The AI Ecosystem Monitor
Where is the money going—and who benefits?
Three arcs, twenty chapters: where capital flows, how the stack works, and what it means for work and daily life. Sourced from Stanford HAI, OECD, McKinsey, IEA, Epoch, and public filings.
$965B on the table: the US still leads
About $965B went into AI in 2025—roughly half as much again as the year before. A big slice of venture dollars still lands with a handful of labs; the US, China, and UK together account for roughly two-fifths of the total.
Sovereign funds are now writing checks that rival whole countries’ private markets—think one Gulf fund in the same ballpark as Germany’s AI total. OpenAI’s 2025 raise was the largest startup round on record.
Three names at the top—open weights are closing in
As of March 2026, Claude, Gemini, and GPT sit in the same narrow band at the top of composite scores—no single winner. Open models like GLM-5 and Qwen are only a dozen points behind, not a generation.
What cost $60 per million tokens in 2023 can be three cents today. Context windows grew from thousands of tokens to millions—enough to hold a whole repo in one prompt.
Patents: China files most; the US gets cited more
China leads on volume; US patents still punch far above on citations. India’s filings are climbing fastest year on year.
| Country | Filings (2024) | Global Share | Avg Citations | CII Score |
|---|---|---|---|---|
| 🇨🇳 China | 60,200 | 40.1% | 1.8 | 0.56 |
| 🇺🇸 United States | 45,100 | 30.1% | 5.8 | 1.80 |
| 🇯🇵 Japan | 8,400 | 5.6% | 3.1 | 0.97 |
| 🇰🇷 South Korea | 7,200 | 4.8% | 2.9 | 0.88 |
| 🇩🇪 Germany | 6,100 | 4.1% | 4.2 | 1.28 |
Unicorns: one lab is worth more than a Wall Street giant
OpenAI’s private valuation has passed major banks. The top 20 AI unicorns sum to hundreds of billions; foundation-model labs take the lion’s share, legal and creative AI are among the fastest-growing corners.
Talent: three job openings for every ML hire
Roughly a quarter of the world’s AI researchers sit in the US; salaries have risen steeply since 2020. Senior people at frontier labs can earn seven figures—not counting equity.
A large share of China- and India-trained AI PhDs work outside their home countries; the US and UK soak up most of the flow. Europe exports talent to Anglo markets too.
Readiness: Singapore punches above its weight
Scores blend strategy, money, rules, talent, and access to chips. The US and China sit at the top; smaller hubs like Singapore still rank high. Many countries in Latin America and Africa still have no national AI plan.
| Country | Strategy | Funding | Regulation | Talent | Compute | Total |
|---|---|---|---|---|---|---|
| 🇺🇸 United States | 95 | 98 | 82 | 94 | 96 | 91 |
| 🇨🇳 China | 96 | 90 | 80 | 91 | 80 | 88 |
| 🇬🇧 United Kingdom | 88 | 78 | 90 | 82 | 79 | 84 |
| 🇸🇬 Singapore | 90 | 74 | 89 | 79 | 75 | 82 |
| 🇨🇦 Canada | 82 | 72 | 85 | 84 | 71 | 79 |
Few countries fab their own accelerators; export rules make that dependence strategic. Most governments still lack a clear AI framework.