The last 24 hours brought a striking mix of corporate maneuvering and policy hesitation: Anthropic doubled down on compute, Google rolled out a cheaper frontier model, and the White House hit pause on its long-awaited AI executive order.
Anthropic locks in $1.25B-per-month SpaceX compute deal
Anthropic expanded its compute partnership with SpaceX, committing to spend roughly $1.25 billion per month through 2029 for access to the Colossus supercomputing infrastructure. The deal arrives as Anthropic tracks toward its first profitable quarter, with Q2 revenue projected to more than double to $10.9 billion. Read more on Axios.
Trump postpones AI executive order at the last minute
The White House delayed the signing of an executive order that would have created a voluntary federal review process for frontier AI models 90 days before release. President Trump said he “didn’t like certain aspects” of the draft and pulled back hours before the scheduled signing, despite weeks of negotiations with OpenAI, Anthropic and others. Full story at CNBC.
Google ships Gemini 3.5 Flash to general availability
Coming out of I/O week, Google moved Gemini 3.5 Flash to GA, priced at $1.50 / $9 per million tokens with a 1M-token context window. The model posts 76.2% on Terminal-Bench 2.1, beating Gemini 3.1 Pro on coding and agent tasks at roughly a third of the price of comparable frontier models. Details from CNBC.
Code with Claude London: agents go private
Anthropic wrapped its first dedicated European developer event with a wave of platform updates, including on-prem agent sandboxes, “MCP tunnels” that let agents reach internal systems without touching the public internet, Fast mode for Claude Opus 4.7 in research preview, and 20+ new legal MCP connectors aimed at law firms. Coverage on Fortune.
OpenAI’s reasoning model cracks an 80-year-old math problem
OpenAI said one of its general-purpose reasoning models autonomously solved a famous geometry problem that had stumped mathematicians for eight decades. The company framed the result as early evidence that frontier models can contribute original mathematical discovery, with potential spillovers into science, engineering and medicine. More via Axios.
What to watch: whether the postponed AI executive order returns with stricter language, and how aggressively rivals respond to Google’s price compression on frontier-tier models.