gpu

PyTorch 2.15.0 (October 28, 2026) makes a rewritten NCCL backend the default; its notes list eager communicators that allocate GPU memory and say nccl-legacy will be removed in the next release.

NVIDIA's TensorRT-LLM v1.4.0rc0, published October 14, 2026, labels two API changes (BREAKING) in its release notes and lists five known issues.

vLLM 0.30.0 adds a GPU-resident weight cache for faster engine restarts and removes GPTQ g_idx; llm-d v0.10.0 then moved its base image to 0.30.0.

AWS and Nvidia announced an expanded partnership to deploy 2 million additional GPUs across AWS data centers in 2027-2028, alongside new Vera CPU, networking, and robotics integrations.

AMD's Adrenalin 26.6.2 driver brings FSR 4.1 machine-learning upscaling to RDNA 3 Radeon RX 7000 cards using an INT8 model, but frame generation remains an RDNA 4 exclusive.

Bloomberg reports Meta is standing up a cloud business under its 'Meta Compute' initiative to sell spare AI capacity; CoreWeave and Nebius shares fell 10.8% and 12.4%.

Intel's data-center inference GPU scales to 480GB of LPDDR5X in a 350W air-cooled PCIe card, sidestepping the HBM shortage that constrains Nvidia and AMD accelerators.

Nvidia's Q1 FY2027 results set new company records across revenue, net income, and free cash flow, driven by Blackwell GPU demand and a near-tripling of networking revenue.

Mesa 26.1.0, released May 6, 2026, delivers over 40 new Vulkan extensions across major GPU drivers, adds VirtIO-GPU native-context support for Intel, and marks the end of VirGL maintenance.

Washington has licensed Alibaba, Tencent, ByteDance, JD.com and distributors Lenovo and Foxconn to buy up to 75,000 H200s each. Not one chip has shipped — Beijing is pushing buyers toward Huawei.

The sustainable footwear brand once valued at $4 billion is pivoting to GPU-as-a-service after selling its shoe assets, raising questions about AI market froth.

AMD's Instinct MI400 family debuts three CDNA 5 accelerators on TSMC 2nm alongside the 72-GPU Helios rack, but a production timeline dispute clouds the second-half 2026 launch.