12 results for “CUDA”
NVIDIA released CUDA Toolkit 13.4 with native Windows on Arm support and a Rubin architecture preview. This gives developers a runway to port CUDA-dependent AI and content apps to Arm64 Windows.
Nvidia has formally opened the cuFile APIs and the vertical storage stack beneath them, moving the interface that lets GPUs read and write storage. Opening cuFile is a strategic departure; this layer has historically lived inside CUDA.
PyTorch Foundation published the Core PyTorch track for PyTorch Conference North America 2026 in San Jose on October 20–21. The lineup centers on torch.compile internals, cross-repo CI for out-of-tree backends, and distributed APIs infrastructure teams will ship against.
Nokia unveiled the industry's first commercial AI-native radio access network (AI-RAN) platform on July 15, built on its anyRAN software and NVIDIA's. Radio access networks account for the largest share of telecom infrastructure spending.
Advanced Micro Devices (AMD) shares rose sharply on Monday after the company announced a major expansion of its strategic partnership with Microsoft (MSFT). This announcement is a critical validation for AMD's AI strategy.
AMD has agreed to invest up to $5 billion in Anthropic under a sweeping infrastructure agreement that ties the chipmaker's equity commitment to. The agreement underscores a shifting dynamic in AI infrastructure procurement: major model.
On 2 September 2026 the PyTorch Foundation released 2.14: nccl2 and NVGEMM land in-tree, fault tolerance becomes a first-class c10d feature, and Apple Silicon gains native linear algebra plus Metal kernels that close an 8.5× single-token decode gap versus the prior MPS path.
Amazon and Nvidia expanded their partnership with a 2 million GPU order for 2027-2028 delivery, tripling the previous commitment. The deal spans Blackwell Ultra, Rubin, and Vera CPUs while Amazon simultaneously builds competing Trainium chips.
NVIDIA Dynamo now offers a preview shadow engine recovery feature that cuts LLM failover from 283 seconds to 7.3 seconds on B200 GPUs. The GPU Memory Service keeps weights resident across process crashes, letting a pre-warmed standby take over almost instantly.
AMD launched its Helios rack-scale AI system at Advancing AI 2026, combining 72 MI455X GPUs with 18 EPYC Venice CPUs to challenge Nvidia's Vera Rubin platform. The system will begin shipping to customers including Microsoft, Meta, and OpenAI later in 2026.
In the high-stakes race to dominate artificial intelligence, two of the world's most valuable companies are pursuing fundamentally different strategies. Apple, known for its consumer devices, is reportedly taking a more cautious approach.
NVIDIA committed to a 20-year land, power, and shell agreement at the PORTS-Pike campus in Ohio, securing 4.25 GW of AI factory capacity for OpenAI as anchor tenant. The deal frames physical infrastructure as a reusable asset across multiple GPU generations, unlocking an estimated $600 billion in NVIDIA compute revenue through 2030.