1. Warsh AI Remarks Put Three Economic Paths on the Table
Federal Reserve Chair Kevin Warsh said on Monday, November 16, that central bank staff have started modeling how artificial intelligence could reshape the US economy. The Warsh AI remarks, delivered at the Fed's "Productivity and Jobs" conference in San Francisco, described three scenarios, including one where unemployment climbs while growth accelerates and another where an AI investment bust drags down stocks.
The exercise does not amount to a prediction. Warsh presented the three paths as illustrations, built by staff economists using simplified models of the economy to test how different outcomes might play out over the next several years. The point, as he framed it, is preparation: the central bank wants to understand a range of possibilities rather than commit to a single view of where the technology leads.
Scenario analysis of this kind is a routine tool for policymakers facing an uncertain shock. Rather than betting on one projection, economists build several internally consistent stories, trace how growth, employment and financial conditions would behave in each, and ask what the appropriate response would be. Simplified models leave out much of the detail of the real economy, which is why Warsh was careful to describe the results as illustrative.
Even so, the disclosure is notable for what it says about the institution's thinking. The body responsible for US interest rates is now formally working through cases in which the most heavily funded technology of the decade either leaves workers behind or fails to deliver the returns that financial markets have already priced in.
2. What Are the Fed's Three AI Scenarios?
The three scenarios are a productivity boom whose gains reach workers, a productivity boom that leaves workers behind, and an AI investment bust that weighs on equity prices and economic growth. Two of them assume the technology works broadly as its backers hope. The third assumes it does not, at least not on the timetable or the scale that investors expect.
What separates the first two is not the size of the economy but who benefits from its expansion. In both, companies become more efficient and total output rises faster than before. The difference lies in the labor market, and specifically in whether people whose tasks are automated can move into new work as quickly as their old roles disappear.
The third scenario turns on finance rather than technology alone. It asks what happens when large sums have been committed on the expectation of very large profits, and those profits arrive late, arrive smaller, or do not arrive.
Laying the cases side by side also clarifies what the Fed would be watching for. Evidence of faster output per worker would point toward one of the first two paths, and the behavior of the unemployment rate would help distinguish between them. Signs that spending on AI infrastructure is outrunning the revenue it generates would point toward the third.
3. The Optimistic Case: Faster Growth That Workers Share
Under the most favorable outcome, AI allows businesses to produce more goods and services with the same resources. Output rises above its historical growth rate, which is the textbook definition of a productivity gain: more comes out of the economy without a matching increase in what goes in.
Crucially, workers take part in those gains in this version of events. Unemployment stays close to the level economists associate with a healthy labor market, meaning that the people who want jobs are generally able to find them even as the nature of work changes.
For a central bank, this is the comfortable case. Stronger productivity lets an economy grow more quickly without the strains that usually accompany rapid expansion, and a stable jobs market means policymakers are not forced to choose between supporting employment and guarding against other risks.
4. The Jobless Boom: Higher Output, Fewer Jobs
A less comfortable possibility keeps the growth but removes the shared benefit. In the second scenario, AI still speeds up the economy, yet companies automate jobs faster than displaced workers can find new ones. Unemployment therefore rises even though productivity is improving.
This is the combination that makes the exercise unusual. Economic growth and employment normally move in the same direction: when the economy expands, firms hire, and when it contracts, they cut staff. A world where output accelerates while more people are out of work breaks that familiar pattern, and it would complicate the reading of almost every indicator the Fed watches.
It would also pose a direct problem for the central bank's mandate. The Fed is charged with pursuing maximum employment alongside stable prices. If headline growth looked strong while joblessness rose because of structural change rather than weak demand, the usual tools for supporting the labor market might be poorly matched to the cause. Warsh did not prescribe a response; he described the scenario as one of several the staff are testing.
5. The Bust: When AI Spending Fails to Pay Off
Disappointment, not disruption, defines the third scenario. Here, AI investments fall short of the returns Wall Street has been counting on. Companies spend heavily on infrastructure, share prices rise in anticipation of enormous profits, and those expectations eventually turn out to have been too optimistic.
The mechanism is a familiar one from earlier investment cycles. Valuations built on anticipated earnings are sensitive to any evidence that those earnings will be smaller or slower than assumed. When confidence shifts, share prices can adjust well before company results do, and the loss of wealth and financing capacity feeds through to spending and hiring.
Warsh said that in such a case equity valuations could fall and set off an economic slowdown. Unemployment would rise more sharply than in the other scenarios. The economy would eventually recover, he said, but it would do so without the productivity boost that had been promised.
That last detail is what makes this the bleakest of the three. In the jobless boom, the economy at least gains the added output that AI makes possible. In the bust, the country absorbs the cost of a correction and a weaker labor market and emerges with little lasting improvement in efficiency to show for the spending.
6. Why the Fed Now Says 'Super Intelligence'
Readers of the speech may have noticed the vocabulary. Warsh referred to the technology as "super intelligence," or SI, rather than artificial intelligence. That choice reflects a change in the central bank's official terminology made in October, which followed an executive order from the Trump administration.
The label should not be read as a scientific judgment. It is a naming convention, not a finding that current AI systems have surpassed human intelligence. The scenarios themselves concern the economic effects of AI adoption and investment, whatever the technology is called.
7. Barr Flags More Than $3 Trillion in Off-Balance-Sheet AI Commitments
Warsh was not the only Fed official to raise concerns on Monday. Governor Michael Barr warned that the boom in AI construction is increasingly being financed through complex arrangements, among them long-term computing leases, energy contracts and financial guarantees.
The Fed estimates that the maximum potential value of these off-balance-sheet obligations exceeds $3 trillion. The figure does not represent debt that exists today. It represents commitments that could come due, which is why the number matters most in the scenario where AI revenue disappoints.
Off-balance-sheet obligations are promises that do not appear as borrowing in a company's headline accounts, such as an agreement to lease computing capacity for many years or to buy power under a long contract. They are a normal part of corporate finance, but they can make it harder for outsiders to see how much a business, and its counterparties, have really staked on a single outcome.
Barr's concern is about where the losses would land. If revenue from AI falls short, the consequences could extend beyond the companies that are building data centers to the landlords, energy suppliers, lenders and guarantors on the other side of those contracts. That is the channel through which the third scenario could spread from a correction in technology shares into the wider financial system.
8. What the Scenarios Mean for Stocks and Crypto
Markets have the most at stake in the third path. Share prices in the technology sector already reflect expectations of large future profits from AI, and the bust scenario is essentially a description of what happens if those expectations are revised down. A fall in equity valuations is the first link in the chain Warsh described, with slower growth and higher unemployment following.
Such a slowdown could also put pressure on riskier assets, including cryptocurrencies. Digital assets tend to be treated by investors as part of the risk end of a portfolio, so an environment of falling valuations and weaker growth is one in which they could come under strain alongside technology stocks.
The first two scenarios carry different implications. A broadly shared productivity boom is the outcome that current valuations appear to assume. The jobless boom is harder to classify, because corporate efficiency would improve while household incomes came under pressure from rising unemployment. None of the three was presented as the most likely, and Warsh attached no probabilities to them.
9. What Comes Next for Fed Policy
No change in interest rates was proposed in Monday's speech. Warsh used the appearance to describe how staff are thinking about AI over a horizon of several years, not to signal a near-term move.
The next scheduled opportunity for a decision is the Fed's policy meeting on December 8-9. Until then, the three scenarios stand as a statement of what the central bank considers plausible enough to model: an economy lifted by AI with workers sharing the gains, an economy lifted by AI with more people out of work, and an economy left to recover from an investment boom that did not deliver.
