Nvidia filed an 8-K on September 2 confirming an $11.9 billion cash purchase of Hugging Face, plus roughly $1 billion in equity-retention grants for staff joining the deal, expected to close in the first half of 2027 pending regulatory review.
Hugging Face CEO Clement Delangue told CNBC that he approached Jensen Huang over the summer because open-source AI "needed more resources, more scale, more visibility." Nvidia has pledged to keep the hub open to competing chipmakers rather than lock it to its own silicon.
The purchase is Nvidia's second-largest ever, trailing only its $20 billion Groq asset acquisition, and values Hugging Face at roughly $150 million in annualized revenue — a striking multiple that underscores how central the model-hosting layer has become to the AI stack.
The Institute of Foundation Models at MBZUAI released K2 Horizon on September 3, a fleet of six Apache 2.0-licensed models sized 0.9B, 3.7B, 7B, 32B, 36B-A4B, and 375B-A23B parameters.
IFM claims the 0.9B, 3.7B, and 7B tiers hit state-of-the-art results at their scale on reasoning, math, coding, and agentic benchmarks. All six ship with day-zero support across vLLM, SGLang, Ollama, and Unsloth, lowering the barrier for smaller labs and independent builders to deploy frontier-adjacent capability without training from scratch.
The release lands days after Nvidia's Hugging Face bid, sharpening the question of how a chipmaker-owned hub will coexist with a growing wave of fully open model fleets that anyone can host on any hardware.
Tesla registered 45 Cybercabs with the Texas DMV ahead of Wednesday's debut at Gigafactory Texas, where the no-steering-wheel, no-pedals vehicle gave invite-only rides through downtown Austin.
The Robotaxi app now supports Cybercab hails alongside its existing fleet of modified Model Ys, though Tesla has yet to say when the low-dozens pilot fleet will scale toward a full commercial rollout.
Sen. Bernie Sanders and Rep. Greg Casar introduced legislation that would permanently prohibit development of superintelligent AI and impose a temporary halt on advanced frontier research until Congress establishes federal safety regulations.
The bill would also stand up a cabinet-level federal AI oversight agency and direct the administration to pursue international agreements against superintelligence development — a marker of how far AI-safety concerns have moved from research labs into active legislative text.
Nvidia unveiled Personal AI Router (PAIR) on September 3, a free, open-source tool that lets multiple PCs on the same network pool their GPUs into a single shared inference cluster — enabling parallel agentic workflows entirely offline.
The beta supports Windows, macOS, and Linux, signaling Nvidia's push to make local, privacy-preserving inference viable for everyday multi-device households rather than only data-center deployments.