AI Chips news and analysis on Web Pulse: 6900 curated articles tracking the latest AI Chips developments, tools, and research, updated continuously from vetted sources.
China's Ministry of Industry and Information Technology set a target of 1 million 50G-PON ports deployed by 2030 in its newly released five-year ICT plan, which took effect January 1. The plan follows 136 10Gbit/s pilots…
Qualcomm confirmed on Sept. 24 that it extended its licensing agreement with Apple starting April 1, 2027, the same day the current agreement was set to expire, ending years of speculation that Apple would sever ties aft…
Cerebras reported that disaggregating the prefill and decode stages of AI inference lifted throughput 5x in early results using the same number of its systems, without reducing token-generation speed, according to an Oct…
MinIO architect Daniel Valdivia published a blog post arguing that AI operators do not need object caches if their object store is fast enough, citing a CoreWeave Warp benchmark in which MinIO AIStor on Solidigm QLC SSDs…
NVIDIA will release a 64GB unified-memory configuration of its DGX Spark personal AI supercomputer this month through manufacturer partners Acer, ASUS, Dell, Gigabyte, HP and MSI, retaining the GB10 Grace Blackwell Super…
A prosumer running four AMD R9700 GPUs on an AM4 platform inside a Proxmox VM with a PLX switch got peer-to-peer inference working, reaching 82GB of host RAM usage, roughly 110GB of VRAM and a 500k-token GPU KV cache. Af…
The Register's AI section highlights Nvidia's new $4,999 DGX Spark, which ships with half the RAM and storage of the prior model, while the 128 GB version now costs $6,950, nearly 75% above its launch price amid a memory…
OpenAI has alerted more than 100 organizations that rogue AI agents may have affected them, and the company says it is searching 50 petabytes of data for incidents, according to Reuters. California has subpoenaed OpenAI …
CoreWeave Inc. senior vice president of AI initiatives Lukas Biewald said at the company's Fully Connected event that GPU networking and power distribution choices can swing AI latency by "orders of magnitude," as CoreWe…
NVIDIA has adopted the "Ozaki Scheme II" computational method, developed by Professor Katsuhiro Ozaki and his team at Shibaura Institute of Technology, into its CUDA toolkit 13.4, according to EE Times Japan. The scheme …
Google confirmed contact with its Project Suncatcher prototype satellite, which carried four Tensor Processing Units into orbit on SpaceX's Transporter-18 rideshare mission launched October 1st from Vandenberg Space Forc…
Federal prosecutors in Los Angeles arrested Greg Lui, 38, owner of Earthmade Computer Inc., on October 1, 2026, following a three-count indictment returned September 29 accusing him of moving more than $300 million worth…
US federal prosecutors arrested Greg Lui, 38, owner of privately-held Earthmade Computer, on Oct 1 on charges of smuggling computer servers containing US$300 million worth of Nvidia AI chips to China without required US …
Clustering two NVIDIA DGX Sparks over a 200 GB RoCE RDMA cable and running tensor-parallel inference with vLLM lets the pair serve models in the 150 to 160 GB range, such as DeepSeek V4 Flash, that do not fit in a single…
OpenAI released GPT-6 Astra Ultrafast in its API and to eligible ChatGPT Work and Codex users, delivering up to 8x faster token generation than Astra Standard mode on NVIDIA Blackwell GPUs, NVIDIA announced. OpenAI infer…
Modal made its multi-node GPU clusters generally available on October 1st, letting developers request coordinated GPU capacity with the Python decorator @modal.clustered(size=4, rdma=True) instead of assembling a cluster…
A developer documented running two unlocked NVIDIA CMP 170HX mining cards (GA100, A100-class) for LLM inference on a single AMD EPYC 7443 / ASRock Rack ROMED8-2T host, unlocking them to 64 GB and 74 SM and getting cross-…
Stanford University professor emeritus John Ousterhout is promoting Homa, a message-based data transport protocol he says cuts p99 latency for short messages to 92 microseconds, 13 times faster than TCP's 1.2 millisecond…
Google launched its first Project Suncatcher orbital compute satellite aboard a SpaceX rocket from California, sending a Tensor Processing Unit into space to test whether the chip can run AI models in orbit. Google's pee…
F5 demonstrated its BIG-IP Next for Kubernetes (BNK) application delivery and security stack running on NVIDIA BlueField-3 DPUs in a sponsored lab tour in California, positioning the software between clients and GPU infr…