Alphabet possesses the most desirable AI infrastructure stack in the industry. The successful third-party agreements with Anthropic and Meta have made internal access a competitive asset.
Google has discreetly developed the most coveted position in AI infrastructure over the past decade with a robust cloud business, its own custom chips, and supply deals that position its TPUs as the main alternative to Nvidia for major external clients.
This strategy’s success has created an unforeseen internal issue for the company.
Bloomberg’s Julia Love reported on Monday that Google’s AI researchers, including teams within Google DeepMind, are competing for access to the computing resources their employer is also providing to Anthropic and Meta.
The 💜 of EU tech
The latest rumblings from the EU tech scene, a story from our wise ol’ founder Boris, and some questionable AI art. It’s free, every week, in your inbox. Sign up now!
The structural reason is simple. Google has committed to investing up to $40bn in Anthropic, encompassing five gigawatts of TPU capacity over five years and access to up to one million seventh-generation Ironwood chips.
A separate supply line mediated by Broadcom will provide an additional 3.5GW of TPU capacity for Anthropic from 2027, augmenting the 1GW the company is already receiving in 2026. Anthropic itself has publicly characterized the Google TPU stack as vital to its training and serving roadmap.
Meta, the other commercial-scale TPU client cited by Bloomberg, signed its own agreement earlier this year. The capacity those commitments occupy is unavailable to Google’s internal model teams without a wait.
DeepMind’s CEO Demis Hassabis noted earlier this year that the limitation works both ways. Part of the bottleneck is hardware, involving ‘a few suppliers of a few key components’, with high-bandwidth memory from Samsung, Micron, and SK Hynix being the most commonly mentioned choke point.
Another part relates to research throughput, as researchers, according to Hassabis, ‘require a significant number of chips to experiment with new ideas at scale’. The hardware constraint is partially outside Google’s control, whereas the internal allocation constraint is not.
The underlying arithmetic is considerable. Alphabet is operating within a guided capex range of $175bn-$185bn for 2026, within a collective Big Tech AI infrastructure expenditure that has surpassed $650bn this year. Google has, based on its statements, been activating significantly more than a gigawatt of new AI compute capacity in 2026.
The <