The Administration's AI Build-Out Has a Fed Problem

In early July, Treasury career analysts completed a draft report comparing the AI market to the dotcom era, warning that AI firms are more deeply entrenched in the economy and that a downturn would send shockwaves through stock markets, chip manufacturers, utilities, and private credit ^[1]. Treasury Secretary Scott Bessent dismissed the finding.

I think they were slightly stunned when I said the biggest risk to AI is China getting ahead of us. — Scott Bessent

The Federal Reserve has now independently identified the same AI infrastructure build-out — announced data center plans exceeding $1.5 trillion — as a primary new inflation threat, with Governor Lisa Cook and New York Fed President John Williams citing rising costs for memory chips, electricity, and copper ^[2]. The Bank for International Settlements issued its own warning in late June: over $1 trillion in hyperscaler capital expenditure across 2025 and 2026 is outpacing earnings, and supply-side bottlenecks in semiconductors and electricity are already contributing to consumer inflation ^[3]. The build-out Bessent is shielding from scrutiny is the same one the Fed and the BIS have independently identified as a systemic risk. The dismissal was not an off-script aside. Since early 2026, the executive branch has systematically designated AI infrastructure — data centers, chips, and the uranium fuel powering them — as critical national security assets, binding the build-out to a strategic competition with China. In April 2026, President Trump signed an executive order designating data centers as critical national security infrastructure, explicitly framing over 1,500 facilities in development as strategic assets tied to US technological dominance ^[4]. The same month, he ordered a quadrupling of US nuclear capacity to 400 gigawatts by 2050, driven by AI data center demand — Constellation is restarting Three Mile Island for Microsoft, Westinghouse signed an $80 billion government reactor deal, and uranium supplier Cameco has positioned itself as a national security asset ^[5]. At a March roundtable, Trump made the frame explicit: the nation leading AI will emerge as the dominant military superpower ^[6]. China plans to invest $295 billion over five years in a nationwide network of AI data centers, mandating at least 80% of core technology be sourced from domestic suppliers like Huawei ^[7]. The security designation does not cause the build-out — commercial AI demand is the underlying driver — but it shields it from constraint. And the build-out is enormous. Fifty-one US utilities plan to spend $1.4 trillion on grid upgrades through 2030, a 21% increase over the prior five-year period, with the industry group PowerLines warning that the capital spending signals future rate increase requests ^[8]. Data center vacancy sits at record lows — 0.3% in Northern Virginia — with a 12-gigawatt global capacity deficit and $2 trillion in cloud backlogs, meaning the demand pressure is structurally real, not speculative ^[9]. The migration of inference from cloud to edge may eventually ease the pressure, but with vacancy near zero and a 12-gigawatt deficit, that relief has not yet materialized. The price spillovers are already arriving. Ohio residential electricity bills are projected to average $800 per month this summer, a 17% increase attributed to data center growth, manufacturing, and electrification outpacing new generation ^[10]. Nine industry trade groups — representing automakers, retailers, and medical device manufacturers — warned the administration that AI data-center demand is consuming a disproportionate share of memory chip capacity, creating significant and sustained near-term price increases for American households, with SK hynix estimating shortages will persist until 2030 ^[11]. Apple and Microsoft have already raised prices on laptops, iPads, and Xbox consoles in response ^[2]. Headline PCE inflation has risen from 2.4% to 4.1%, with the Fed's semi-annual Monetary Policy Report identifying AI semiconductors and data-center infrastructure, alongside tariffs and the US-Iran conflict, as primary drivers ^[12]. The Fed is internally divided but tilting hawkish. Chair Kevin Warsh has publicly disputed the notion that AI will cause long-term inflation ^[13]. But Governor Cook and New York Fed President Williams have identified the $1.5 trillion in AI data center investments as a primary source of price pressure, and Governor Christopher Waller has invoked the 2021 mistake of not responding to inflation sooner ^[13]. Markets are pricing a roughly 90% probability of a rate hike by December 2026 ^[13]. The administration is not unaware of the cost spillover. In his February 2026 State of the Union, Trump introduced the Ratepayer Protection Pledge, a voluntary agreement with Amazon, Google, Meta, Microsoft, OpenAI, Oracle, and xAI to self-fund their power needs. He acknowledged that AI data center expansion is driving up household electricity bills and said the companies had an obligation to cover their own power costs. The pledge came with a guarantee ^[14].

The country needs to build new data centers quickly to maintain its competitiveness on AI and national security—but AI companies shouldn’t leave American ratepayers to pick up the tab. — Anthropic

The pledge has no enforcement mechanism. Critics have dismissed it as an unenforceable handshake ^[14]. Senators Richard Blumenthal and Josh Hawley have introduced the GRID Act to legally mandate that data centers source power outside the grid; Senator Bernie Sanders has proposed a nationwide moratorium ^[14]. State regulators are moving on their own: the Pennsylvania Public Utility Commission advanced a model tariff targeting large data centers to prevent them from shifting expenses onto residential and small business customers, with Chairman Stephen DeFrank warning that generations of Pennsylvanians could be burdened by the cost of abandoned or unused infrastructure ^[15]. Local moratoriums on data centers have emerged in California, Maine, South Carolina, Texas, Minnesota, and Iowa, driven by concerns over electricity consumption, water usage, and noise ^[16][17]. The geometry is this: the executive branch has designated AI infrastructure a national security priority and subordinated financial-stability risk to the competition with China. The central bank is now signaling it may have to raise rates to contain the inflationary pressure from that same build-out. The only mechanism the administration has offered to bridge the gap — a voluntary pledge with no enforcement teeth — is the knot in the middle. If the Fed moves, the rate hike will land on the same AI equities whose valuations are financing the security build-out.