A super intelligence buildout that reaches orbit, lands in China and breaks ground across Europe: we start the week with the infrastructure and the agents moving at the same time.
OpenAI plans a post midterm push for AI compute campuses. Reporting on November 1 from Axios and TechCrunch says OpenAI will launch a policy and public push after Tuesday’s midterms that reframes data centers as AI compute campuses. Proposals have been stockpiled since August, the push starts in the lame duck Congress opening November 9, and OpenAI is also moving ChatGPT and Codex to one shared usage allowance from December. Why it matters: the next contest for super intelligence is as much about power, permits and public consent as it is about models, and OpenAI is preparing to argue that case in Washington.
Apple Intelligence reaches China in December, with Qwen inside. Apple has confirmed a first batch of Apple Intelligence features will ship in China with iOS 27.3 in December, with developer beta testing starting November 4. The release includes Writing Tools, live translation and camera based Visual Intelligence, powered by Apple models alongside Alibaba’s Qwen through a local partnership. Why it matters: one of the world’s largest phone markets finally gets Apple’s on device intelligence, and it arrives in a distinctly local form shaped by regulation and partnership.
Google runs a data center chip in orbit. Google says a prototype satellite for Project Suncatcher, launched October 1, powered up a TPU in orbit and ran its Gemma 4 model, sending a generated haiku back to Earth. Google calls it the first AI response in orbit from a data center class TPU. The limits are real: about one kilowatt of power, one TPU at a time, and sessions of roughly 15 to 20 minutes. Why it matters: it is a small, verified first step toward computing in space, where sunlight is abundant and the hard problems become radiation and cooling.
DeepSeek open sources a super agent prototype. The DeepSeek Harness team has published a technical exploration of super agent architecture and open sourced its research prototype on GitHub under an MIT license. It is a prototype validating an approach, not a finished commercial product, and some designs are not yet fully implemented. Why it matters: open agent architecture gives builders everywhere a shared starting point to test, critique and improve, instead of leaving agent design locked inside a few labs.
Europe gets a new NVIDIA powered AI factory network. SWI Compute launched on November 2 with a target of 2.3 gigawatts of AI factory capacity by 2029 across sites including Ireland, Britain, Denmark, Spain and Italy. Its first large scale UK facility is planned online by the end of 2027, starting at 30 megawatts with nearly 7,000 NVIDIA Rubin GPUs, and the company says it has ordered about $1 billion in NVIDIA systems. Why it matters: Europe is building sovereign compute at factory scale, and the map of where super intelligence runs is widening beyond the United States.
