Artificial Intelligence News
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
Arm is arguing that countries building sovereign AI should prioritise a flexible technology ecosystem, scalable computing and power efficiency, not simply buy more infrastructure. Its proposed approach gives governments several routes to develop local capability, from designing chips themselves to using Arm’s ready-made silicon.
Read the report & discussion ↗Watch Desk
The AI economy depends on more than chips, code and data centres: it also needs materials such as helium, yttrium and silicon carbide. In The Heart of the Matter, the author argues that physical supply constraints can limit how quickly the industry expands.
vLLM Watch
vLLM has added bi-directional KV-cache transfer for multi-turn AI workloads, letting prefill servers reuse context held by decode servers instead of repeatedly computing it. In AWS’s benchmark, the change cut time to first token by up to 3.3 times for one tested model, though the gains depend on…
Meta AI Watch
Meta has unveiled new networking designs for AI data centres, including a liquid-cooled switch rack and a fabric intended to connect larger GPU clusters with fewer network hops. The announcement also includes open implementations and a new standards initiative, putting infrastructure choices that are usually buried in the…
Meta AI Watch
Meta says it is using otherwise idle capacity across its inference fleet to improve ad recommendations, while keeping the computing power reserved for ordinary service protected. The approach offers a practical way to do more AI work with hardware already in the system, rather than ordering more capacity.
Intel AI Watch
Intel has outlined a broad data-centre push for AI agents and inference, combining Xeon rack designs, a new GPU and software intended to work across its CPUs and accelerators. The standout detail for AI operators is the GPU’s memory capacity: Intel says its own card will have 160GB,…
NVIDIA Watch
NVIDIA and Princeton’s Dao AI Lab have developed ReplaySSM, a method for speeding up inference in hybrid AI models by changing how their state is stored. NVIDIA reports up to 1.80× faster SSM decoding kernels at batch size 256, with larger gains for speculative decoding on tested 120B-class…
Palantir Technologies Watch
Ondas and Palantir have formed Regna AI, a new venture to develop specialised AI models for governments, defence organisations and businesses. Its first planned deployment puts AI on connected devices to help coordinate autonomous missions, including when communications are limited or disrupted.
Watch Desk
ASUS has unveiled a compact server built around NVIDIA Vera CPUs for agentic AI and reinforcement-learning workloads. The XA P2N-E2 is designed to handle the less glamorous, essential work around AI models: running code, using tools, sandboxing tasks and processing data.
NVIDIA Watch
NVIDIA and Cisco have announced a new pluggable optical module design aimed at connecting more computing equipment inside AI data centres, with up to 12.8 terabits per second of capacity. The proposed USFP standard targets more bandwidth in a familiar-sized module, potentially helping AI operators scale networks without…
Watch Desk
Taiwan-based AI compute operator KONST says it has raised $30 million in Series B funding led by ADATA Technology, with the money earmarked for expanding its AI data-centre buildout across Asia. The raise backs a business that aims to sell not just computing capacity, but the tools to…
Unsloth Watch
Unsloth’s new beta lets local and hosted models browse the web, while adding several useful tools for training and running models on personal hardware. The browser controls and performance changes make this a substantial update for people building local AI workflows, not just another dot on a release…
Alibaba Qwen Watch
Alibaba researchers have detailed QwenGyre, a framework for training AI agents on long software-engineering tasks. It dynamically shifts GPUs between generating agent runs and training, targeting expensive stretches when hardware would otherwise sit idle.
Microsoft AI Watch
Microsoft plans to deploy its Maia 300 custom AI chip at multi-gigawatt scale from 2027, positioning it as the primary inference system for its products and services. The announcement signals a major expansion of Microsoft’s in-house AI compute, with potential consequences for how its services are built and…
AMD Watch
AMD has unveiled Helion, an open-standard rack reference design aimed at AI data centres, with a CPU-only configuration capable of holding up to 32,768 Zen 6 cores. The rack is planned for the first half of 2027, giving server makers a design to build from rather than a…
Watch Desk
Rimac Technology and ECOBLOX are developing modular data-centre systems for AI computing, combining battery storage, power electronics, cooling and IT infrastructure. Their first planned installation, at Rimac’s campus near Zagreb, is due in November, putting the partnership’s infrastructure pitch on a real site rather than leaving it in…
Watch Desk
Around 300 people rallied in Caernarfon against plans for a large AI data centre near the Menai Strait, according to Wales Online. The proposal promises jobs, but has prompted questions about the scale of the development and its local impact.
Watch Desk
Cambridge start-up Neela Biotech has raised £2.1 million to develop an AI-guided process that turns food, farm and forestry waste into sustainable aviation fuel feedstock. The funding is intended to support a pilot-scale trial at a biogas plant over the next two years, putting existing infrastructure at the…
Advantech Watch
Advantech and Leopard Imaging are showing a machine-vision system that runs camera capture and AI inference locally on an edge computer. The demo combines four live camera feeds with person detection across eight additional video streams, without sending the video to the cloud.
Watch Desk
PhysicsX has established its Asia-Pacific hub in Singapore, where the London-based company aims to use AI models to speed up engineering simulations and industrial work. The expansion puts a substantial AI-for-engineering business closer to manufacturing and semiconductor customers, though its productivity claims still need to be judged by…
AMD Watch
AMD’s GAIA 0.25.0 release candidate brings a larger local model, a much leaner Windows desktop install and tighter checks around agent permissions and files. It is a substantial preview for testers, not the finished release, and several changes may require users to rebuild data after upgrading.
