Alphabet Overhauls AI Leadership as Hassabis Steps Down from DeepMind CEO Role

What happened

Alphabet announced a significant reorganisation of its AI division on Wednesday. Demis Hassabis, who co-founded DeepMind in 2010 and led it through its 2014 acquisition by Google, stepped down as Google DeepMind CEO to become Alphabet's chief scientist and DeepMind chairman. Koray Kavukcuoglu, previously serving as CTO, took over day-to-day responsibilities with the title of senior vice president rather than CEO, whilst retaining his parallel role as Alphabet's chief AI architect.

Simultaneously, four prominent Google AI leaders departed the company. Jeff Dean, Google DeepMind's chief scientist and one of Google's earliest employees, left alongside Sanjay Ghemawat, Oriol Vinyals, and Quoc Le to launch Discovery Loop, a public benefit corporation focused on AI-driven breakthroughs in machine learning, science, and engineering. Google committed to serving as a founding investor in the venture.

The timing proved sensitive as the flagship version of Google's latest Gemini model remains unreleased despite a planned June launch. This has sparked investor concerns about Google falling behind rivals Anthropic and OpenAI, which recently recruited star Google AI staff including Noam Shazeer (who joined OpenAI in June) and Nobel Prize-winner John Jumper (who moved to Anthropic).

Why it matters

The restructuring represents a pivotal shift in how Alphabet manages its AI strategy during an intensely competitive period. Hassabis cited the closeness of artificial general intelligence (AGI) as the impetus for his role change, allowing him to focus on long-term research and societal impacts whilst maintaining his CEO role at Isomorphic Labs, DeepMind's drug discovery spinout.

The departure of Dean and his collaborators marks significant talent loss. Dean and Hassabis were considered Google's two leading AI figures for years. Dean co-founded Google Brain in 2011, which merged with DeepMind in 2023 to form Google DeepMind after OpenAI's ChatGPT caught Google off guard. His collaboration with Ghemawat produced landmark infrastructure projects that still power Google today.

Senior leaders in Google Cloud, Alphabet's main AI revenue engine, welcomed Kavukcuoglu's appointment as positive for commercialisation efforts. Kavukcuoglu had been more directly involved than Hassabis in collaborations between DeepMind and Cloud, and his appointment is expected to steer Gemini towards a more product-oriented direction. Some people who worked with Hassabis said he put scientific pursuits ahead of making money for Google, though allies rejected this characterisation.

Bigger picture

Google faces mounting competitive pressure in AI. Whilst Gemini 3 released last November closed the gap with leading models from Anthropic and OpenAI, subsequent advances from those rivals and momentum for cheaper Chinese open-source models have put Google on the defensive. At Google's developer conference in May, executives touted Gemini's cost advantage rather than state-of-the-art capabilities.

The talent exodus compounds these challenges. Shazeer's departure to OpenAI came less than two years after Google paid billions to hire him and his team. Jumper, who won a Nobel Prize alongside Hassabis in 2024, joined Anthropic the same week. Dean, Vinyals, Le and former OpenAI chief scientist Ilya Sutskever coauthored a foundational 2014 research paper establishing that scaling up data to pre-train AI systems would send them to new heights.

The industry has matured rapidly from academic research to a central strategic priority commanding enormous infrastructure investments. Training frontier AI models requires clusters of specialised hardware consuming vast amounts of electricity. Alphabet has committed billions to expanding data centre capacity and developing more efficient computing architectures. The leadership changes attempt to balance ambitious scientific goals with commercial pressures and operational execution.

What to watch

Monitor progress on the delayed Gemini 4 model, which Hassabis touted as showing great progress. The flagship version's release timing will signal whether the restructuring improves execution.

Watch Discovery Loop's development as it focuses on AI-driven solutions for complex challenges with Google as a founding investor and cloud computing partner. The startup's success could influence perceptions of Google's ability to retain top talent.

Track whether Kavukcuoglu's product-oriented approach accelerates Gemini's integration into Google services including Search, YouTube, and hardware. Senior Cloud leaders expect improved collaboration.

Observe competitive dynamics with Anthropic and OpenAI as they continue advancing their models. Google's ability to close capability gaps whilst leveraging cost advantages will prove crucial.

Hassabis indicated he would spend more time with Isomorphic Labs pursuing drug discovery applications, potentially signalling increased focus on demonstrating AI's value in healthcare.