Home / Indexes / Geography of AI
Thank you for reading this post, don't forget to subscribe!
Ask where the AI revolution is happening and the honest answer is: overwhelmingly, in one country. The United States drew roughly $109 billion in private AI investment in 2024 — about 12× China’s total and 24× the UK’s (Stanford HAI) — and controls around three-quarters of the world’s AI computing power (Epoch AI). This pillar maps the real geography of AI: who owns the money, the machines, the models, and the talent — and where the “only in America” story breaks down.
“America is in the very beginning of a revolution that we own, the AI revolution… Every single relevant company in the world is in this country. There is no other place to do technology really at scale besides America right now… Europe has basically decided to regulate its anemic and non-existent tech scene out of production… Who wants to build a company in China or Russia? It’s happening right here.”
It’s a provocative claim. So we checked it against the data. On capital, compute, and corporate scale, Karp is directionally right. On the absolutes — “every single relevant company,” “not happening” anywhere else — the numbers say it’s overstated. Here’s the sourced picture.
The money: America owns the capital
US private AI investment hit $109.1 billion in 2024 — nearly 12× China’s $9.3 billion and 24× the UK’s $4.5 billion (Stanford HAI, 2025). The gap is widening: in generative AI specifically, US private investment exceeded China plus the EU and UK combined by $25.4 billion in 2024. Scaling up to all corporate spending, global AI investment reached roughly $581 billion in 2025, of which the US accounted for about $285.9 billion against China’s $12.4 billion (Stanford HAI, 2026). Venture capital tells the same story: the US absorbs roughly 52% of all global VC versus Europe’s ~5%, and in AI startups specifically the US drew about $47 billion in 2024 to Europe’s ~$11 billion (Silicon Valley Bank, 2025).
The machines: America owns the compute
Compute is the hard constraint on frontier AI, and it is concentrated even more sharply than money. The US hosts about 74.5% of the world’s AI-supercomputer performance, with China a distant second near 15% and the entire EU around 5% (Epoch AI, 2025). This is the physical backbone behind Karp’s “no other place to do technology at scale” — training a frontier model requires GPU clusters that, today, mostly sit on American soil.
The companies: America owns the scale
By market capitalization, the corporate map is almost monochrome. Of the ten largest AI-focused companies in 2025, all but one (China’s Cambricon) are US-based, led by Nvidia (Statista, 2025). Europe’s single breakout, Mistral AI, raised the continent’s largest-ever AI round — a ~$2 billion Series C toward a ~$14 billion valuation (Crunchbase, 2025) — yet that remains an order of magnitude below leading US labs. This is the part of Karp’s claim the data supports most cleanly: on commercialized scale, there is currently no peer to the US cluster.
China: the rival Karp waves away
“Who wants to build a company in China?” is where the claim frays. China led the world in AI research output in 2023 with 23.2% of global AI publications, and filed roughly 69.7% of all global AI patents — versus 14.2% for the US (Stanford HAI / WIPO). The capability gap has collapsed: China’s best model now trails the top US model by just 2.7% on the LMArena leaderboard, down from ~31.6% in 2023 (Stanford HAI, 2026). And China’s DeepSeek reported training its R1 reasoning model for just $294,000 (peer-reviewed in Nature), after which R1 became the most-downloaded open-weight model on Hugging Face. China is not a non-participant; it is a genuine peer competitor, especially in research and open models. Full breakdown: AI in China 2026 →
Europe: research strength, scale-up failure
Europe’s problem isn’t talent or science — it produces about 15% of the world’s AI publications — it’s converting that into scaled companies. European institutions released just 3 notable AI models in 2024 (Stanford HAI). More than half of the world’s unicorns are US-based versus under 10% in the EU, and of 147 unicorns founded in Europe, roughly 40 have relocated their headquarters abroad, mostly to the US (Fortune / Draghi report context, 2026). The 2024 Draghi report on European competitiveness argued that regulatory density — the AI Act and GDPR among the culprits — deters investment; by 2025 Mario Draghi was calling to pause the AI Act, and the EU has delayed its high-risk rules toward 2027–2028. On Europe, in other words, Karp is closest to right — not because Europe lacks ability, but because it struggles to scale what it builds. Full breakdown: AI in Europe 2026 →
The talent: American institutions, global brains
About 59% of the world’s top-tier AI researchers work at US institutions (MacroPolo, 2024) — a concentration that mirrors the capital and compute. But the same data carries a warning inside Karp’s triumphalism: roughly 38% of that elite US-based talent was originally trained in China. America’s lead is real and it is also imported — dependent on continuing to attract the world’s researchers rather than home-growing all of them. The Tortoise Global AI Index 2024 ranks the US #1 and China #2 by a wide margin, then Singapore, the UK, and France.
On scale, mostly yes. Capital (~12× China), compute (~75% of the world’s), and market-cap leadership (10 of the top 10) all confirm a genuine, structural US advantage — the kind he compares to the early nuclear era. On the absolutes, no. “Every single relevant company” overlooks China’s DeepSeek and near-parity models, its ~70% of AI patents, and Europe’s real (if unscaled) research base. The accurate version: the AI revolution is being commercialized in America first and fastest — but it is being contested, especially by China.
Go deeper
- AI in China 2026 — patents, open models, and compute under export controls
- AI in Europe 2026 — the scale-up gap and the regulation debate
- Sovereign AI Index 2026 — national compute buildouts, country by country
- Global AI Regulation Tracker — the EU AI Act and the global rulebook
- AI Investment & Funding Statistics 2026 — the capital flows in detail
Frequently asked questions
Is the AI revolution really only happening in America?
Mostly, by scale. The US leads private AI investment (~$109B in 2024, ~12× China), holds ~75% of global AI supercomputing, and hosts nearly every top AI company by market cap. But China is a genuine peer in research and open models, so “only” is an overstatement.
Where does China stand in AI?
Second, and closing fast: ~23% of global AI publications and ~70% of AI patents in 2023, with its best model now trailing the top US model by just 2.7% on LMArena (down from ~32% in 2023).
Why does Europe lag in AI?
Not for lack of research (~15% of global AI publications) but for lack of scale-up: just 3 notable models in 2024, a heavy regulatory load the Draghi report blames for deterring investment, and unicorns relocating to the US.
Sources
- Stanford HAI, 2025 AI Index Report (Economy; R&D), Apr 2025 — https://hai.stanford.edu/ai-index/2025-ai-index-report/economy
- Stanford HAI, 2026 AI Index Report, Apr 2026 — https://hai.stanford.edu/ai-index/2026-ai-index-report
- Epoch AI, AI Supercomputers performance share by country, May 2025 — https://epoch.ai/data-insights/ai-supercomputers-performance-share-by-country
- Statista, Leading AI companies globally by market cap, 2025 — https://www.statista.com/statistics/1619086/top-ai-firms-globally-by-market-cap/
- MacroPolo (Paulson Institute), Global AI Talent Tracker 3.0, 2024 — https://archivemacropolo.org/interactive/digital-projects/the-global-ai-talent-tracker
- CNN Business / Nature, DeepSeek R1 training cost ($294K), Sep 2025 — https://www.cnn.com/2025/09/19/business/deepseek-ai-training-cost-china-intl
- Silicon Valley Bank, AI Industry Trends in Europe, 2025 — https://www.svb.com/business-growth/global-expansion/ai-industry-trends-in-europe/
- Crunchbase News, Mistral’s $2B Series C is Europe’s largest AI round, 2025 — https://news.crunchbase.com/venture/europe-largest-ai-round-mistral-seriesc/
- Euronews, Draghi calls for pause to EU AI Act, Sep 2025 — https://www.euronews.com/my-europe/2025/09/16/draghi-calls-for-pause-to-ai-act-to-gauge-risks
- Tortoise Media, The Global AI Index 2024, Sep 2024 — https://www.tortoisemedia.com/2024/09/18/the-global-artificial-intelligence-index-2024
- Alex Karp, CEO of Palantir — public remarks (quote as supplied; exact wording not independently verified in press reporting).