Kevin Warsh has warned that the Federal Reserve is largely unaware of who is financing the massive AI boom that could reshape the global economy. The debate over artificial intelligence and monetary policy is already under way, but the central bank may be missing a more immediate challenge.
Scale of the AI investment surge
According to Morgan Stanley, nearly $3 trillion of AI-related infrastructure will be invested worldwide by 2028. The firm estimates a $1.5 trillion external financing gap that is already straining construction, semiconductor supply, electricity and skilled labour. In the short term, these pressures can lift resource utilisation and prices, while over the longer term automation and organisational change are expected to boost potential output and lower unit costs.
Why financial understanding matters
The temptation is to treat every sign of pressure as an inflation problem that requires higher interest rates. Monetary policy, however, does more than curb demand; it also influences the investment and innovation that determine future supply. A 2018 paper in the Journal of Monetary Economics by Patrick Moran and Albert Queralto showed that when technology adoption is endogenous, policy changes affect firms' incentives to develop new technologies and thus future productivity.
Lessons from past cycles
In the mid-1990s, unemployment fell below what policymakers then considered the natural rate. While some in the Fed pushed for tightening, Chairman Alan Greenspan entertained the idea that faster productivity growth had raised the economy's speed limit and largely resisted further rate hikes. Unemployment continued to fall and inflation stayed subdued.
That episode raises a question for today: how much of the 1990s productivity boom would have been missed if the Fed had continued tightening until the economy matched its models? The answer is unknowable, but the risk is clear. Inflation caused by excessive accommodation eventually appears in the data, whereas lost productivity from stalled investment never does. With AI, the damage could be permanent because data centres, power capacity, human capital and financing expertise create cumulative advantages that are hard to relocate.
Financial architecture of the AI boom
Traditional monetary-policy models give financial variables little independent weight, focusing mainly on inflation, employment and the output gap. Recent work with Sergey Sarkisyan shows that credit spreads contain policy-relevant information about financing distortions and firms' cost of capital that inflation and the output gap miss.
The Federal Reserve's original mandate was to protect financial stability after recurrent banking panics, yet inflation and employment have come to dominate its models. Financial stability should sit alongside price stability and employment, and at times should take precedence. Leverage, funding fragility and severe distortions in capital allocation can cause lasting economic damage that modest deviations of inflation or employment cannot.
What the Fed should do next
Understanding how AI investment is financed, the growing role of private markets, complex links among borrowers and intermediaries, and where leverage and maturity risk reside, requires better data and new models. Since 2008 the Fed has built deep expertise in banks, housing and mortgages, but the next vulnerability is unlikely to resemble the last crisis. Private-market funding structures deserve comparable analytical depth.
The lesson from 2008 is that the central failure was not simply an incorrectly set policy rate, but a failure to appreciate the leverage, complexity and interconnectedness of a rapidly changing finance system until it became systemic. AI is not subprime mortgages, and predicting another financial crisis would be unwarranted, but the institutional lesson is clear: when financial innovation outpaces models, understanding where risk accumulates must become a central concern of the Fed.
Higher rates alone cannot substitute for a clear picture of the financing ecosystem. Reflexive tightening could expose hidden leverage while raising the cost of the productive investment needed for AI to deliver its gains, potentially eroding the United States' technological leadership.
In short, the Fed must match its tools to the problem: address persistent inflation where it appears, but also restore financial stability to its proper place in the policy framework to avoid blinding itself to the AI investment cycle that could shape the next decade of growth.