# The Two Curves Anthropic's revenue is bending upward faster than almost any company in history. We must choose a path that leads to accelerated socioeconomic development. On the distance between what the machines can do and who is using them. - Region: 999 - Updated: 2026-05-27 - URL: https://xecon.dev/999/deep-dives/the-two-curves - Type: Deep dive · Analysis ## The climb On May 28, 2026, Anthropic closed a US$65 billion funding round at a US$965 billion valuation — eclipsing OpenAI's $852 billion for the first time, and putting a maker of thinking machines within sight of a trillion dollar valuation. Days earlier, the company's annualized revenue run-rate had crossed US$47 billion. A year ago it booked about $10 billion for the entire year. (Chart: Anthropic annualized revenue run-rate, US$ billions (run-rate, not trailing revenue)) - Run-rate ($bn): 9, 14, 19, 30, 47 Source: Anthropic Series H; CNBC; VentureBeat Two caveats. First, a run-rate is the latest month annualized. This is a photograph of a moving object. The trajectory can change. Second, no exponential runs forever; the question is not whether growth slows but from what altitude. Even a sharp deceleration leaves a company that arrived in years where others took decades. Among the round's backers, notably, sits Temasek — Singapore's sovereign investor. Hold that name; it returns below. > The revenue curve measures what the machine can sell. It says nothing about who, on the other side of the world, can use it. What world awaits us as a gigantic new economic force matures? ## The divide Now lay a second curve over the first. Anthropic's own Economic Index measures where and how its model is actually used. Its central instrument is the AI Usage Index. This shows Claude usage divided by its share of the working-age population. A value of 1.0 means a place uses the tool in exact proportion to its size; above 1.0 is over-representation, below is under. (Chart: AI Usage Index, selected economies, 1.0 = usage proportional to working-age population) - Singapore: 5.53 - U.S.: 4.58 - U.K.: 3.71 - Ghana: 0.36 - Nigeria: 0.27 - India: 0.28 Source: Anthropic Economic Index, February 2026 dashboard [Off the chart, literally] Singapore now leads the world at 5.53× its population share, having overtaken Israel. The United States sits at 4.58. Ghana (0.36) and Nigeria (0.27) are on the index. This is good. Liberia is not. With fewer than forty observed conversations it falls below the index's 200-conversation floor, and is left unranked. Where the gap is widest, the data thins to nothing. The pattern is not noise. Usage tracks income tightly. A 1 percent rise in GDP per working-age capita is associated with roughly a 0.7 percent rise in usage per capita. Usage is concentrating, not spreading: the top twenty countries' share of per-capita usage rose from 45 to 48 percent in a single reporting window. The gap between rich and poor widens. ## What income explains, and what it doesn't Income is the first explanation, and it is incomplete. Move the slider below. It raises a region's GDP per working-age capita and projects its usage index forward at the Economic Index's own 0.7 elasticity, starting from Nigeria's 0.27 and measured against Singapore's 5.53. (Interactive: Interactive · the adoption gap — Projection applies the AEI's ~0.7 GDP-to-usage elasticity to Nigeria's 0.27 base, against Singapore's 5.53 frontier. Illustrative, not a forecast.) Even doubling income per capita leaves Nigeria far short of parity, let alone the frontier. The rest of the gap is structural. Language coverage, payment rails that accept the subscriptions, reliable power and connectivity, and the simple presence of people who have done the work before. The composition of usage differs too, and it is revealing. (Chart: What Ghana uses Claude for, % of conversations · top tasks) - Coursework: 8.5 - Biz strategy: 3.9 - Web dev: 3.8 - Teaching: 3.2 - Maths: 3.1 Source: Anthropic Economic Index, Feb 2026 dashboard (Chart: What Nigeria uses Claude for, % of conversations · top tasks) - Coursework: 6.1 - Social media: 5.1 - Web dev: 4.5 - Video/audio: 4.2 - Marketing: 3.5 Source: Anthropic Economic Index, Feb 2026 dashboard Two countries, two emphases. Ghana leans on the classroom with coursework, teaching materials, mathematics. Nigeria leans on the studio and the storefront with social-media content, video, marketing. Both have a web-development base. Both are the tools of people teaching themselves, which is the most hopeful reading available. The autodidacts have a new superpower. Where the institutions have not yet arrived, the individuals are not waiting for them. ## The Singapore exception Singapore uses these tools at more than five times its population share. First in the world, ahead of the United States, from an island with no natural resources and roughly a quarter of Nigeria's people. That is the same playbook that carried Singapore from poor port to high-income economy in one generation, now pointed at artificial intelligence. Development minded strategists must take note. Pay attention to what the proven economic development accelerators are up to. These moves are deliberate. Singapore's Smart Nation strategy and its National AI Strategy commit the state to tripling its pool of AI practitioners to 15,000, to crowding in compute, and to over S$1 billion for AI research and talent across 2025–2030 — with a National AI Council chaired by the prime minister and national missions in finance, manufacturing, connectivity, and health. The country treats AI fluency the way it once treated English and engineering. The tools of cognition and creation are national infrastructure, built on purpose. [On both sides of the curve] Let's focus on Temasek, the Singaporean state investor in Anthropic's $65 billion round. Singapore funds the frontier and uses it more intensively than anyone. It sits on both curves at once. The lesson for developing regions is not Singapore's size or its politics. The sequence is humanist. People, human capital comes first. Add this to institutions that work, openness to the world, and a plan measured in decades. ## What would Lee Kuan Yew say after playing with Claude Code? Hand the architect of Singapore's ascent Lee Kuan Yew the reins of a developing region or nation for a decade. Show him agentic AI tools. Perhaps he would land on three key insights. [People are the only resource] "We have no natural resources," he said; "our people are our only resource." In the age of thinking machines the lesson sharpens: treat AI fluency as the new literacy, build it at mass scale, and import the talent you lack without ego. Build the muscle that complements the machine. Focus on judgment, deployment, and data. [Clean, capable institutions before all else] Lee held that corruption and patronage, not poverty, are the binding constraint. He made Singapore one of the cleanest governments on earth, through design and funding. An AI-enabled patronage network simply industrializes graft and surveillance. AI laid over honest, merit-staffed institutions could compound trust and expedient service delivery. A key question to ask is how can these tools accelerate service delivery and identify bottlenecks with cheap and abundant data analysis. [Be pragmatic, open, and patient] His test for any policy was not ideology but "would it work?" Test ideas, adopt what demonstrably works, and plan in decades, not budget cycles. Applied to AI: move AI efforts out of short-term election-cycle dynamics. Fund a decade of talent and endow local institutions. One honest asterisk belongs here. Lee's results are inseparable from a soft-authoritarian, one-party state, and that is the contested part of his record. The transferable core is the three principles above — human capital, clean institutions, pragmatic openness — not his methods of political control. A democracy can choose the principles without the cage. ## Let's bring it home Bronzeville and Chicago live inside that U.S. usage index at 4.58. The national number hides the neighborhood, and within-country usage is still concentrating. For the South Side the question is not whether the country is at the frontier; it is whether the block is. Ghana and Nigeria are on the board, at roughly a third of parity. Liberia is not yet on the board at all. There is a thread of hope in the newest data. Experience compounds: users with six months of tenure show about a 10 percent higher success rate, an edge that survives controls for task and country. Access is the first lever. Tenure — the unglamorous accumulation of practice — is the second, and it is the one a community can build for itself. This is a new skill and priority. You don't know what to do with these tools without trying them yourself. How do we get more doers on the flywheel? Charity begins at home. The thought experiment was fun, but we want a fairer, richer world for as many as practical. We want this as fast as possible. Perhaps the journey starts simply. If you are using these tools and find that they are accelerating you, maybe take a couple of hours a week to help someone else get on board. Jump on a call. Install the tools, start the iterative improvement. Most people are cooperative and well meaning. The more people we get building, pursuing their goals outside of this initial concentration of wealth and activity, the better for us all. > Two curves: what the machines can do, climbing vertically; who can use them, nearly flat across the diaspora. The space between is the work. [Thinking Machines] A standing xecon series on how artificial intelligence meets — or misses — the work of accelerated development across the places we cover. We return to the usage index each time the Economic Index updates, and measure the four places against the frontier. ## Sources 1. [Anthropic, "Anthropic raises Series H" (May 28, 2026)](https://www.anthropic.com/news/series-h) 2. [CNBC, "Anthropic valued at $965 billion, eclipsing OpenAI" (May 28, 2026)](https://www.cnbc.com/2026/05/28/anthropic-open-ai-startup-value.html) 3. [Anthropic Economic Index dashboard (February 2026 data)](https://www.anthropic.com/economic-index) 4. [Anthropic Economic Index: Uneven geographic and enterprise AI adoption (September 2025)](https://www.anthropic.com/research/anthropic-economic-index-september-2025-report) 5. [Anthropic Economic Index: Learning curves (March 2026)](https://www.anthropic.com/research/economic-index-march-2026-report) 6. [Singapore National AI Strategy (NAIS 2.0) / Smart Nation](https://www.smartnation.gov.sg/initiatives/national-ai-strategy/) 7. Lee Kuan Yew, From Third World to First (2000) ## Editor's notes - 2026-05-27: Draft for review. Revenue is annualized run-rate (not trailing revenue); the funding round, valuation, and AUI figures are sourced to the linked reports. Liberia is below the index's 200-conversation reporting threshold and is therefore unranked.