Economic Impact of AI and the Geography of Innovation

Forecasts of AI’s payoff differ several-fold, so the first skill is to know what kind of number you are reading. Goldman Sachs Research estimated in April 2023 that generative AI could raise global GDP by 7% (almost $7 trillion) and lift productivity growth by 1.5 percentage points over ten years. McKinsey put the potential value of generative AI at $2.6 to $4.4 trillion a year across 63 use cases it analyzed (June 2023). Both are scenarios: they assume how fast AI is adopted and then compute what follows. They are not measurements, and they do not agree with every economist (see the box below).

The key mechanism is productivity. AI augments human work — allowing one professional to produce more in the same hours. In economies facing severe demographic pressure — aging populations in the US, Europe, Japan, China, and South Korea are shrinking the working-age share of the population — AI-driven productivity growth is potentially the primary mechanism sustaining economic output as workforces contract. This makes AI not just a commercial opportunity but a genuine policy imperative for governments facing demographic decline.

But the distribution of AI’s economic benefits will be deeply unequal, at least initially. The companies building and controlling AI infrastructure and foundational models — primarily a small number of US technology companies — will capture an enormous share of early value creation. Countries with the talent, capital, and policy environment to deploy AI quickly will grow faster. Countries dependent on tasks that AI automates without clear alternative employment pathways will face significant labor market disruption.

Software travels at light speed, so innovation should theoretically happen anywhere. In practice it happens in about a dozen postcodes, because of what economists call the agglomeration effect: engineers switch jobs across the street carrying ideas with them, investors can meet five startups before lunch, universities feed talent in, one generation’s successful founders become the next generation’s angel investors (chapter 4.7: Venture Capital and Private Equity — The Pre-IPO World), and failure carries no stigma because the ecosystem instantly reabsorbs you. Each ingredient strengthens the others — which is why clusters, once formed, are extraordinarily hard to dislodge, and why governments worldwide keep trying (and mostly failing) to build “the next Silicon Valley” by decree.

Silicon Valley is the archetype, born of an unusual recipe: Stanford University pushing professors and students toward industry, Cold War defense spending seeding the first electronics firms, the 1957 founding of Fairchild Semiconductor — whose alumni (“Fairchildren”) went on to found Intel and dozens more, giving the valley its silicon and its name — and the invention of modern venture capital on Sand Hill Road. Seventy years later it still commands the world’s deepest pool of risk capital and has been supercharged by the AI era: Nvidia, OpenAI, Google, Meta, and Apple — the very companies whose capex dominates chapter 7.3: The Financial Flows — Where the Trillions Are Going — all sit within an hour’s drive of each other, a concentration of market value with no historical precedent.

Bengaluru is the century’s great counter-example — proof that clusters can be built outside California. Public-sector science (HAL, ISRO, IISc) seeded the talent base; Texas Instruments’ 1985 arrival proved satellite links could put Indian engineers on global projects; Infosys and Wipro industrialized software services through the 1990s; Y2K and the telecom boom scaled it; and each wave compounded into the next — the GCC era (chapter 10.6: From BPO to GCC — India’s Global Delivery Economy), then India’s startup explosion, with Bengaluru hosting the largest share of the country’s hundred-plus unicorns, and now a fast-growing role in global AI engineering, from hyperscaler research teams to GCC-based model and data work. The city is, in effect, where several chapters of this guide physically live. Hyderabad (Microsoft’s and Google’s largest campuses outside the US, plus pharma), Pune, Chennai, and the NCR complete India’s supporting cast.

The rest of the map, briefly: Shenzhen — hardware’s capital, where China’s first special economic zone (1980) grew from fishing villages into the place a prototype becomes a mass-manufactured product in weeks, home to Huawei, Tencent, BYD, and DJI, and the muscle behind chapter 9.1: China — The State-Capitalist Superpower‘s manufacturing dominance. Hsinchu, Taiwan — home of TSMC’s headquarters, R&D and some of its most advanced fabs, on an island whose concentration of leading-edge production is the geographic single point of failure that makes chapter 7.4 a chapter about geopolitics rather than merely chips. Tel Aviv — where military signals-intelligence units function as the world’s most effective startup incubator, especially in cybersecurity. Seattle (Microsoft, Amazon — the cloud’s home town), London (Europe’s fintech capital and DeepMind’s birthplace), and Boston (biotech) round out the network.

Now stitch the geographies together, because they work as one machine: a Bengaluru startup is typically funded from Silicon Valley and Singapore, structured through Luxembourg or Singapore vehicles (chapter 9.5: The Safe Havens — Why Small, Stable Countries Hold the World’s Money), banked across the centers of chapter 9.6: The Great Financial Cities — A Map of Where Money Lives, and exits via an IPO in Mumbai or New York (chapter 4.4: The IPO Process — How a Company Goes Public) — at which point its early employees become the angel investors of the city’s next wave. Innovation geography and financial geography are two halves of a single circulatory system; a country that hosts one and not the other captures only half the value, which is precisely why India’s simultaneous bets on its tech cities and on GIFT City are, at bottom, the same bet.