Nobody’s exclusive anymore

The exclusives are over. TechCrunch reports OpenAI and Microsoft jointly announced on Monday April 27 that Microsoft has lost exclusive access to OpenAI products and IP, retaining only a nonexclusive license through 2032 and a roughly 27% ownership stake in the for-profit entity. Microsoft moves from full revenue share to a capped revenue share through 2030, even as it reported $7.5 billion in quarterly revenue from its OpenAI position last quarter and committed $250 billion in additional cloud purchases in October. The immediate beneficiary is Amazon: OpenAI’s previously contested $50 billion AWS investment is now legally cleared, Andy Jassy says OpenAI models reach Bedrock “in the coming weeks,” and a joint AWS-OpenAI event is scheduled in San Francisco for today. The single most consequential alliance in commercial AI just got rewritten from gatekeeper to one cloud among several.

While OpenAI is dismantling exclusivity at the cloud layer, Google is signing a fresh one at the customer layer. The Decoder reports Google signed a classified AI contract with the Pentagon covering “any lawful government purpose,” over the objection of more than 600 employees—many from DeepMind—who urged Sundar Pichai to refuse the deal. The contract states the system “is not intended for domestic mass surveillance or autonomous weapons without appropriate human oversight,” language legal experts quoted in the piece say carries no enforceable weight, and Google has agreed to help the government adjust its safety filters on request. Google joins OpenAI’s February deal and an earlier xAI agreement as Pentagon AI suppliers; OpenAI retained full control of its safety stack in its version, and Anthropic was excluded after demanding contractual guarantees against mass surveillance and autonomous weapons. Three of the four U.S. frontier labs are now inside the classified perimeter; the fourth opted out.

And while incumbents reshuffle suppliers and customers, capital is quietly funding a route around the LLM paradigm itself. TechCrunch reports ex-DeepMind reinforcement learning lead David Silver raised $1.1 billion in seed funding for Ineffable Intelligence at a $5.1 billion post-money valuation—Europe’s largest seed ever, co-led by Sequoia and Lightspeed with Nvidia, Google, Index, DST Global, the UK Sovereign AI Fund, and the British Business Bank also in. There is no product, no revenue, and no public roadmap; Silver, the researcher behind AlphaGo, AlphaZero, AlphaStar, and AlphaProof, wants to build a “superlearner” that skips human-data pre-training entirely and learns from experience in simulation. The round follows Yann LeCun’s AMI Labs, which closed $1.03 billion in March, and Tim Rocktaschel’s Recursive Superintelligence, which pulled in roughly $500 million to $1 billion this month at a $4 billion valuation—all on similar non-LLM theses. The realignment isn’t only horizontal across clouds or customer-side at the Pentagon—it’s vertical across paradigms, and the seed checks now look like Series C rounds did two years ago.