↗
IBM and OpenAI Form Enterprise AI Partnership With GPT-5.6
IBM will embed OpenAI's GPT-5.6, Codex, and ChatGPT Work into IBM Consulting Advantage and launch a dedicated OpenAI Practice.
N THE NEWSWIRE
Latest
dispatches.
Fresh signal from the moving edge of hardware, technology, and the systems underneath.
01 ALL STORIES
↗
↗
DeepSeek Ships Official V4 Pro With DSpark Decoding
DeepSeek ships the official V4 Pro release (0813) with MIT-licensed weights, DSpark speculative decoding, and large agentic benchmark gains.
↗
Nvidia Releases Nemotron 3.5 Lightning for Agentic AI
Nvidia's 30B-parameter Nemotron 3.5 Lightning open model targets long-running agentic workloads, joined by the NeMo Switchyard routing library.
↗
CME Group Sets October 5 Launch for GPU Compute Futures
CME Group will launch the first regulated compute futures on October 5, tracking Nvidia H100 and B200 rental prices on NYMEX.
↗
Sony and TSMC Finalize Image Sensor Joint Venture in Japan
Sony and TSMC signed a binding deal for a Kumamoto image sensor joint venture, with Sony as controlling shareholder and production planned for 2029.
↗
Nvidia and Six Investors Plan $500 Billion for AI Compute
Nvidia is building compute financing platforms with six major investors to mobilize over $500 billion for AI infrastructure.
↗
AMDs Ryzen 5 439 and Ryzen 7 449 Ship Without an NPU
AMD adds Ryzen 5 439 and Ryzen 7 449 mobile chips with Zen 5 cores but no NPU, aiming at cheaper mainstream laptops.
↗
Microsoft Plans Maia 300 AI Chip Launch as Soon as Next Month
Microsoft is reportedly preparing to unveil its Maia 300 AI chip as soon as next month, ramping production and courting cloud customers like Anthropic.
↗
Intel Launches $15 Billion Stock Offering for AI Growth
Intel is selling $15 billion in common stock to fund capital spending, citing surging AI compute demand, with underwriters able to buy $2.25 billion more.
↗
Meta Releases Muse Glimmer, a 30B Open-Weight Agentic Model
Meta's Muse Glimmer packs 30 billion parameters into an Apache 2.0 agentic model that runs locally, with a 131K-token context and 4-bit weights under 20 GB.