Google DeepMind
Alphabet's AI research lab developing advanced models like Gemini and AlphaFold.
Updated April 2026
Overview
- Website
- deepmind.google
- Founded
- 2010
- Headquarters
- London, UK
- Segment
- Frontier Foundation Model Labs
- Posture
- Closed-Source Frontier
Product overview
Google DeepMind builds frontier AI models including multimodal LLMs (Gemini family), protein folding (AlphaFold), video generation (Veo), music (Lyria), robotics (Gemini Robotics), and world models (Genie), powering Google products like Search, Assistant, YouTube, and Cloud services used by billions worldwide. Distinct from other labs by deep Alphabet integration, unparalleled compute, interdisciplinary approach spanning reinforcement learning and neuroscience, and responsibility focus toward AGI benefiting humanity.
Revenue model
As Alphabet subsidiary, contributes indirectly via AI enhancements to ad revenue (Search/YouTube) and consumer products (Gemini app); directly via Google Cloud Vertex AI usage-based API pricing for Gemini models (e.g. Gemini 2.5 Pro input $1.25-$2.50/1M tokens, output $10-$15/1M tokens; provisioned throughput $2,000/GSU/month yearly).
Moat
Google DeepMind's key competitive moat is its extensive portfolio of proprietary AI technologies and patents, including breakthroughs like AlphaFold for protein folding, AlphaCode for competitive programming, and AlphaEvolve for algorithmic discoveries, which create high barriers to entry through cutting-edge capabilities in machine learning that competitors struggle to replicate. This is reinforced by strategic withholding of research publications to protect innovations and Google's massive scale advantages in compute resources, data access, and integration across products like data centers and Android, enabling rapid deployment and efficiency gains unmatched by rivals.