regions. This helps with performance, but it also supports compliance. Jenkins notes that almost half of large APAC organisations struggle with differing data rules across markets, which makes local processing more important. Emerging partnerships are now shaping the next phase of AI infrastructure in the region, especially for workloads that depend on low-latency responses. Security is built into these systems from the start, Jenkins says. Zero-trust controls, data-aware routing, and protections against fraud and bots are becoming standard parts of the technology stacks on offer. The infrastructure needed to support agentic AI and automation Running agentic systems – which make many decisions in sequence – needs infrastructure that can operate at millisecond speeds. Jenkins believes the region’s diversity makes this harder but not impossible. Countries differ widely in connectivity, rules, and technical readiness, so AI workloads must be flexible enough to run where it makes the most sense. He points to research showing that most enterprises in the region already use public cloud in production, but many expect to rely on edge services by 2027. That shift will require infrastructure that can hold data in-country, route tasks to the closest suitable location, and keep functioning when networks are unstable. What companies need to prepare for next As inference moves to the edge, companies will need new ways to manage operations. Jenkins says organisations should expect a more distributed AI lifecycle, where models are updated across many sites. This requires better orchestration and strong visibility into performance, cost, and errors in core and edge systems. Data governance becomes more complex but also more manageable when processing stays local. Half of the region’s large enterprises already struggle with the variance in regulations, so placing inference closer to where data is generated can help. Security also needs more attention. While spreading inference to the edge can improve resilience, it also means every site must be secured. Firms need to protect APIs, data pipelines, and guard against fraud or bot attacks. Jenkins notes that many financial institutions already rely on Akamai’s controls in these areas. (Photo by Igor Omilaev) Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and co-located with other leading technology events. Click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post APAC enterprises move AI infrastructure to edge as inference costs rise appeared first on AI News. View the full article

We’ve all seen the headlines: a third of US college students say they use ChatGPT for writing tasks at least once a month. The share of US teens turning to the same tool for schoolwork doubled between 2023 and 2024. Generative AI tools overall are a fixture of life for seven out of ten teens. The advent of ChatGPT and its competitors was supposed to put even the best essay writing services out of business. After all, generative AI can create an essay in seconds. So, why pay a professional to take care of it? Yet, three years after the launch of ChatGPT, academic help services are still going strong. Here’s why US students continue to choose expert help over AI-generated content, and the four services they trust with their assignments. How students actually use AI tools When it first made the news, ChatGPT was called “the death of the English essay.” Now, that kind of language seems like a promise of an apocalypse that (predictably, in hindsight) never came. Today, students don’t use generative AI tools to generate whole essays. Across multiple surveys, brainstorming, outlining, research, and test prep emerge as the main use cases for AI. For example, the survey from University of California Irvine found that: 66% use AI to learn more on a specific topic/subject 56% use it to prepare for tests 55% use it to find academic sources 46% use it for note-taking Only a third of respondents (31%) reported turning to AI tools to write essays. The percentage went even lower for scholarship and college application essays (21%). Why students still opt for top essay writing services While AI tools are great at generating long texts in a blink of an eye for free, that’s where their benefits typically end. Unlike professional writers, AI simply can’t: Grasp all the intricacies and subtleties of the expectations toward an essay, especially if it’s meant for a college or scholarship application Capture the customer’s authentic voice based on samples of their previous writing Write an essay that’s truly distinct and memorable: AI tools regurgitate cliché narratives and generic statements Come up with qualitatively new ideas and arguments: AI can only repeat the opinions already out there Verify the essay is 100% factually correct: AI tools can hallucinate facts, and many don’t even include precise sources of information Potential AI checks are another concern that pushes some students to hire a top essay writing service instead of using AI. For one, Turnitin automatically checks all assignments for both plagiarism and AI content now. Some educators take it on themselves to run AI content scans, too. An essay written by a professional will pass those checks without a hitch, which can’t be said about an AI-generated one. 4 best online essay writing services students trust Which platforms score the highest among the best online essay writing services trusted by US students? Here’s your snapshot of four such platforms: ServiceBest forRating (Sitejabber)EssayProOne-stop help4.4/5 based on 31,122 reviewsWritePaperIn-depth research5.0/5 based on 1,019 reviewsMyPaperHelpPersonalised writing4.8/5 based on 364 reviewsPaperWriterCollaborative approach4.9/5 based on 848 reviews EssayPro: Best for one-stop help EssayPro is the essay writing service online with the most extensive track record on this list. As of writing, it has over 30,000 reviews on Sitejabber alone and completes 300,000+ assignments annually. But that’s not what makes it the best essay writing service. Based on customer reviews, students prefer EssayPro to AI because its essay help remains affordable, all while being more in-depth, insightful, and creative than AI content. The fact that its writers specialise in 140+ subjects and 50+ paper types helped EssayPro secure its popularity, too. Pricing Custom writing: Starts at $10.80/page Rewriting: Starts at $7.56/page Editing: Starts at $7.56/page Proofreading: Starts at $5.40/page Pros 350+ writers specialising in 140+ subjects Good price-quality ratio with transparent pricing and no surprise fees Solid track record of on-time delivery Free plagiarism and AI reports Full control over who works on your essay Cons Not all orders can be completed in 3 hours No over-the-phone customer support WritePaper: Best for in-depth research AI tools can’t do the kind of research, analysis, and synthesis that experts at WritePaper do every day. That’s what makes it the best college essay writing service for essays and other papers that have to be insightful and present advanced, nuanced arguments on a complex topic. According to the best essay writing service reviews, WritePaper’s experts are especially good at delivering in-depth essays in research-intensive disciplines that require advanced reasoning. Those include nursing, philosophy, psychology, and history. Pricing Custom writing: Starts at $10.80/page Rewriting: Starts at $7.56/page Editing: Starts at $7.56/page Proofreading: Starts at $5.40/page Pros Solid argumentation and research skills among writers 115+ subjects covered Around-the-clock support and help Thoroughly researched essays with advanced reasoning Diverse formatting options (MLA, APA, Chicago, etc.) Cons Potentially overwhelming writer selection process Graphs and tables cost extra MyPaperHelp: Best for personalised writing While all services on this list provide custom writing services, MyPaperHelp is a paper writing service frequently praised for its personalised approach to orders. Its experts readily work with samples and adapt the style and tone of voice to the essay’s context and purpose. They also build on the ideas, suggestions, and whole outlines added to the order form. This makes MyPaperHelp the best essay writing website for any essay that has to be highly personal in nature. Think scholarship and college application essays or creative writing assignments that focus on personal experiences rather than academic research. Pricing Custom writing: Starts at $10.80/page Rewriting: Starts at $7.56/page Editing: Starts at $7.56/page Proofreading: Starts at $5.40/page Pros Wide range of writing styles and tones of voice supported Essays fully adapted to their context and purpose High-quality creative writing assignments Possible to attach samples to the order form that writers build on Unique, authentic writing that doesn’t rehash generic ideas or clichés Cons You have to be very precise with your instructions to leave no room for misunderstandings Originality reports aren’t provided by default; you have to request one (although they are free) PaperWriter: Best for collaborative approach Yes, US students turn to PaperWriter for many reasons, but direct writer communication is the most frequently cited one. So, if two-way communication with the writer is important, PaperWriter is definitely worth considering. PaperWriter’s experts routinely reach out to customers via direct chat whenever they need to clarify the requirements or ask for additional information. That makes PaperWriter the best essay writing service for students who want their essays to reflect their thoughts, ideas, and opinions to the letter. Pricing Custom writing: Starts at $10.80/page Rewriting: Starts at $7.56/page Editing: Starts at $7.56/page Proofreading: Starts at $5.40/page Pros Direct writer chat with end-to-end encryption Strict privacy policy that protects confidentiality Responsive writers who proactively communicate with customers Unlimited free revisions without mandatory waiting time Review work services available Cons Writer selection may be a bit time-consuming A collaborative approach is, by definition, also somewhat time-consuming Final thoughts: AI can’t rival human creativity & expertise AI tools may be becoming more ingrained in the learning process, but that doesn’t mean they’re ready to replace human creativity and expertise altogether. Yes, they can help you outline an essay or brainstorm ideas. But only professionals can come up with truly fresh ideas or develop a complex argument on a topic that requires hours of research. So, it’s safe to say that essay writing services aren’t going anywhere any time soon. They will continue supporting students in their studies, more so than AI tools. If you’re looking for the best essay writing service, Reddit and other social media platforms are a good place to start. Independent review platforms like Sitejabber and Reviews.io can also come in handy. Image source: Unsplash The post 4 best essay writing websites students choose over AI appeared first on AI News. View the full article