The Fed Turned the AI Boom Into Its Inflation Problem
In five months the Federal Reserve made data-center construction a monetary-policy variable — and it is now hostage to the very boom it is hiking against.
There is an odd sort of comedy in a central banker ranking his preferred inflations. In October, Chicago Fed President Austan Goolsbee did exactly that. Inflation from tariffs or war, he said, is the good kind, because it goes away on its own.
If it’s coming from overheating demand, from AI data center build out, that’s a little more disturbing and we’d have to address that. — Austan Dean Goolsbee
The other kind — overheating demand from data-center construction — is the one the Fed would have to answer. Five months earlier that answer didn't exist. Now it is policy. In mid-July, Fed Governor Lisa Cook and New York Fed President John Williams flagged more than $1.5 trillion in announced data-center plans as a primary new inflation threat, pointing to memory-chip costs that had already pushed Apple and Microsoft to raise prices on laptops, tablets, and Xbox consoles [1]. By late September the Fed had voted, lifting its benchmark rate to 3.75% to 4%, with Cook noting electricity and water were each up about 5% year over year as data centers bid up shared resources [2]. In November, Vice Chair for Supervision Michelle Bowman put a number on the whole thing: AI-related price pressure accounted for roughly one-fourth of above-target core inflation over the year through September, while Governor Christopher Waller pushed for a half-point December hike and the White House's Council of Economic Advisers called that an overreaction [3]. The mechanism is physical before it is monetary. The buildout bids up scarce chips, power, water, copper, and the HVAC and construction crews needed to raise a data center, until what began as sector overheating goes aggregate [4]. Then there is the demand side — Goolsbee's worry that consumers will begin spending ahead of AI gains they have not received yet.
Two forces, the building up of data centers ... and consumers spending out of future wealth, those can easily overheat the economy ... The Fed's job is to prevent this thing from overheating the economy. — Austan Dean Goolsbee
The center of gravity is that the buildout gets hit twice. Its own borrowing did part of the Fed's work for it. In late July the top five AI firms' debt issuance coincided with the 30-year Treasury yield touching 5.28%, the highest since 2007, while the Fed still held rates steady [5]. Analyst Casey Sprake offered a read on who was really doing the tightening.
the bond yield curve is actually doing the the work for the Fed in that — Casey Sprake
He argued the rising cost of debt was restraining the economy more effectively than the central bank's own decisions.
the market is being sort of punch drunk around anything AI CapEx spend. — Casey Sprake
And every hike the Fed does make reprices the next tranche of borrowing. Some neocloud bonds already yield 15%, Goldman Sachs has nearly doubled its 2027 AI project-finance issuance forecast to $545 billion, and markets put 93% odds on a December hike [6]. The bite is selective: Microsoft and Alphabet ride on their balance sheets, while the leveraged periphery — neocloud providers and the special-purpose vehicles that fund them — absorbs the rate [7]. The Fed's own October stability report flags the same thing, warning of AI debt growing through opaque financing structures and concentrated in a handful of companies [8]. Amazon's Andy Jassy has noted that data-center capital is spent two years before the servers can be monetized [9]. The loop was forecast back in January, when Royal London Asset Management named the instrument that would eventually prick the bubble.
You need a pin that pricks the bubble and it will probably come through tighter money. — Royal London Asset Management
Morgan Stanley projected U.S. inflation above 2% through 2027, driven by two specific costs.
The costs are going up not down in our forecast, because there’s inflation in chip costs and inflation in power costs. — Morgan Stanley
In October it reversed. CoreWeave bonds priced at a 63% default probability, the Kospi fell 20% in a week, and the Nasdaq 100 shed nearly 6% [10]. Bond yields dropped as AI-financing worries eclipsed hot inflation data and near-$98 oil. Warsh's remarks in Bangkok were read as signaling a pause on hikes — exactly when the market's own tightening bit [10][11]. A central bank that had spent the summer promising to cool the buildout flinched the moment it actually cracked. The Fed's leeway is hemmed in on every side. There is a federal debt above $40 trillion with annual interest costs over $1 trillion, and economist Robin Brooks argues Warsh and Treasury Secretary Scott Bessent operate under a de facto accord to keep long-term yields from blowing out [11]. Analysts warn the hikes barely touch AI capital spending, which is largely insensitive to interest rates, while crushing rate-sensitive housing instead [12].
one size doesn't fit all here with regard to monetary policy. — Jack Janasiewicz
The White House's own economist, NEC Director Kevin Hassett, argues there should be no further tightening at all, saying AI's supply-side benefits are being underestimated [2].
The growth benefit of it is way bigger than you might think. — Kevin Hassett
Outside the Fed, the make-the-buildout-pay logic has gone mainstream. North Carolina's utilities commission put the principle in plain terms.
These risks must be borne either by existing ratepayers, who have no responsibility for the imposition of the risks, or the data centers, who are imposing these risks. — North Carolina Utilities Commission
Grant County, Washington proposed 2027 rate increases of 24% for large data centers versus 3.5% for households and farms [13]. Illinois consumer advocates are fighting grid-upgrade costs being passed to residential bills [14]. The IMF's Kristalina Georgieva, projecting 4.7% global inflation driven partly by AI demand, reached the same conclusion as the Fed [15].
This calls for a tightening of monetary policy in many economies. — Kristalina Georgieva
The Swiss National Bank's Petra Tschudin warned of redirected investment flows and chip shortages pushing prices up in the near term [16].
Shortages can occur, for example with chips, causing prices to rise. — Petra Tschudin
China's central bank is running the opposite experiment, committing to subsidize its own AI industry — while Fed Chair Kevin Warsh has characterized AI as a positive shock to the economy even as his committee hikes against it [17].
AI was a positive economic "shock" that could "fundamentally" change monetary policy. — Kevin Warsh
The bluntness is the point. Goldman expects AI capital-expenditure growth to slow for the first time in 2027 [18]; token prices are falling even as the physical inputs inflate [19]; and the October inflation reading that locked in the December hike was driven largely by energy, not chips [20]. A tool this broad cannot aim at the buildout without grazing everything else. Muzinich & Co.'s chief expects rates to keep rising on sticky inflation — unless a correction in the AI market forces the Fed to cut. Policy is now indexed to the AI cycle in both directions [21]. Which is what makes the December 8–9 meeting the bind it is: a central bank trying to fight inflation created by an investment boom it cannot afford to break, with October's flinch already on the record.
- 1. Federal Reserve Flags AI Infrastructure as New Inflation Threat
- 2. Federal Reserve Officials Warn AI Boom Drives Inflation
- 3. Federal Reserve Officials Signal Rate Hikes Amid AI Inflation
- 4. Austan Goolsbee Warns AI Expansion May Overheat Economy
- 5. AI Capital Spending Drives US 30-Year Bond Yields Higher
- 6. Fed Rate Hikes and AI Debt Drive Bond Volatility
- 7. AI Boom Drives Inflation and Potential Federal Reserve Rate Hikes
- 8. Federal Reserve Warns of AI Investment Financial Risks
- 9. Tech Giants Shift S&P 500 Risk Toward AI Speculation
- 10. Neocloud AI Collapse Triggers Market Shift and Fed Pause
- 11. Federal Reserve Faces Scrutiny Over Debt-Driven Monetary Policy
- 12. Analyst Warns Federal Reserve Faces Two-Speed Economy
- 13. US Utilities Implement Targeted Rate Hikes for Data Centers
- 14. Illinois Data Center Growth Drives Up Residential Power Costs
- 15. Fed and IMF Warn AI Boom Drives Global Inflation
- 16. Swiss National Bank Warns AI May Increase Short-Term Inflation
- 17. Central Bankers Evaluate AI Impact on Global Monetary Policy
- 18. Goldman Sachs Advises Shift Toward AI Adopter Stocks
- 19. Morgan Stanley Warns Open-Weight AI Models May Pressure Pricing
- 20. October CPI Rise Makes December Fed Rate Hike Likely
- 21. Muzinich & Co. CEO Warns Geopolitics Complicate AI Investments