A morning where open models, sovereign bets and very large cloud numbers all pointed the same way: capable super intelligence is becoming infrastructure, and everyone wants a piece of it on their own terms.
Mistral opens a trillion parameters. France’s Mistral released the downloadable weights for Large 4 on October 29, its 1 trillion parameter multimodal flagship trained in Europe, under a modified MIT licence priced at $0.70 per million input tokens for the hosted version. Mistral reports it effectively ties the open weight leader GLM-5.3 on coding while using 47 percent fewer tokens per task. Why it matters: open weights cannot be switched off remotely, so this is Europe’s strongest sovereignty play yet, and a real option for teams that want frontier capability on infrastructure they control.
Canada buys into Cohere. The Government of Canada announced on October 29 that it will take an equity stake of up to about C$1.6 billion in Cohere, in stages, backing the home grown lab as secure, sovereign AI. Why it matters: governments are moving from regulating super intelligence to owning a share of it, and Canada is putting serious capital behind a domestic champion.
The cloud giants report, and the spending is the story. Alphabet reported third quarter revenue of $127.6 billion on October 28, with Google Cloud revenue of $30.7 billion, up 103 percent, and $51.4 billion spent on property and equipment in the quarter alone, while Amazon reported third quarter revenue of $201.9 billion, up 12 percent, on strong AWS demand. Why it matters: the buildout is no longer a promise in an investor deck, it is the largest capital deployment in tech, quarter after quarter.
Gemini becomes a place to make things, not just ask things. On October 29 Google added the ability to create, preview and edit documents, PDFs, presentations, spreadsheets and code files directly inside the Gemini app. Why it matters: the assistant is quietly becoming the workspace itself, which changes where our drafting, deck building and everyday documents actually live.
OpenAI rewrites its own rulebook. OpenAI published a new Model Spec organized around five values, alongside fresh work on detecting misalignment in frontier models, as agent deployments accelerate. Why it matters: as agents start acting in real systems, the written rules for how they should behave, and how we catch them drifting, become part of the product, not fine print.
LG puts a date on the home robot. On its October 28 earnings call, LG Electronics said it plans to enter the home embodied AI and humanoid robot market by 2030. Why it matters: the race for the home is moving from screens and speakers to machines that can move through a room, and one of the world’s biggest appliance makers just set its timeline in public.
