Fetching from the wire…
Public story · 2026-07-23 · high
OpenAI signed its own deal with the DOE and the national labs the same week, making federal science funding a two-lab compute contest.
Why now: Google's and OpenAI's commitments landed in the same week, turning what looked like a single donation into a two-lab contest for federal science compute.
Google committed $40 million in tokens and cloud credits, not cash, to the federal government's Genesis Mission, per a DeepMind announcement. The stakes: federal researchers end up running their work on whichever company's infrastructure backs the commitment, not on hardware picked for the science.
That structure keeps the money inside Google's own systems. It funds compute for federal researchers running on Google's stack, rather than a check the government could spend with any vendor.
OpenAI signed its own commitment with the Department of Energy and the national labs the same week, per its own announcement. Two labs, two structures, one federal science program.
That timing turns Genesis Mission into a two-lab compute contest. Federal research funding used to mean grants a lab could spend on whatever hardware fit the work. Now it's tokens and credits tied to whichever vendor signed first.
Neither announcement says what research gets prioritized in exchange. DeepMind's post covers Google's commitment, OpenAI's covers its own, and neither says how the two split inside the same program.
I'd watch whether DOE ever publishes a usage split between the two vendors. That number would show who's actually winning the contract, not just who announced first.
Each link below shares sources, entities, or timing with this story.
The U.S. Department of Energy formalized agreements with OpenAI, NVIDIA, Google DeepMind, and 21 others under a $293M initiative spanning advanced manufacturing, biotech, nuclear energy, and quantum science. The most concrete government-AI lab coordination mechanism to date. DOE
The pilot puts evaluation inside Confidential Space on Google Cloud's Confidential Computing so Gemini Flash Lite's weights stay private from evaluators while evaluator prompts stay private from Google (DeepMind). Partners are the Singapore AI Safety Institute, OpenMined, AVER...
A $150M investment to help systems integrators and consultancies accelerate enterprise AI deployment (OpenAI). It builds the channel and go-to-market layer OpenAI needs to compete against Microsoft, Google, and Anthropic for enterprise agents, where deployment support, not mod...
Anthropic filed a federal lawsuit to overturn its Pentagon "supply chain risk" designation, which bars Claude from all Department of Defense work. The designation came after Anthropic demanded restrictions on mass surveillance and autonomous weapons applications. In a rare sho...
The single biggest AI governance story ever unfolded today. After Anthropic CEO Dario Amodei drew two absolute red lines — no mass domestic surveillance of Americans and no fully autonomous weapons — the situation escalated rapidly: - 5:01 PM ET: Pentagon deadline passed witho...
Zoph co-founded Thinking Machines Lab with Mira Murati and served as CTO, left in January 2026 with Luke Metz to return to OpenAI, was put on enterprise AI sales, and was later revealed to have been fired after five months (TechCrunch). Google to OpenAI to Thinking Machines to...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.