Last Thursday, the yield on the 30-year US Treasury bond broke above 5.5 percent for the first time since June 2004, and touched 5.53 percent on Friday. It was below 5 percent as recently as early July. The 10-year reached 5.22 percent, a level last seen in 2007. The selloff was global — Japan’s 10-year yield hit its highest since 1996 — and it was not only about the Federal Reserve, though markets now put better than 75 percent odds on another hike in October, up from 49 percent a week earlier.
The structural story arrived the same week the yields did. In its September 21 commentary, BlackRock put the AI buildout and government borrowing on the same side of the capital ledger: annual US financing demand could exceed $7.5 trillion by 2030, driven mainly by AI’s capital needs — the most intense competition for funding, the firm says, since the financial crisis and Covid. AI and data-center bonds already make up about 14 percent of US investment-grade issuance this year, up from 5 percent last year and 1 percent over the previous decade. Goldman Sachs made the same argument days earlier: data centers, power plants, transmission lines, and government deficits are all “long-duration claims on savings,” drawing from the same well — and Goldman now expects record investment-grade issuance of $2.3 trillion this year, a quarter of it from AI.
For most of the past fifteen years, capital was nearly free. This week the market posted a new price for the future. One financial historian wrote the book on what that price means.
The book that explains the price
Edward Chancellor is a Cambridge-trained financial historian, a former Lazard banker, and the author of Devil Take the Hindmost, the classic history of financial speculation. In 2022 he published The Price of Time: The Real Story of Interest, a history of interest from ancient Mesopotamia through John Law’s Mississippi bubble to the credit booms of this century.
His argument is that interest is not a technical banking fee. It is the price of time — the exchange rate between the present and the future — and because every valuation involves the future, it is the one price inside every other price. Warren Buffett’s old image makes the same point: interest rates act on asset values the way gravity acts on matter. When the price of time falls to zero, financial gravity is switched off, and anything can float.
What zero built
Chancellor’s history shows what happens when central banks hold rates near zero for a decade, as they did after 2008: time becomes free, and any payoff, however distant or speculative, can justify any valuation today. Capital flows into projects that make sense only at zero. The unproductive survive alongside the productive — his “zombie” companies, kept alive by cheap refinancing rather than earnings.
The AI buildout is the last great project of that mindset, and here the timing matters. Through 2024 and 2025, the buildout was funded mostly from cash piles accumulated in the zero-rate era — spending that never had to face a bond market. Only this year did it migrate onto one, that jump from 1 percent to 14 percent of investment-grade issuance. The largest capital project of the age arrived to borrow at scale in the exact season the price of time snapped back to a two-decade high. The buildout was designed in a world where waiting cost nothing. It is being financed in a world where waiting costs 5.5 percent.
The 5.5 percent hurdle
Here is what the repricing does, in one number. At a zero discount rate, a dollar of profit in 2040 is worth a dollar today. At 5.5 percent, it is worth about 47 cents. Every AI project promising distant returns just watched half its present value evaporate — not because the technology changed, but because the price of time did. And every project must now clear not just its own costs but the alternative: a 5.5 percent return from the US government, risk-free.
Chancellor’s framework says this works as a filter, not a verdict. Projects offering a genuine leap in productivity will clear the hurdle and find funding. Infrastructure built on hype and zero-rate arithmetic will not. Note that the institutions raising the alarm are not calling a crash: BlackRock remains overweight US equities and AI on earnings strength even as it warns about the fight for capital, and Goldman’s point is a mechanism, not a top. The reckoning Chancellor describes is selection pressure, and selection takes time.
What to watch next
The first hard evidence arrives in late October, when the big tech firms report earnings and — more important — capital spending guidance. This will be the first guidance written in a 5.5 percent world. Watch for stretched timelines, “phasing,” and joint ventures that move debt off balance sheets: the polite vocabulary of projects that no longer clear the hurdle.
Then watch the long end of the bond market rather than the Fed. If October’s hike lands, short rates will follow policy — but the 30-year yield is the market’s own verdict on the fight for capital, and it answers to no chairman. And keep an eye on that 14 percent issuance share. If AI borrowing keeps climbing while yields hold above 5 percent, the collision is still building. If issuance stalls, the filter has started to bite.
Readers of this newsletter have watched this story assemble in installments: Minsky gave us the anatomy of the AI debt, Dalio the political collision over rates. Chancellor supplies the final piece — the price. We have moved from innovation at any cost to innovation at a specific cost. So far, that cost is 5.5 percent, and everything built on the assumption that time was free is about to find its true value.
This week’s book: Edward Chancellor, The Price of Time: The Real Story of Interest (Atlantic Monthly Press, 2022).
Sources: Bloomberg (Treasury yields, Sept 24–25, 2026); CNBC (bond selloff and October hike odds, Sept 24, 2026); Axios (yield moves, Sept 24, 2026); BlackRock Investment Institute weekly commentary (Sept 21, 2026); Goldman Sachs research via press reports (Sept 2026). Book: Edward Chancellor, The Price of Time (2022).