Research institute · Research orgs · San Francisco, California · Est. 2022
Epoch AI is an independent, data-first research nonprofit investigating the future of artificial intelligence—specifically the drivers of AI progress, the evolution of AI capabilities, and the economic and societal impacts of AI. Its work emphasizes building shared, evidence-based datasets and evaluation infrastructure (including compute, hardware, and infrastructure tracking) and publishing research openly so that policymakers, researchers, and industry can reason from consistent information rather than anecdote.
Epoch also runs benchmarking and an aggregation layer for model capability assessment, and it provides mission-aligned commissioned research and evaluations for companies, nonprofits, and government bodies.
How quickly are Anthropic and OpenAI growing their revenue?
The Nvidia-sized hole in US GDP statisticsGDP growth is understated by about 0.3 percentage points because of missing value from fabless chipmakers, primarily Nvidia.
9 Big questions benchmarks can help answerA good benchmark asks more than just “can AI do this specific task?”
Will financing bottleneck AI compute? An Anthropic case studyWhy Anthropic's buildout suggests that financing is unlikely to be the immediate blocker to frontier AI compute growth.
The Epoch Brief - July 31, 2026Expanding FrontierMath: Open Problems, how "parallelizability" determines a technological singularity, the realities of AI energy use, and signs of AI uplift
OpenAI accidentally hacked Hugging Face — should we have seen it coming?Expert assessments and cyber benchmarks led us to expect that frontier models were capable of executing this kind of cyberattack
The Epoch Brief - July 8, 2026A board game AI can't master, a cyber-disclosure spike after Claude Mythos, GPT-4's record run atop the ECI, and what's missing from AI futurism discourse
The missing half of AI futurism debatesWhy we should think a little harder about what it takes to build a Dyson Sphere
The Epoch Brief - June 26, 2026Our new long-horizon coding benchmark, hyperscaler cash flows, tracking AI R&D automation, and Chinese lab strategies
What we learned from 1,604 Chinese AI job postingsInferring Chinese AI labs’ strategies from their job descriptions
Toward an O*NET for AI R&DProposing a new way to track AI research automation
The Epoch Brief - June 12, 2026Increased cyber vulnerability reports, Mythos' cyber hype, Fable 5's lead on FrontierMath v2, record-setting data centers, and what wealth distribution could look like after AGI
Epoch reports that since mid-2024 the record for the largest AI data center by IT power capacity has doubled every ten months, and it identifies xAI’s Colossus 2 as the current record-holder (at an estimated 950 MW).
Will Huawei catch up to Nvidia by 2030?Epoch estimates Huawei will produce less than 4% as much AI compute as Nvidia in 2026 and discusses constraints tied to export controls and supply-chain bottlenecks (including HBM), arguing it is unlikely Huawei catches up by 2030 under the stated assumptions.
The ECI frontier has advanced by 14 points per year since the introduction of reasoning modelsEpoch describes a linear ECI frontier advancement since September 2024 reasoning model introduction, comparing 14 ECI points/year for reasoning models to a lower pace for non-reasoning models.
Announcing FrontierMath ErdősEpoch announces FrontierMath Erdős: a benchmark of 68 significant Erdős problems, open as of August 2026, curated by Thomas Bloom and formalized in Lean, with the goal of proving/disproving problems within a fixed budget.
An update on AI’s most important numberEpoch’s Gradient Updates commentary focuses on how quickly Anthropic and OpenAI are growing their revenue.
The Nvidia-sized hole in US GDP statisticsEpoch argues US GDP growth is understated by about 0.3 percentage points due to missing value from fabless chipmakers, primarily Nvidia, and traces the accounting gap through Nvidia’s fabless chip supply chain.