Google DeepMind: the AI lab behind Gemini and AlphaFold

Google DeepMind is the artificial intelligence research and development unit of Google, part of Alphabet Inc. It builds the Gemini family of AI models and created AlphaGo and AlphaFold, and it describes its mission as “to build AI responsibly to benefit humanity”.

The unit sits at the centre of Google’s AI plans. Its models power the Gemini app, which Alphabet says passed 750 million monthly active users at the end of 2025, and two of its scientists shared the 2024 Nobel Prize in Chemistry.

Overview and Founding

DeepMind started in London in 2010. Its chief executive, Demis Hassabis, co-founded it, and press reports in 2014 named him, Shane Legg and Mustafa Suleyman as the people running the young company. The company says it set out with an interdisciplinary approach to general AI.

An early result was DQN, a program that learned to play 49 Atari games using only the raw pixels on the screen as input. Google DeepMind says this work made it a pioneer of deep reinforcement learning, in which a system learns by trial and reward.

Google agreed to buy DeepMind in January 2014. The lab remained in London, where the Nobel committee still lists Google DeepMind as the home of its two laureates.

Google had a second AI group of its own. The Brain team began in 2011 at X, Alphabet’s research lab, and in 2017 its researchers invented the Transformer, a model architecture for processing language and other sequences.

In April 2023 Google chief executive Sundar Pichai combined the Brain team and DeepMind into one unit, Google DeepMind, with Hassabis as chief executive. Pichai wrote that the group would bring together two leading research teams in the field.

Key Milestones

• 2010: DeepMind is founded in London.
• 2014: Google agrees in January to acquire DeepMind.
• 2015: AlphaGo beats the three-time European Go champion 5-0 in October.
• 2016: AlphaGo beats Lee Sedol 4-1 in Seoul in March.
• 2019: AlphaStar beats a top professional player at StarCraft II.
• 2020: AlphaFold2 is presented, predicting the 3D structure of proteins.
• 2023: Google DeepMind is formed in April, and Gemini 1.0 is launched in December.
• 2024: AlphaFold 3 is released in May with Isomorphic Labs, and Hassabis and John Jumper share the Nobel Prize in Chemistry in October.

What It Does

Google DeepMind does basic research and also builds the models used across Google’s products. Its best-known current work is Gemini, a family of multimodal models that handle text, code, audio, images and video. The first version, launched in December 2023, came in three sizes: Ultra, Pro and Nano.

The latest generation, Gemini 4 Argon, is described on the unit’s site as built for science, mathematics, robotics, cybersecurity and coding work. Its global release to developers is covered in our report Google opens Gemini 4 Argon to developers worldwide at $2 per million input tokens.

Science is the second pillar. AlphaFold2 made it possible to predict protein structures at scale, and AlphaFold 3 extended this to DNA, RNA and small drug-like molecules. A free AlphaFold Server lets researchers use the model for non-commercial work.

The unit also works with Isomorphic Labs, a separate Alphabet company that Hassabis leads and that uses AI to design new medicines.

Google co-founder Sergey Brin has worked with the researchers building Gemini since his return to technical work, Fortune reported in 2025.

In Numbers

• AlphaFold Database: 200 million protein structures (2024).
• AlphaFold2 users: more than two million people in 190 countries (2024).
• Gemini app: more than 750 million monthly active users (fourth quarter of 2025).
• Gemini models: more than 10 billion tokens a minute through direct customer API use (fourth quarter of 2025).
• AlphaGo against Lee Sedol: watched by more than 200 million people (2016).

Leadership

Demis Hassabis is Co-founder and Chief Executive Officer of Google DeepMind. He also serves as Founder and Chief Executive Officer of Isomorphic Labs.

When the unit was formed in 2023, Jeff Dean was named Chief Scientist for Google Research and Google DeepMind, reporting to Pichai.