Companies like Google and Microsoft face a dilemma: use their compute to develop their own AI models or rent it out for high returns. The profitable infrastructure-as-a-service model often wins, starving internal research teams and creating a conflict that slows their model development, an issue pure-play labs like OpenAI don't face.

In a significant strategic misstep, Google sold a large volume of its custom TPU accelerators to rival Anthropic. Immediately after, demand for Google's own Gemini model surged, leaving Google compute-constrained and trying to secure more capacity from a sold-out TSMC.

High-profile departures from DeepMind and Google selling 20%+ of its TPU capacity to competitor Anthropic show a strategic shift. Google is prioritizing its profitable cloud (GCP) business over winning the frontier model race, effectively becoming a "picks and shovels" provider for the AI industry rather than a leading research lab.

Cloud providers like Amazon and Google benefit regardless of which AI model wins. By structuring deals as large-scale compute commitments in exchange for equity (e.g., with Anthropic), they profit from cloud usage fees, drive adoption of their in-house silicon, and gain visibility into data center capex recovery, effectively hedging their bets across the entire AI ecosystem.

Google's cloud division (GCP), incentivized to sell compute, is allocating scarce TPU chips to external customer Anthropic. This directly constrains Google's own AI lab, Gemini, hindering its progress in the hyper-competitive AI race and revealing significant internal friction between business units with conflicting goals.

Google's apparent failure to keep pace with OpenAI and Anthropic might not be an accident. It could be a strategic choice to cede the current LLM battle and focus resources on what it believes is the next frontier: "world models." This is a high-risk gamble, betting that a future breakthrough will leapfrog today's technology.

While the public focuses on the race for the best AI model, Google has a profitable alternative: being the 'picks and shovels' provider. By selling its powerful cloud infrastructure and TPUs to rivals like Anthropic and OpenAI, Google can generate massive revenue from the AI boom regardless of whether its own models are state-of-the-art.

Google is allocating its massive $200B CapEx to data centers, a high-confidence return on investment, rather than the riskier frontier model development. This de-prioritization is pushing top AI researchers to leave and launch their own ventures, where capital for model building is abundant.

Google DeepMind's leadership changes and model slowdowns suggest a strategic pivot. Instead of competing on frontier models, Google may prioritize its cloud platform (GCP), aiming to become the compute provider for other AI firms, mirroring Microsoft's Azure strategy.