The business behind the AI race

Follow the capital behind the models.

Compare company-wide financial context for five public companies connected to model research in our Epoch AI source. Financial scale adds business context. It does not determine model quality or change a ranking.

Evidence boundary

Model authorship is established by the reviewed Epoch AI artifact. Corporate financial observations come from exact public-parent SEC 10-K filings. The two evidence classes are not blended into a score or causal claim.

Meta AI

Meta Platforms

Research and development expense

$57.372BFY ended 2025-12-31

Llama 4 Scout establishes the model-research link. The SEC metric remains company-wide.

View company profile

Google DeepMind

Alphabet

Research and development expense

$61.087BFY ended 2025-12-31

Gemini 3.1 Pro establishes the model-research link. The SEC metric remains company-wide.

View company profile

Microsoft Research

Microsoft

Research and development expense

$32.488BFY ended 2025-06-30

CODEFUSION (Python) establishes the model-research link. The SEC metric remains company-wide.

View company profile

Amazon AI

Amazon

Revenue from contracts with customers, excluding assessed tax

$716.9BFY ended 2025-12-31

Amazon Nova Pro establishes the model-research link. The SEC metric remains company-wide.

View company profile

NVIDIA Research

NVIDIA

Research and development expense

$18.497BFY ended 2026-01-25

Nemotron 3 Ultra establishes the model-research link. The SEC metric remains company-wide.

View company profile

What we leave out

Investment is not ownership.

OpenAI: Our sources do not establish a public-parent ownership relationship. Microsoft is treated as its own research organization, not as OpenAI's parent.

Anthropic: Our sources do not establish a public-parent ownership relationship. Amazon is treated as its own research organization, not as Anthropic's parent.

xAI, Mistral AI, and DeepSeek: Our sources do not contain exact SEC filing facts for these private organizations.