Our Funding

Epoch AI is a nonprofit, tax-exempt 501(c)(3) organization. Our research advances evidence-based discussions on the drivers, progress, and consequences of AI development and deployment, promoting informed policy and decision-making.

Funders

As an independent nonprofit, we rely on the ongoing support of donors to continue our research.

Open Philanthropy | General Support

April 2023$6,922,565 USD

Read more

February 2023$188,558 USD

Read more

June 2022$1,960,000

Read more

Jaan Tallinn | General Support

January 2025$600,000 USD DAF donation

Likith Govindaiah | General Support

January 2025$400,000 USD DAF donation

Leopold Aschenbrenner | Pilot for New Benchmark

January 2025$200,000 USD awarded via Manifund

Carl Shulman | General Support

March 2023$100,000 USD DAF donation

We value every contribution as a vouch of confidence in our work and vision, and we pledge to use it to further our mission of building a better shared understanding of AI. If you would like further information about specific funding needs, please don’t hesitate to contact us.

Donate

For more information, email us at info@epoch.ai.

Partnerships

We partner with leading organizations to produce impactful research and tools that advance the understanding and responsible development of AI.

OpenAI | Developing the FrontierMath benchmark

2024

Sentinel Bio | Tracking developments in biological AI models

2024/25

Google Deepmind | Joint report

2024/25

EPRI | Joint report on current trends in the energy demand for AI training

2024/25

AI Index | Joint report on the cost of AI development and investigation on all AI applications

2024/25

Department for Science, Innovation and Technology | Consultation

April 2023

Considering partnering with Epoch AI?

Contact us

Conflict of Interest

Epoch AI invests part of its funds in semiconductor and AI stocks. This increases our capacity to advance our mission in scenarios where our work matters most. Epoch AI is dedicated to conducting research that is impartial, objective, and meets a high standard of methodological integrity.