Google reorganizes DeepMind leadership as engineering talent exits for independent labs
The departure of Jeff Dean and other senior fellows to form Discovery Loop marks a structural shift for Google, prioritizing cloud infrastructure revenue over first-party model supremacy.
Julian Reeve
Aug 12, 2026 · 1 min read
Four of Google’s most senior technical architects announced their departure on August 5 to establish an independent research firm, Discovery Loop. The group includes Jeff Dean, the former Google Chief Scientist who initiated the TPU program, alongside Google Fellows Sanjay Ghemawat and Quoc Le, and Gemini co-lead Oriol Vinyals. Their exit coincides with a broader leadership overhaul at DeepMind, where co-founder Demis Hassabis has stepped back from daily operations, leaving former CTO Koray Kavukcuoglu to lead the division.
The restructuring reflects a decisive pivot in how Google allocates its most valuable industrial resource: compute. For several quarters, internal tension has mounted between the Gemini research teams and Google Cloud Platform (GCP) over access to high-end hardware. That conflict appears resolved in favor of the cloud division. Google has committed more than 20% of its total TPU shipments from the third quarter of 2026 through late 2027 to Anthropic, effectively prioritizing external rental revenue over the compute requirements of its own frontier models.
This shift in capital allocation follows a visible cooling in Gemini’s market performance. First-party API token growth decelerated from 60% in the first quarter of 2026 to 38% in the second. As first-party model performance has slipped relative to competitors, the Gemini Enterprise Agent Platform—formerly Vertex—has sustained its momentum primarily through the integration of third-party models.
The emergence of Discovery Loop follows a pattern established by other senior Google veterans who have raised external capital to procure Nvidia hardware through GCP. By transitioning from an internal lab to a cloud customer, these teams bypass internal compute rationing while contributing to GCP’s accelerating revenue growth. For Google, the strategy sacrifices the immediate pursuit of state-of-the-art model rankings to secure its position as the primary infrastructure provider for the rest of the industry.