Federal Reserve lacks visibility into $3 trillion AI financing

The Federal Reserve is navigating the artificial intelligence investment boom with incomplete information about how trillions of dollars in capital are flowing through the financial system, according to analysis published in Fortune.

Joao Gomes, a finance professor at the Wharton School and visiting scholar at the Federal Reserve Bank of Philadelphia, argues that while policymakers debate whether AI-driven resource constraints justify higher interest rates, they are overlooking a more pressing concern: the opacity of AI's rapidly evolving financing ecosystem.

Morgan Stanley projects nearly $3 trillion in global AI infrastructure investment through 2028, with an estimated $1.5 trillion external financing gap. That capital is already flowing into construction, semiconductors, electricity infrastructure, and specialized labor markets.

The productivity versus inflation dilemma

Federal Reserve Chair Kevin Warsh has emphasized that AI could boost productivity and expand economic capacity, though the investment surge may strain resources before those benefits materialize. Gomes contends this framing risks triggering premature monetary tightening that could suppress the very innovation needed to deliver future productivity gains.

He points to the 1990s as a cautionary example. When unemployment fell below what models predicted was sustainable, pressure mounted inside the Fed to raise rates. Chairman Alan Greenspan resisted, entertaining the possibility that accelerating productivity growth had raised the economy's speed limit. Unemployment continued falling while inflation remained subdued.

The counterfactual question remains unanswerable: how much of the 1990s productivity boom would have been lost if the Fed had tightened aggressively? With AI, Gomes warns, the stakes may be higher. Data centers, power infrastructure, specialized talent, and financing expertise create cumulative advantages. If that investment shifts overseas, lowering rates years later may not bring it back.

Why it matters

The Federal Reserve's traditional focus on inflation and employment may be insufficient for an investment cycle driven by complex private-market financing structures. Unlike inflation, which eventually appears in data, financial vulnerabilities can remain hidden until they trigger crises—and foregone productivity from suppressed investment never shows up in statistics at all. If the Fed misjudges the nature of AI-related economic pressure, it risks either enabling hidden financial instability or permanently damaging U.S. technological competitiveness.

Financial stability as a blind spot

Gomes argues that financial stability has drifted to the periphery of Fed policymaking, despite being the institution's original mandate following recurrent banking panics. Standard monetary policy frameworks give financial variables limited independent weight, focusing instead on how credit conditions forecast inflation and employment.

In recent research with Sergey Sarkisyan, Gomes demonstrates that credit spreads contain policy-relevant information about financing distortions and capital costs that inflation and output gaps miss. He contends that leverage, funding fragility, and severe capital allocation distortions can inflict more lasting damage than modest deviations from inflation or employment targets.

The AI boom makes this especially consequential. The Fed needs deeper understanding of private market growth, increasingly complex borrower-intermediary relationships, and where leverage and ultimate exposures actually reside. It needs better data and models for how losses might cascade if expected revenues disappoint or expensive computing infrastructure becomes obsolete faster than anticipated.

The 2008 lesson

Gomes draws parallels to the 2008 financial crisis, when the central failure was not an incorrectly set interest rate but rather policymakers' inability to appreciate the leverage, complexity, and interconnectedness of mortgage finance until consequences became systemic. While AI is not subprime mortgages, the institutional lesson applies: when financial innovation outpaces models, understanding risk accumulation must be central to Fed operations.

Reflexive tightening could produce the worst outcome—exposing leverage the Fed does not fully understand while raising costs for productive investment necessary for AI to deliver expected gains. The result could combine financial vulnerability with lasting loss of U.S. technological leadership as investment and expertise develop elsewhere.

Gomes emphasizes this is not an argument for easy money. Persistent inflation requires monetary response, and central bankers should not pick which AI projects deserve funding. The task is matching instruments to problems and restoring financial stability to its proper place in the Fed's framework without unnecessarily impairing capital formation.

The analysis was first reported by Fortune in a commentary piece by Gomes.