The IMF's chief economist warned this week that AI investment isn't just a productivity story. It's also a wealth effect that can keep inflation sticky and interest rates higher than founders want.
Pierre-Olivier Gourinchas didn't bury the lede. Speaking to Bloomberg in a report published June 26, the IMF's chief economist said the AI investment boom is producing "tremendous valuations" in US equity markets. Those valuations are making investors and consumers feel richer, more confident, and more willing to spend before the productivity gains from AI have clearly shown up in the data.
That's the uncomfortable part of this story. AI may eventually help companies produce more with fewer people, fewer wasted hours, and better software. But the spending is happening now. The data centers are being built now. The chips, power contracts, cooling systems, servers, and private funding rounds are already moving through the economy. Productivity is still the promise. Demand is the fact.
The IMF's April 2026 World Economic Outlook put global growth near 3.1 percent in its mild scenario for this year, while global inflation under that same scenario was projected at 4.4 percent. The Guardian's coverage of the IMF report tied that inflation pressure partly to energy shocks from the Middle East conflict, but the broader point holds: this isn't the clean disinflationary path rate-cut optimists were hoping for. Gourinchas is now adding AI to the list of reasons central bankers can't relax.
His comparison is the late 1990s internet bubble, and it's a useful one if you don't overwork it. Both periods pushed stock valuations and capital gains higher, which helped fuel spending. The difference, as Gourinchas noted, is that today's AI investment isn't mainly debt-financed, so a correction would be less likely to run straight through the banking system. Shareholders would take the first hit. Banks may not.
Still, founders shouldn't take too much comfort from that. A wealth-driven spending surge can keep inflation pressure alive even if it doesn't look like a 2008-style financial accident. Your company doesn't need a banking crisis to suffer. It only needs capital to stay expensive for longer than your model assumed.
The Fed is watching the same thing. The Wall Street Journal reported this week that New York Fed President John Williams said monetary policy is "well positioned" to restore inflation to 2 percent, while also naming robust demand for AI-related technology goods as one of the current inflation drivers. The Fed's preferred inflation gauge rose 4.1 percent year over year, according to the same report. That's not background noise. That's the number sitting between founders and cheaper money.
For early-stage companies, interest rates aren't an abstraction. They shape venture math, exit valuations, customer budgets, bridge rounds, down rounds, and how long a board will tolerate a burn rate. Crunchbase data cited in April showed global venture funding hit roughly $300 billion in the first quarter of 2026, with AI deals driving the surge. But the headline figure flatters the market. Big checks are going to OpenAI, Anthropic, xAI, Waymo, and a small group of companies that can plausibly absorb billions at once. Plenty of founders outside that circle are still pitching into a market that asks for AI upside, lower burn, and proof of revenue all at the same time.
Here's the thing: the AI boom is making some venture portfolios look better on paper while helping keep the cost of capital high for the companies those portfolios are supposed to fund. A partner can sit across from you with impressive marks from AI bets and still tell you the bar for deployment is higher than last year. Both can be true. The wealth effect benefits asset holders. The rate effect lands on asset builders.
The physical build-out is enormous. CNBC and the Financial Times have reported Wall Street estimates showing the largest technology companies planning hundreds of billions of dollars in AI-related capital spending in 2026. Axios also reported June 26 that Goldman Sachs expects roughly $7.6 trillion to be invested globally in AI infrastructure from 2026 to 2031, across compute, data centers, and power. That money doesn't float above the real economy. It pulls on electricity grids, land, construction crews, transformers, memory chips, and Nvidia GPUs. You can call that investment. You can also call it demand.
Gourinchas's warning is two-sided, and it should stay that way. If AI productivity gains arrive quickly, the pressure eases and today's higher rates start to look like a temporary tax on the build-out. If the gains disappoint, or if AI revenue curves bend down before the capex curves do, the correction could be ugly without ever becoming a banking panic. A lot of companies would still be left on the wrong side of expensive infrastructure and expensive capital.
Frankly, founders don't need another victory lap about the AI boom. They need to price the bill. If your plan only works with cheap money, faster rate cuts, and investors willing to underwrite distant productivity gains, revisit the plan now. The boom is real. So is the rate environment it helps sustain.
Also read: Washington handed Anthropic back its most powerful model, and the precedent it set is bigger than the lockdown • The AI trade's week of reckoning arrived and the market's message to founders is clear • Washington's Escalating Hardware War With China Is Forcing a Reckoning for America's AI Buildout
