One week after DeepMind’s leadership handover, Google is going further than a change of faces. Reuters reports that at an August 6 all-hands meeting, management told employees that several non-technical teams would leave DeepMind and report directly into Google’s corporate structure. The research teams stay. The shell around them — the support, operations and coordination layers that let DeepMind feel like a lab apart — is being pulled back into the mothership.
Employees learned about the scale of the shake-up only when the leadership change went public. Now they are left with questions Google will not answer: which teams are moving, who decides the fate of remaining projects, and how much funding survives for research that does not directly serve Gemini.
What Actually Happened: Three Movements in One Week
The restructure is really three separate power shifts arriving at the same time.
1. Teams move to corporate reporting. Reuters says the moved teams are non-research — support functions rather than scientists. Google has not confirmed the list or the reasoning. Read against the broader reorganization, the direction is clear: fewer layers between DeepMind and Google’s product units, so Gemini can reach Search, Cloud and every other surface faster.
2. Koray Kavukcuoglu consolidates control. The new DeepMind leader took over oversight of model development in 2025. Google confirmed last week that he now holds final say over major DeepMind decisions. Four people familiar with the matter told Reuters that some DeepMind leaders have watched their influence shrink in recent months. Notably, the person who has been the main liaison with Google Cloud — the company’s primary AI revenue engine — has been Koray all along, not Demis Hassabis.
3. Sergey Brin returns to the AI power center. The 52-year-old co-founder holds no formal management title, but he has re-entered Gemini model training and resource allocation. After Anthropic demoed Claude Mythos in April, Brin joined an internal meeting with hundreds of employees and demanded speed. In recent months he has pushed core AI staff to catch the frontier and steered resources toward ‘recursive self-improvement’.
From Independent Lab to In-House Model Division
Since Google acquired it in 2014, DeepMind kept a remarkable degree of autonomy: its own culture, its own teams, its own research agenda. The structural signals now point the other way. DeepMind may keep its brand and its research staff, but its organizational form is becoming what the market would call a model R&D department inside Google — not a semi-independent lab.
This is the resolution of a tension that has existed since day one. An independent lab optimizes for long-horizon science. A product organization optimizes for shipping. For twelve years Google let the lab side win most arguments. The August restructure, the leadership handover and Brin’s return are all the same answer: product wins now.
What It Means: Faster Gemini, Pricier Ambiguity
For Gemini users and builders, the restructure is good news on cadence. The 3-week iteration cycle we saw with Gemini 3.7 Flash’s aggressive pricing is a preview of what a product-aligned DeepMind can ship. Removing coordination layers tends to compress model-to-product time.
For researchers, the calculus changes. Non-Gemini research will have to justify its existence against product deadlines. The long-horizon science chapter of DeepMind’s story is closing, and the industry is watching whether other frontier labs follow the same path. The pattern is not unique: the market is no longer pricing in lab autonomy anywhere.
For the talent market, the signal is blunt. Researchers who joined DeepMind for independence now face a choice between the brand and the mission. Meanwhile a single person — Koray — becomes the bottleneck for direction, which concentrates both speed and risk.
What to Watch Next
Three signals will tell you whether this restructure is working: whether Gemini’s shipping cadence accelerates further, whether Google Cloud revenue keeps climbing as the AI revenue engine, and whether non-Gemini research gets quietly defunded or publicly defended. None of this happens in a vacuum — the same squeeze is hitting the AI agent startup funding winter, where capital is also demanding product over promise.
DeepMind the brand will survive. DeepMind the independent lab will not. That transition, more than any single model release, is the real story of Google’s AI strategy in 2026.
