From the archive
AI hardware provides the computing power, memory and networks used to train and run artificial intelligence models. Follow news on GPUs, AI accelerators, data centres, chip design and energy demand. Read how changes in infrastructure affect performance, availability and cost, and distinguish measured results from vendor claims about what the next generation of computing will deliver.
Lead story
Amazon has joined the AI safety argument at last, and it has picked a careful spot to stand in. The company says models should be released only when they are safe and fully tested, but it has stopped short of backing the slowdown call that swept up its…
Read the report & discussion ↗Watch Desk
The AI industry has quietly changed what it does for a living: less building models, more running them. That shift from training to inference is now forcing a rethink of the chips and memory everything else depends on, and it is redrawing who buys silicon from whom. The…
Intel AI Watch
Intel is meeting only about half of the CPU demand its customers are asking for, according to the company's own chief executive, who says he has been apologising to the bosses who call him about it. Lip-Bu Tan tied the shortfall to AI on the livestream of Splunk's…
Watch Desk
The Federal Reserve has raised interest rates for the first time in three years, and AI's build-out took it where it hurts. The sector leans on borrowed money, from hyperscaler campuses to would-be neoclouds, and short and medium-term borrowing costs have just gone up, with another rise projected…
TSMC Watch
Samsung Electronics is drawing up plans to hand commodity DDR5 production to outside partners, and weighing whether the base dies inside next-generation HBM4 should come from TSMC, according to industry sources. The quietly enormous point: the one-stop, everything-in-house model that defined the memory industry is cracking under AI…
Watch Desk
The House has voted 417 to 3 to pass the Ratepayer Protection Act, a bipartisan bill that would require state utility regulators to consider rules making data centres cover some of the costs of the new power generation they set off. After months of Congress arguing about artificial…
Watch Desk
Qualcomm's plan to sell AI silicon to Amazon has acquired the two things investor stories need most: dates and digits. Shares in the chipmaker climbed to a two-month high as details of its multi-generational AI data-centre deal with Amazon emerged, and the numbers are substantial. Qualcomm expects roughly…
AMD Watch
AMD has filed its broadest MLPerf Inference submission ever, and it has arrived with fighting words attached. The company says its Instinct MI355X GPUs beat selected NVIDIA B200 and B300 submissions on the GPT-OSS-120B language model at eight GPUs, led the 72-GPU GB200 result too, and the brand…
Cerebras Watch
Cerebras CEO Andrew Feldman says the company’s route through the AI-chip market is not a slightly cheaper Nvidia alternative. In an interview with Jack Altman, Feldman argued that a challenger needs a radically better product to survive the years it takes to design, manufacture and scale new hardware.…
Watch Desk
AI workloads are forcing technology leaders to rethink long-standing AWS, Azure and Google Cloud commitments, as GPU scarcity, shifting prices and data-transfer costs make production plans harder to forecast. The practical lesson is blunt: an AI roadmap is not financially credible until its workload, location and capacity assumptions…
Watch Desk
Cisco has unveiled a broad Splunk expansion spanning private AI infrastructure, agent-cost monitoring and security automation. The central idea is refreshingly practical: companies need somewhere controlled to run models, a way to see what agents are costing them and better tools for handling faster attacks.
AWS AI Watch
AWS has published a practical route for teaching Qwen3-8B to tag product catalogues consistently. The walkthrough uses SageMaker to fine-tune the model on a nine-category schema, optimise it with verifiable rewards, then process catalogue jobs asynchronously. That is a more concrete proposition than simply asking a frontier model…
Hugging Face Watch
Getting LeRobot running on an NVIDIA Jetson Orin is less a pip install than a small expedition through Python versions, CUDA libraries and C++ ABI compatibility. A new Hugging Face Community Article lays out the author's route to a working container, including the traps that turned a supposedly…
Groq Watch
Groq is rebuilding itself as an AI-inference cloud after its signature chip architecture and much of its engineering leadership moved to NVIDIA. The company still has substantial infrastructure and customers, but its new reliance on NVIDIA systems makes the old challenger look rather more like a specialised cloud…
Watch Desk
Cornelis Networks is opening a new front in AI infrastructure: the Intel spin-off says it is moving into scale-up networking for rack-scale systems, raising $205 million and exploring a collaboration with Qualcomm. The important bit is strategic, not merely financial: Cornelis wants to give AI-platform builders an open…
Watch Desk
Pinterest says its new multimodal AI infrastructure makes visual search 7.3 times faster overall and starts responses 85 times faster in benchmark testing. Built with NVIDIA technology, the system is intended to improve discovery, shopping, content understanding and safety across a platform handling more than 80 billion searches…
NVIDIA Watch
NVIDIA says it has made dropless Mixture-of-Experts training dramatically faster in JAX, taking DeepSeek-V3 from 103 to 1,068 TFLOPS per GPU on GB200. The company also reports 97% scaling efficiency across 1,024 GPUs on GB300 NVL72 hardware, turning a specialist optimisation story into a useful glimpse of where…
