Latest news, developments and reporting on AI Chips & Compute.
Chip startup xMEMS Labs has started mass production of the XMC-2400 and XMC-4800, two solid-state coolers that use a vibrating silicon membrane to move air in AI glasses, smartphones and data center storage.
Kearney ranks the UK among Europe’s slowest-growing data center markets, forecasting 17.6% annual demand growth to 2030, versus 36% in Ireland. High land and energy costs and slow grid connections are cited as obstacles.
Microsoft has opened preorders for the Surface Laptop Ultra RTX Spark, an AI-focused Windows 11 laptop starting at $2,599.99. TechRadar also named top 2026 Windows 11 laptops and a $1,449 Gigabyte RTX 5070 Ti deal.
TechRadar published an analysis warning that enterprise IT remains optimized for AI training while inference becomes the dominant workload. It cites IDC, Deloitte, McKinsey and Gartner on inference demand, agent growth, network strain and the rising role of CPUs.
The IMF warned Tuesday that inflation, debt and AI-related borrowing could cause sovereign bond yields to spike in a disorderly way, while global resilience is nearing its limits. DBS's CEO said Singapore's predictability is an advantage as Asian tech shares fell on AI financing worries.
Seoul's KOSPI fell 5.46% to 6,035.35 on Oct. 13, extending losses to a fifth session as technology and chip stocks slumped on renewed AI and inflation worries. The won weakened to 1,345.1 per dollar after foreigners sold a net 4.51 trillion won of local shares.
Lee stresses U.S. alliance; North Korea reports missile test; Seoul stocks tumble; IMF lifts Korea 2026 growth forecast.
Asian technology and semiconductor shares fell Tuesday, tracking a U.S. chip sell-off, as SK Hynix, Samsung and SoftBank slumped amid concerns over AI financing and competition.
The IMF raised its 2026 global growth forecast to 3.1%, citing AI investment that offset Middle East war shocks, while warning of disorderly bond-yield spikes and a possible AI share-price correction. It lifted South Korea's growth outlook to 3.3% on semiconductor exports.
Samsung Electronics unveiled next-generation memory and storage strategies at OCP Global Summit 2026 in San Francisco, targeting AI infrastructure performance. The company outlined high-performance memory, CXL-based expansion and next-generation SSD storage, and plans a joint CXL demonstration with AMD.
Microsoft said it will expand Azure Maia AI systems to several gigawatts from 2027 and make them its main inference platform, while OnePlus, Google, CMF, LineageOS and Meta announced new devices, software and AI models.
Applied Materials has opened a $5 billion semiconductor research facility in Silicon Valley, with partners including TSMC, Intel, Samsung, Micron and SK hynix, aiming to cut development cycles by at least 30%.
Intel CEO Lip-Bu Tan and Broadcom CEO Hock Tan said at a Singapore financial conference that capital access and cost of capital are becoming major constraints on AI expansion, even as chip shortages ease. They also discussed AI training costs, open-source models and quantum computing as funding continued for AI infrastructure and robotics startups.
Kioxia announced an upgraded GP-series SSD that exceeds 20 million random-read IOPS using third-generation XL-FLASH memory, and also exhibited the LD4 QLC E1.L drive with PCIe 5. TechRadar reported the new GP drive roughly doubles the speed of the August 2026 GP1.
Alibaba's Qwen team has introduced QwenGyre, an online reinforcement learning framework that trains coding agents on rollouts running for hours and consuming close to one million tokens, with reported end-to-end speedups of up to 1.85x over a colocated baseline.
At Dell’s AI Leadership Symposium, executives said enterprise AI governance is shifting to runtime enforcement as always-on agents spread, while cloud token budgets are running out faster than planned and pushing on-premises and open-weight models back into deployment plans. The discussions were broadcast by SiliconANGLE Media’s theCUBE, according to SiliconANGLE.
Microsoft introduced Maia 300, a four-chiplet inference accelerator that it says can process 40% more tokens per megawatt than leading GPU-based systems. The chip uses HBM4 memory, a dataflow architecture and Ethernet-linked superpods, with production deployment planned for next year.
Apple is expected to launch a 14-inch M6 MacBook Pro later this month, with 9to5Mac reporting a dual 16-core Neural Engine, a 12-core CPU and GPU, and a 2nm process as the chip's main changes.
Dell is adapting its AI partner ecosystem for Silicon Valley customers who prefer building their own stacks, a Dell executive says in an interview on theCUBE.
Amazon is reportedly in talks to acquire AI startup Decart for up to $7 billion. Decart's DOS software optimizes AI models for specific chips, potentially aiding AWS and its Trainium push.
IDC: Q3 2026 PC shipments fell 20.1% to 62.7 million units, the worst third quarter on record, as AI-driven memory costs squeeze demand.
Cisco and Nvidia unveiled silicon, open rack designs and optical modules at OCP Summit to ease AI data-center scaling.
A US custom PC maker says retail GPU prices are set to rise significantly, as reports point to higher AMD VRAM costs, rising component prices and Nvidia diverting GB202 chips away from the RTX 5090.
Microsoft unveiled Maia 300, its third-generation AI chip, aiming to cut reliance on Nvidia for AI inference.