The Federal Reserve is weighing how artificial intelligence spending affects the economy. Chair Kevin Warsh has said AI could raise productivity over time. But he also warned the investment boom is putting pressure on resources right now.
That pressure is already showing up in several markets. Construction, semiconductors, electricity and skilled labor are all being pulled into AI projects. This can push prices higher in the short term.
Morgan Stanley estimates nearly $3 trillion in global AI infrastructure investment through 2028. The bank also points to a $1.5 trillion gap between what's needed and what's currently financed through normal channels. That gap is being filled through other means, including private credit markets.
A Wharton finance professor argues the Fed doesn't fully understand who is funding this boom. He says the financing system behind AI is complex and still developing. Regulators may not have a clear picture of where risk is sitting.
A Lesson From the 1990s
The article draws a comparison to the mid-1990s. At that time, unemployment fell below levels the Fed considered normal. Many inside the central bank wanted to raise rates.
Instead, then-Chair Alan Greenspan held off. He believed productivity growth had changed the economy's underlying limits. Unemployment kept falling while inflation stayed low.
The author asks what would have happened if the Fed had tightened anyway. That outcome can't be tested after the fact. But the point is that lost investment and innovation don't always show up in economic data the way inflation does.
The concern is that AI investment could shift elsewhere if U.S. policy makes it too expensive. Once data centers and expertise move to another country, they may not come back easily. That could have long-term effects on U.S. competitiveness.
Financial Risks Are Harder To See
The piece argues that Fed policy models focus heavily on inflation and jobs. Financial stability gets less weight in comparison. That's despite the Fed being originally created to prevent banking panics.
New research cited in the article suggests credit spreads carry information about financing risk that inflation data misses. This includes signals about how much it costs companies to raise money. The author says this data deserves more attention from policymakers.
Since 2008, the Fed has built deep expertise in banks, housing and mortgages. That knowledge remains useful. But the author says the AI financing system looks nothing like the mortgage market that caused the last crisis.
The 2008 crisis wasn't only about interest rates being set wrong. It was also about regulators not understanding how connected and leveraged the financial system had become. The author says a similar knowledge gap could exist today with AI financing.
Raising interest rates without first understanding this system carries its own risk. Higher rates could expose leverage regulators don't fully see. At the same time, they could raise the cost of investment that AI needs to deliver on its promised gains.
The author isn't calling for looser monetary policy. He says persistent inflation will still need a response from the Fed. He also says central bankers shouldn't be in the business of picking which AI projects get funded.
His main point is about preparation. He says the Fed should build better data and models around private financing structures now. Waiting until a problem shows up in the numbers may be too late to respond effectively.
