The Allocation Turn: When Compute Is Rationed by Capacity, Not Price
Google told Meta it could not buy all the Gemini capacity it wanted, and Meta started rationing tokens. AI compute now clears on allocation, not price.
5 articles tagged with Hyperscalers
Google told Meta it could not buy all the Gemini capacity it wanted, and Meta started rationing tokens. AI compute now clears on allocation, not price.
Microsoft Agent 365 hit GA May 1 2026 at $15/user/month, with Entra Agent IDs and third-party agents from Adobe, SAP, Zendesk, and Manus folded into Defender and Intune oversight. Five days after Azure-OpenAI exclusivity ended, Microsoft is racing to be the AI control plane — with the EU AI Act August 2 deadline 91 days out.
On April 27 2026 Microsoft and OpenAI ended Azure exclusivity. OpenAI keeps a capped 20 percent payment to Microsoft through 2030 but is freed to ship production workloads on AWS, GCP, and Oracle. The same week, Anthropic locked in $40B from Google and $5B from Amazon against a $100B compute commitment. The cloud-and-foundation-model relationship that powered the post-2023 build-out is being unwound — and most enterprise AI architectures still assume the world that ended last Monday.
Half of Alphabet's and Amazon's headline Q1 2026 profits did not come from selling cloud, ads, or commerce. They came from marking up the carrying value of an Anthropic stake whose next funding round is being structured by the same companies booking the gains. The accounting is legal, the cash is not real, and the circularity is now the dominant feature of Big Tech earnings. This is what mark-to-market AI economics looks like at $900 billion.
On April 29, 2026, the four largest hyperscalers reported Q1 earnings into a market that had spent three months absorbing the news that helium production fell off a cliff and DRAM costs more than doubled. The announced seven-hundred-billion-dollar AI capex story collided with the realized cost of building it. The gap between those two numbers is the new strategic axis of the AI infrastructure economy, and it is widening faster than the earnings calls have admitted out loud.