US vs China AI labs
Every tracked lab grouped by its HQ country (parsed from labs.json), with combined and average figures labelled as computed by us.
Data as of the week of 26 Oct 2026 (updated 26 Oct 2026, 08:09, Bucharest time). Every figure links to its source; see the methodology.
Where each lab is headquartered
| Lab | HQ (from labs.json) | Country |
|---|---|---|
| Anthropic | San Francisco, US | United States |
| Google DeepMind | London / Mountain View | United States |
| OpenAI | San Francisco, US | United States |
| Meta Superintelligence Labs | Menlo Park, US | United States |
| xAI (SpaceXAI) | Palo Alto / Memphis, US | United States |
| Alibaba (Qwen) | Hangzhou, CN | China |
| Moonshot AI (Kimi) | Beijing, CN | China |
| Z.ai (Zhipu) | Beijing, CN | China |
| DeepSeek | Hangzhou, CN | China |
| Mistral AI | Paris, FR | France |
Country is whatever follows the last comma in the HQ field (an HQ naming "Mountain View" counts as US; see below).
Google DeepMind lists its HQ as London / Mountain View; we count it as US because its parent, Alphabet, is based in Mountain View, California.
By country
| Metric | United States (5) | China (4) | France (1) |
|---|---|---|---|
| Labs | Anthropic, Google DeepMind, OpenAI, Meta Superintelligence Labs, xAI (SpaceXAI) | Alibaba (Qwen), Moonshot AI (Kimi), Z.ai (Zhipu), DeepSeek | Mistral AI |
| Best composite score | Anthropic 97.2/100 | Alibaba (Qwen) 53.6/100 | Mistral AI 27.7/100 |
| Average composite score (computed by us) | 75.7/100 (5/5 scored) | 49.9/100 (4/4 scored) | 27.7/100 (1/1 scored) |
| Best Arena rank | Google DeepMind #1 | Moonshot AI (Kimi) #18 | Mistral AI #111 |
| Best AA Intelligence Index | Anthropic 57.6 | Moonshot AI (Kimi) 55.1 | Mistral AI 38.4 |
| Combined revenue run-rate (computed by us) | $147B (3/5 reported) | $9.3B EST (3/4 reported, includes an estimate) | $1B (1/1 reported) |
| Combined valuation (computed by us) | $2.07T (3/5 reported) | $102B (2/4 reported) | $24B (1/1 reported) |
| Combined Polymarket 2026 odds (computed by us) | 94.1% (5/5 in market, sum of each lab's Yes price in the same market (outcomes are mutually exclusive, so this is the market-implied chance that one of these labs wins; prices are not normalised)) | 1.5% (4/4 in market, sum of each lab's Yes price in the same market (outcomes are mutually exclusive, so this is the market-implied chance that one of these labs wins; prices are not normalised)) | 0.2% (1/1 in market, sum of each lab's Yes price in the same market (outcomes are mutually exclusive, so this is the market-implied chance that one of these labs wins; prices are not normalised)) |
Sums and averages are plain arithmetic across the labs in each group that have a figure; they are not adjusted for accounting differences or market structure. See each lab's page for the basis and source of every input.
FAQ
Do US or Chinese AI labs score higher on average?
US-based labs average higher: US 75.7/100 vs China 49.9/100 (computed by us, over the labs with a score in each group).
Which single lab scores highest, US or China-based?
Anthropic, at 97.2/100.
Which group reports more combined revenue?
US-based labs, $147B (computed by us; only 3 of 5 US labs and 3 of 4 China labs report a figure, includes an estimate).
Which group does Polymarket favour for the best AI model at the end of 2026?
US-based labs, summing to 94.1% (computed by us; sum of each lab's Yes price in the same market (outcomes are mutually exclusive, so this is the market-implied chance that one of these labs wins; prices are not normalised)).
Which labs have no parseable country?
Every tracked lab has a parseable country.