📈 Data to start your week
Exponential View Hannah Petrovic ● Covered by 3 sources
Monday's numbers show AI money flowing to a tiny few, not the whole field. Big labs cash in per token; workers, cities and even VCs stay stuck in old patterns.
Based on reporting by Exponential View, Hannah Petrovic — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Start with the money. On Vercel's AI Gateway, OpenAI, Anthropic and Google pull in 90 percent of spend while handling only 52 percent of the tokens processed. Do the math and each token from the Big Three brings in more than eight times the revenue of a token from anyone else. That's not a rounding error, that's a moat.
Meanwhile the labor story looks less dramatic than the headlines suggest. AI touches 68 percent of US occupations in some form, but inside any given job it's only handling about a fifth of the actual tasks. That's selective use, not replacement. And the productivity gains showing up in US data mostly trace back to companies squeezing more out of equipment they already own, not some leap in total factor productivity — the kind of gain that comes from genuinely new ways of organizing work.
A smaller but telling data point: users testing Cursor's Router in Auto Intelligence mode rated its output as good as Fable's, at roughly 60 percent lower cost. That's the kind of efficiency story that doesn't make front pages but quietly reshapes what tools people reach for.
Outside AI, the roundup drifts into economics and demographics that echo the same theme of concentration. Five percent of US venture capital firms generate 90 percent of investment profits. The Bay Area holds 91 percent of generative AI unicorn market cap and still claims 39 percent of the market cap across all unicorns, AI or not. Even a completely unrelated finding — Denmark's GLP-1 drugs cutting long-term sick leave by 17 percent over four years — shows the same pattern: the savings went to employers and public budgets, not to the workers taking the drugs.
Throw in that the number of kids under five in big US cities has dropped 15 percent over a decade, even where population is otherwise growing, and Latin America's EV sales are gaining ground on the US despite starting late, thanks to tax incentives. None of these are AI stories exactly, but together they paint a picture of an economy where gains cluster fast around whoever already has scale, capital or a head start, while everyone else picks up whatever's left.
My take — AI-written commentary, not fact-checked reporting
None of this should shock anyone who's watched tech markets for more than five minutes: winners take an outsized cut, and the rest fight over table scraps. What's worth sitting with is how thin AI's actual footprint still is inside most jobs — a fifth of tasks touched isn't transformation, it's a tool getting used where it's convenient. The real economic story might not be AI at all, it might be companies finally running the equipment and capital they already had a bit harder, which is a far less exciting headline than anyone in Silicon Valley wants to write.
Read more about this at: Exponential View