Recent Federal Open Market Committee (FOMC) minutes, released under new Fed Chair Kevin Warsh, reveal a significant debate among policymakers regarding the future path of interest rates. While there was unanimous agreement to maintain the policy rate between 3.5% and 3.75% for the fourth consecutive meeting, a clear division emerged. Nine participants either projected no further rate adjustments or anticipated rate cuts if inflationary pressures subsided, while another nine foresaw at least one quarter-point rate increase by year-end if inflation remained elevated. This marks a rare instance where artificial intelligence (AI) capital expenditure was explicitly cited as a structural inflationary driver.
The minutes highlighted concerns among many participants that inflation remains elevated, with upside risks to price stability still significant. Broader price pressures were noted, and several flagged AI-related demand as a potential source of increased costs for technology products and electricity. This challenges the traditional view that technological advancement is inherently disinflationary in the short term. The AI buildout is seen as an infrastructure cycle, requiring substantial investment in electricity, semiconductors, grid upgrades, cooling, copper, skilled labor, financing, and permitting, which can be inflationary before yielding productivity enhancements.
Mohamed El-Erian, Allianz Chief Economic Adviser, supports the view that massive AI investment demands will strain financial markets. He asserts that the bond market cannot fund the extensive capital expenditure needed by tech platforms, governments, and other corporations without higher yields. For example, Amazon has guided to roughly $200 billion in capital expenditures across 2026 for AI infrastructure. El-Erian points to Amazon's recent lackluster bond issuance as a potential sign of market strain, suggesting that investors had to sell other assets to buy Amazon's new bonds. Microsoft's Q3 FY2026 capital expenditure hit $30.88 billion, an 84.39% year-over-year increase, with commercial remaining performance obligations reaching $627 billion, up 99% year-over-year. NVIDIA also posted Q1 FY2027 revenue of $81.62 billion, up 85.2% year-over-year, with data center revenue of $75.25 billion and $119 billion in supply-related commitments.
The Federal Reserve has also noted the concentration of U.S. investment in AI infrastructure, with real business fixed investment rising 5.5% last year and 11% in the first quarter, predominantly driven by AI-related build-outs. US tech investment as a share of GDP is nearing its 2000 dot-com bubble peak, reaching 11.33% in Q1, just 0.16 percentage points below the 11.49% peak. This raises concerns about potential overcapacity if AI demand doesn't keep pace with investment, similar to historical infrastructure booms that led to idle capacity.
Despite the potential for short-term inflationary pressure and market strain, the minutes did not offer concrete forward guidance on interest rate policy. Policymakers are assessing various scenarios and plan to await clearer data signals before committing to a specific rate path. This reflects Chair Warsh's focus on evidence-based decisions and avoiding rigid commitments, departing from previous communication strategies that might have hinted at an easing bias.
