This was a week where AI's physical and financial ambitions both got bigger, with infrastructure deals and hardware announcements doing more talking than any single model release.
The biggest story was consolidation: Nvidia agreed to acquire Hugging Face for roughly $12.9 billion, nearly triple the open-source model hub's 2023 valuation and Nvidia's largest acquisition ever. The deal pushes Nvidia further from pure silicon into cloud and developer tooling, tightening its grip on AI infrastructure just as rivals build out their own stacks. Anthropic, meanwhile, opened a research preview of its Model Hardware Standard, a shared specification letting Claude and other agents safely operate physical equipment like robots and lab instruments — initially scoped to research labs and manufacturers so safeguards can be tested before a wider rollout.
OpenAI had a two-track week on infrastructure and product. It launched ChatGPT Work, a $20-per-month tier aimed at general office workers rather than engineers, reporting 98% internal adoption but acknowledging that non-technical workflows need simpler permissions and configuration than its developer tools. On the hardware side, early benchmarks for OpenAI's first custom inference chip, Jalapeño (built with Broadcom), showed 1.5x-1.9x higher throughput and up to 3.6x lower latency than current GPU systems on models like GPT-OSS-120B and DeepSeek R1, at notably lower power per rack — a clear signal OpenAI wants to control its own inference economics rather than depend entirely on Nvidia.
Apple pushed local AI compute forward with the M6, its first 2-nanometer chip, and the quad-die M5 Ultra, which delivers up to 4.5x the peak GPU AI performance of the M3 Ultra and supports 512GB of unified memory for running large models entirely on-device. On the scrutiny side, Anthropic's pitch to IPO investors — a $30 trillion total addressable market estimate ahead of a targeted $2 trillion valuation this October — drew skepticism from analysts who note current revenue run rates fall well short of what that math implies. And Microsoft expanded Defender with new capabilities to assess risk posture for AI agents themselves, reflecting growing enterprise anxiety about agents operating with real permissions inside business systems.
To watch next week: how markets and regulators react to the Nvidia-Hugging Face deal, and whether more detail emerges on Anthropic's IPO numbers as October approaches.