Two findings landed this weekend that, read together, tell a single story about the AI boom — and it is not the usual one about capability or valuations. It is about the bill.

The first is economic. New analysis from the Federal Reserve Bank of Minneapolis finds that the AI build-out's hunger for memory and computer hardware has pushed the price of video and information-processing equipment up 12.2% year over year through July — adding roughly 0.4 percentage points to core PCE inflation, as much as the entire contribution of the tariffs imposed early last year. A category of goods that reliably got cheaper for a decade has swung hard the other way. The spillover from the data-centre supercycle is no longer confined to hyperscaler balance sheets; it is showing up in the price of a laptop. (Our full report on the Fed finding.)

The second is societal. Bill Gates — for years one of AI's louder optimists — used a roughly 6,000-word essay to warn that the world has no real plan for the transition. He names three near-term risks: vanishing entry-level jobs, emboldened criminals armed with cheaper and more capable tools, and developmental harm to children raised alongside AI companions. His framing is stark: AI will be "the greatest equalizer ever invented, or the worst source of injustice." The prescription is the tell — a founder who built his career on moving fast is now calling for limits. (Our full report on the Gates essay.)

Set side by side, the two rhyme. Both describe the same gap: how fast the capability is arriving against how slowly the systems around it — economic, regulatory, social — are adapting. The Fed measures that gap in tenths of a percentage point on a price index. Gates measures it in the risks nobody has a plan for. Neither is a story about whether AI works. Both are stories about who absorbs the cost while it does.

The theme has been building all week. Capital keeps pouring into the physical layer — a16z's new $1.1bn Machine Age Fund, aimed squarely at the hardware build-out, is the latest vehicle — while a growing backlash trails the datacentres it funds. This weekend the conversation caught up with what that capital does on the way through: to prices, to jobs, to the people who never signed up for the experiment.

None of this is an argument that the boom is a mistake. Productivity gains at the scale the investment implies could dwarf a few tenths of a point on inflation, and Gates himself is careful to hold the upside open. But the optimist's case increasingly rests on a promise — that the benefits arrive before the costs compound — and this weekend put a number and a warning on the costs.

One deal that would redraw the week's map stayed in the wings: a reported roughly $12.9bn acquisition of Hugging Face by Nvidia remained unconfirmed over the weekend, with no signed agreement announced. If it lands, it is the build-out's next chapter. For now it is a report, not a done deal.

The choices, as Gates put it, are the point. This weekend the case for making them deliberately picked up two new pieces of evidence.