Our latest work

Filter

Topic
Type
Agentic inference capacity grew 4.6× in a year, versus 2.6× for compute capacity
Data Insight
Nov. 18, 2026
Agentic inference capacity grew 4.6× in a year, versus 2.6× for compute capacity

Our Inference Performance Index and chip sales data show agentic inference capacity grew nearly twice as much as 8-bit compute.

By Jason Li

Anthropic’s compute spending is increasing by 4x per year, outpacing OpenAI’s 3x
Data Insight
Nov. 18, 2026
Anthropic’s compute spending is increasing by 4x per year, outpacing OpenAI’s 3x

Anthropic spent nearly $10 billion on compute and infrastructure in the third quarter of 2026, or close to $40 billion on an annualized basis, up from just $433 million in 2023.

By Josh You

How good are Fable 5.2 and GPT-6.2 Astra at real work?
Report
Nov. 17, 2026
How good are Fable 5.2 and GPT-6.2 Astra at real work?

We tested Fable 5.2, GPT-6.2 Astra, and four other models on real Epoch work tasks. None met our employee standard, including the two that top our ECI.

By Kelly Hong

Incident report: AI agents attempted exploits on our website during a third-party alignment evaluation
Report
Nov. 13, 2026
Incident report: AI agents attempted exploits on our website during a third-party alignment evaluation

Three October 2026 events involving Trajectory Labs, PBC, in which agents probed our site and impersonated staff. No private data was accessed.

By The Epoch AI Team

Inference Performance Index: changing the way we measure inference capacity
Report
Nov. 9, 2026
Inference Performance Index: changing the way we measure inference capacity

Comparing AI hardware for inference beyond FLOP/s with memory and inter-chip communication

By Jason Li

What Huawei’s Horinger data center reveals about China’s compute gap
Report
Nov. 6, 2026
What Huawei’s Horinger data center reveals about China’s compute gap

Huawei’s AI data center in Horinger, Inner Mongolia, uses Ascend chips and has an estimated AI computing capacity that is just 7% that of the world’s largest AI data center, SpaceX’s Colossus 2 in Tennessee.

By Ben Cottier

Broadcom TPU compute shipments doubled from Q2 to Q3 2026, reaching 2 million H100-equivalents
Data Insight
Nov. 4, 2026
Broadcom TPU compute shipments doubled from Q2 to Q3 2026, reaching 2 million H100-equivalents

Broadcom’s shipments of TPU by unit volume increased by about 65% quarter over quarter to roughly 800,000 chips, while compute output doubled, driven by a shift to more powerful chips.

By Venkat Somala and Josh You

Updating the Epoch Capabilities Index methodology to use a Bayesian model
Report
Nov. 4, 2026
Updating the Epoch Capabilities Index methodology to use a Bayesian model

A Bayesian model gives the ECI better-justified uncertainty estimates and lets us compare models directly.

By Emilia Álvarez and Alexander Barry

LLM use is acknowledged almost as often as specialized AI tools in biology preprints
Data Insight
Oct. 30, 2026
LLM use is acknowledged almost as often as specialized AI tools in biology preprints

17% of biology preprints acknowledge LLM use, nearly matching specialized tools like AlphaFold. Epoch AI analyzes trends in AI acknowledgment in research.

By Tara Abrishami

AI can turn security fixes into working exploits before users receive an update
Report
Oct. 29, 2026
AI can turn security fixes into working exploits before users receive an update

GLM 5.3 turned published-but-unshipped Firefox security fixes into working one-click exploits for six releases, at roughly $1,000 each.

By Aiden Ament, Yafah Edelman, Alexander Barry, Timothée Chauvin, Lee Bousfield, and Jaeho Lee