On a Thursday morning in June 2026, the stock market repriced the value of human thinking.
Accenture is not a household name unless your household contains a consultant, but it may be the purest publicly traded bet on white-collar expertise ever assembled: nearly eight hundred thousand people whose entire product is advice — analysis, integration, judgment, thinking by the hour, at industrial scale. On June 18, 2026, its shares suffered the worst single-day fall in the company’s history. Not because of a scandal. Not because the quarter was a disaster. Because of what clients were signaling about the quarters to come: they were pausing exactly the kind of work Accenture sells, and the question moving through the analyst call was polite, professional, and lethal — how much of this will a machine soon do for approximately nothing?
Whatever the stock does next, that morning belongs in this book, because it was the market pricing a single idea in real time: intelligence — the thing every salary in the knowledge economy is paid for — is becoming an industrial input with a collapsing price.
This chapter is about what that force does to money. And it’s here, closing Part II, because almost everyone — including people who agree with every previous chapter — has its effect on this book’s thesis exactly backwards.
The backwards version goes like this: AI makes everything cheaper. Cheaper is deflation. Deflation makes each dollar buy more. So AI is the dollar’s rescue — the one trend big enough to bail out the arithmetic of Part I.
Half of that is true. Which is what makes the whole of it dangerous.
Concede the premise fully, because it’s real. AI is deflationary the way electricity was deflationary: not one cheaper product, but a cheaper input to nearly every product. Drafting a contract, reading a scan, writing the software, routing the freight, answering the customer — each is a task whose marginal cost is heading toward the price of the electrons involved. The cost of running these systems has been falling at rates that make Moore’s Law look leisurely, and everything built from thinking gets cheaper as thinking does.
If the story ended there, this book would end here too — a genuine deflationary miracle, compounding against the printer, and the honest move would be to tell you to hold your dollars and enjoy the discount.
The story does not end there, because an economy is not a price index. It is a political system with a printer attached. And the fifty years this book has already walked you through are a complete record of what that system does under stress.
Follow the chain one link at a time, and notice that no link requires a prediction — only the response function Part I already documented.
Deflationary for goods means brutal for labor, because labor is the cost being cut. The savings have to come from somewhere, and mostly they come from salaries. What’s different this time is whose. The great automation waves of the last century landed on muscle — the field hand, the line worker, the typist. This one lands on the credentialed middle: the paralegal, the junior analyst, the coder, the consultant, the copywriter, the radiologist’s reading queue. People with mortgages, degrees, expectations — and, critically, people who are numerous, articulate, organized, and registered to vote. The displaced weavers of 1811 could break looms. Displaced knowledge workers can break coalitions.
Now ask what the political system does when a large, loud class of voters loses income — or merely fears losing it, which arrives years earlier and votes just as hard. You don’t have to guess. Every financial shock for forty years was answered with liquidity; that was Part I. And 2020 added the missing piece of machinery: the government learned it can deliver money directly to households, at national scale, in days. The checks landed in Venmo once already. The pipes aren’t hypothetical anymore. They’re installed, tested, and popular.
So the chain runs: displacement — or its shadow — arrives; transfers expand. Retraining funds, expanded credits, pilot basic incomes; a sitting congressman has already proposed a wealth tax aimed at fortunes above fifty million dollars, part of a broader agenda framed explicitly around who wins from AI. Whether any single proposal passes is beside the point; the direction of pressure is the point. Deficits — already running around six percent of GDP at full employment, before a single displacement wave — grow. The printer, which Chapter 5 showed is now fiscal, permanent, and bipartisan, accelerates. Economists have begun building formal models of exactly this loop — abundance in production, deficiency in demand, transfers as the bridge — but you don’t need the math. You need the last fifty years, and you’ve just read them.
AI doesn’t break the fiscal machine. It feeds the machine the one input it runs on: a population that needs the checks to keep coming.
There’s a quieter mechanism underneath all this, and it answers the deflation argument permanently, so give it a page.
Suppose the optimists are right and AI pushes the price of nearly everything down. Here is the uncomfortable fact about the monetary system you live in: it is explicitly designed not to let that happen to your dollars. The Federal Reserve targets two percent inflation — not zero, and never negative. When technology pushes the price level down, policy’s mandate is to push it back up until the number reads two again. Deflation isn’t treated as a dividend to savers; it’s treated as a malfunction, and the cure for the malfunction is issuance.
This isn’t a conspiracy theory; it’s the stated framework, and you’ve already lived through a version of it. For thirty years, globalization and electronics handed the dollar an enormous purchasing-power dividend — the falling-TV-price decades, when the tradable-goods half of the price index deflated year after year. Did your dollars buy more over those decades? They did not. The technology dividend was offset — absorbed as room to run everything else hotter — and the melt continued on schedule: housing, tuition, medicine, insurance, assets. The saver’s share of the productivity miracle was, by policy, zero.
Now scale that up. AI may be the largest deflationary impulse in the history of the two-percent target. If the framework holds — and no one in either party proposes changing it — the offset will be the largest accommodation in the history of the printer. The gains are real; the goods get cheaper; the currency doesn’t get sounder, because soundness is precisely what the framework is built to prevent. Cheaper goods, easier money, and the gap between the two flows — where it always flows — into everything that can’t be printed.
So much for the theory. Close the chapter the way a trader would: ignore what everyone says and watch what the biggest players do.
The most profitable companies in human history are sitting on cash machines that throw off hundreds of billions of dollars a year. They could hold that cash in Treasury bills and collect a riskless four-something percent. Instead, they are locked in the largest capital-expenditure arms race ever recorded — converting cash into chips, land, water rights, transformers, turbines, and decades-long power contracts as fast as the physical world will absorb it, at a combined pace in the hundreds of billions per year, and increasingly borrowing to do it, in size that is visibly bending the shape of the corporate bond market. One of them paid to bring a shuttered nuclear plant back from the dead to feed its datacenters.
Strip the technology out and look at the trade underneath. Entities with the world’s deepest visibility into what’s coming are dumping the melting asset — cash — for scarce real inputs: energy, silicon, land, grid capacity. That is the debasement trade of this entire book, executed by the most sophisticated treasurers alive, in the largest size available. They would tell you they’re buying compute, and they are. They are also telling you, with every dollar spent, what they think a dollar held is worth.
You cannot buy a gigawatt. But you can notice what the people who can are choosing — and notice that the assets in this book’s subtitle belong to the same category they’re stampeding into: the small set of things whose supply does not answer to demand.
The wrench, then, hits twice. The political response to AI prints more dollars. The industrial response bids up everything scarce that dollars chase. A richer economy and a weaker currency, arriving together — which sounds like a contradiction until you remember it’s just Chapter 1 again: the economy grows, the ruler shrinks, and everything real gets repriced in melting units.
Part II is complete. You’ve seen the exits: a money whose supply is fixed by rule, a financial system that replaces trust with mathematics, a dollar that wins the job nobody profits from and loses the one everybody depends on, and now a technological revolution that accelerates the fiscal endgame it was supposed to rescue us from. What remains is the practical part — Part III — and it opens with the question every reader of a book like this is entitled to ask with narrowed eyes: fine — but why now? The next chapter answers it the only honest way: with conditions you can check, not promises you can’t.
AI makes the economy richer and the currency weaker — at the same time.
[Charts for this chapter: (1) the cost of intelligence — inference cost per unit, 2022–2026, log scale, falling; (2) the response function — U.S. transfer payments as a share of GDP, 1970–2026, stepping up at each crisis; (3) the arms race — combined hyperscaler capex vs. their aggregate cash returned to shareholders, 2019–2026. Endnotes: Accenture episode (June 18, 2026 — news pipeline); Northern Trust “Is AI Inflationary or Deflationary?” (June 2026); “Abundant Intelligence and Deficient Demand” (arXiv 2026) for the displacement→transfers model; Khanna wealth-tax proposal (July 2026); Goldman credit commentary on AI-driven IG issuance; Microsoft/Three Mile Island PPA coverage. Tone note: the falling-TV-decades offset beat is the chapter’s load-bearing economic claim — have the fact-check pass confirm the framing (durable-goods deflation vs. services/asset inflation, 1995–2020) survives scrutiny. All inline [VERIFY] tags resolved at fact-check pass 2026-07-26 — see VERIFICATION-partII.md. Sourcing notes: Accenture fell ~18% on Jun 18, 2026 (record one-day drop per Bloomberg; 798,739 employees as of May 31, 2026); Khanna’s Jul 2, 2026 essay endorses a 2% tax above $50M but does not itself invoke AI — the AI framing is from his broader “AI democratist” writings, hence the softened phrasing; big-four hyperscaler 2026 capex guidance ~$725B combined (~$700–900B incl. Oracle); Goldman: AI-related share of US IG issuance 1% (2024) → 7% (2025) → ~18% (H1 2026); deficits corrected to “around six percent” (FY24 6.4%, FY25 5.8%, FY26 proj. 5.8%, CBO).]