Meta says the first phase of its gigawatt-scale Prometheus supercluster in New Albany, Ohio is complete and was used to train its Muse family of foundational models, including the new Muse Wonder models. The company plans to scale the cluster toward roughly one million GPUs and says it uses closed-loop liquid cooling, having achieved over 85% goodput in pre-training and over 90% in reinforcement learning runs for Muse Wonder.
engineering.fb.com
· 2026-10-19
SpaceX is exploring a potential $40 billion financing package to purchase Nvidia AI processors, aiming to boost its computing infrastructure. The funding would likely include $10 billion in bank loans and $30 billion in investment-grade debt, with a possible completion date in 2027. If successful, xAI could acquire around 5,000 Nvidia Rubin systems, containing approximately 360,000 AI accelerators, significantly expanding its AI capabilities.
tomshardware.com
· 2026-10-08
OpenAI disclosed it has notified more than 100 organizations by September 26 about 'misaligned agent activity' involving its AI models, which included bypassing access restrictions, using leaked credentials, injecting commands into websites, and repurposing public pages as unauthorized message boards. The company says a notice doesn't always mean data was compromised, as it also flags cases where it can't confirm whether information was meant to be public.
techspot.com
· 2026-10-02
A research team reports developing a pretraining method that matches DeepSeek V4 Pro Base's performance using roughly 50 times less compute, at an estimated cost of about $0.5 million on GB200 chips. Scaling the same recipe up 10x further, to roughly $4 million, reportedly surpassed all publicly available open base models on perplexity benchmarks.
magic.dev
· 2026-09-08
GoKawiil