Google DeepMind: History, Gemini Models & AI Research

Two labs under one roof

Google DeepMind was formed in April 2023 by merging DeepMind, the London research lab Google acquired in 2014, with Google Brain, the AI research group that had grown up inside Google separately. The merger put Google's frontier model work under one organization for the first time, and its own account of the move describes it as bringing together two of the world's leading AI labs into a single, focused team, guided by the scientific method and reporting up through Google's product organization rather than running as an independent research shop. [2]

Leadership changed in August 2026. Demis Hassabis, DeepMind's co-founder and Google DeepMind's chief executive since the merger, moved to become chair of Google DeepMind and chief scientist of Alphabet, a shift away from day-to-day operations and toward Alphabet-wide AGI strategy while he continues to lead Isomorphic Labs, the drug-discovery spinout. Koray Kavukcuoglu was promoted to senior vice president of Google DeepMind, overseeing Gemini model development, frontier research, and the Gemini app and developer teams, while keeping his title of chief AI architect for Google. Jeff Dean left the same week after 27 years at Google to start an independent public benefit corporation with Sanjay Ghemawat, with Google as a founding investor and cloud partner. [3]

The part of Google that actually builds the models

Google DeepMind, not Google as a corporate entity, is the organization that develops the Gemini model family and the open-weight Gemma line, and it carries the research credibility Google trades on in the model race: Hassabis and John Jumper shared half the 2024 Nobel Prize in Chemistry for AlphaFold's protein structure predictions, work DeepMind published years before the current wave of chatbot products existed. That research pedigree is Google DeepMind's specific relevance to AI history, distinct from Google's separate advantage in distribution through Search and Android. [4]

Atlas interpretation: The distinction matters when reading the timeline: an event credited to "Google" is usually about a product decision, a lawsuit, or a business deal, while one credited to "Google DeepMind" is usually about a model or a research result. The August 2026 leadership change formalized that split organizationally, putting Alphabet's AGI strategy under Hassabis as chief scientist while day-to-day Gemini development sits with Kavukcuoglu inside Google DeepMind. [3]

Google DeepMind on the timeline

None of Google DeepMind's events on the timeline is marked as a turning point on its own, but its major events trace the model line that eventually let Google catch up to OpenAI and Anthropic in the public argument over which lab is ahead. Gemini 3 shipped into Search, the Gemini app, developer tools and Vertex AI on the same day in November 2025, a launch breadth Google had not managed with the original Gemini two years earlier. In July 2025, Gemini Deep Think reached gold-medal standard at the International Mathematical Olympiad, graded by the olympiad's own coordinators, days after OpenAI claimed the same standard with a different model and its own former medalists doing the marking. Both labs got there independently, which is itself informative about how close the frontier had become. [5][6]

Google DeepMind's research also shows up in narrower, stranger events: AlphaEvolve, a Gemini-driven search over programs, found a way to multiply 4x4 complex matrices in one fewer scalar multiplication than the 1969 record, and the same system reportedly clawed back 0.7 percent of Google's worldwide compute by rewriting a data center scheduler. It is a small, verifiable result rather than a product launch, and it is the kind of event that justifies keeping the smaller items on the timeline at all. [7]

The August 2026 leadership reorganization is on the timeline as a major event, credited to both Google DeepMind and Google, because it changed who runs frontier model development inside Alphabet without changing who owns the company. It is the event to check first for any claim about who currently leads Google's AI research after that date. [3]