The U.S. economy has thus far proven resilient, even amid a dramatic weakening of the labor market. Higher tariffs and reduced immigration reduce output and raise prices, but the enormous scale of investment in artificial intelligence has more than offset these effects. Spending on this new technology is estimated to have contributed a full percentage point to U.S. GDP growth in the first half of this year, accounting for more than half of the 1.6% reported. The current excess demand for computer capacity has incentivized hyperscalers—firms such as Meta, Google, Amazon, Microsoft, and Oracle that provide large-scale infrastructure and services to users of AI—to ramp up capital investment by hundreds of billions of dollars.
Meanwhile, the effects of higher tariffs have been delayed, first by a surge in imports before the tariffs went into effect that forestalled price increases, and then by a 4-month pause in reciprocal tariffs. The current average tariff rate is estimated at around 15%, well above the 2.5 % that existed before the inauguration but significantly below the 25%+ announced on Liberation Day. The delayed impact of tariffs has prevented inflation from rising significantly so far this year: it remains around 3%, one percentage point above the Fed’s 2% target. Most inflation forecasts, however, show a significant increase due to higher tariffs for the rest of this year and into next.
Reduced immigration has stalled U.S. labor force growth, a restraint on economic growth. While GDP is usually viewed as the sum of spending components (C+I+G), it can also be measured on the supply side as labor force growth multiplied by productivity (growth in output per hour worked). The investment in AI will add to productivity growth, offsetting at least a portion of the negative effect on the U.S. economy.
It is already clear that job growth has slowed dramatically, even though the government shutdown prevented the Bureau of Labor Statistics (BLS) from releasing the September employment report on October 3. The BLS report for August already showed significantly slower employment growth: the annual revision showed much smaller job increases over the past year through March, while the 3-month average gain through August was only 29,000, down from 100,000 3 months earlier. Labor market data for September was released by ADP, a company whose employment survey is less comprehensive than the BLS’s. It showed a decline of 32,000 jobs, and the August figure was revised from +54,000 to -3,000. While firms are not hiring many people, they don’t appear to be firing them either, at least not yet. Jobless claims remain low, and other surveys continue to show small layoff rates.
Despite the significant weakening in the labor market, the economy is holding up. Estimates for Q3 GDP growth center close to 3%, almost double the pace of the first half of the year, reflecting the accelerating pace of AI investment.
While the explosion in AI spending has thus far outweighed the adverse effects on U.S. economic growth from higher tariffs and reduced immigration, it’s far from clear how long this can continue. It is essential to recognize that AI investment must grow at the same pace, or faster, to continue providing the boost to GDP growth it has delivered so far this year. If AI spending remains at its current level—which is already very high—it would not add to GDP growth at all.
The projections of the hyperscalers suggest that their AI spending growth will slow, suggesting that the maximum boost to GDP may already be behind us. Any reduction in AI spending would have serious consequences for the U.S. economy, as the adverse effects of reduced immigration and higher tariffs are just beginning to show. That said, spending by the energy industry is likely to accelerate, making up for at least some of any slowdown by the hyperscalers. Significantly more computational power will be required to unlock the productivity benefits of the massive AI investment already underway.
The damage from government shutdowns is only temporary. Shutdowns reduce economic activity but generally don’t last long, and the economy rebounds right after they’re over, so financial markets tend to ignore them. Federal workers who’ve been furloughed—the CBO estimates around 750,000 – are on unpaid leave and barred from working. Essential workers are required to work without pay, but all federal government workers get back pay when the shutdown ends. Estimates of the negative effect on GDP are about half a percentage point per month. The current shutdown could cause more permanent damage if President Trump follows through on his threat to fire instead of furlough many government employees. Still, so far, that number is not large enough to have a material effect.
Federal Reserve Policy and Independence
The Fed’s dual mandate from Congress is to promote “maximum employment and price stability”. Since the recent weakness in employment has been much more pronounced than the rise in inflation, the Fed responded by cutting rates by a quarter point at its September 17 meeting, its first rate cut this year. The Fed also released the policy rate forecasts of its Federal Open Market Committee members at that meeting, revealing an unusual divergence in opinion for the remaining two meetings this year. Seven projected no more rate cuts, two favored one more cut, and nine favored two more. Stephen Miran—recently appointed to fill a vacant seat by President Trump—projected 150bps of rate cuts before year-end, which is unrealistic.
As expected, the FOMC reduced its policy rate by another quarter-point at its subsequent meeting on October 29. Markets are pricing in a high probability of yet another small interest rate cut at its final meeting of the year on December 10, which doesn’t seem unreasonable. But market expectations for significant further monetary easing next year seem overly aggressive, unless a recession becomes likely. The current target federal funds rate range of 3.75-4% is close to a neutral rate, implying that monetary policy is neither stimulative nor restrictive. If AI-related spending continues to support solid economic growth and inflation rises due to higher tariffs and tepid labor force growth, the Fed is likely to stand pat.
The overt attack on Fed independence has lately attracted more attention than its plans for interest rate changes. No previous President has publicly attacked the Fed Chair and its policies anywhere near the extent the current Administration has. An independent central bank—immune to short-term pressures to stimulate the economy—is critical to economic stability. There is overwhelming empirical evidence that central banks controlled by politicians lead to very poor economic outcomes, particularly high and volatile inflation.
The President has tried to influence Fed policy not only through verbal attacks but also by seeking to put loyal people on the Federal Reserve Board. His appointment of Stephen Miran—the Chair of his Council of Economic Advisors—to fill a temporary vacancy on the Board makes him the first person to take a seat on the Fed Board while retaining a position at the White House (he’s on a temporary leave of absence). The President also attempted to remove Governor Lisa Cook from the Board, but the Supreme Court delayed a decision until it hears legal arguments in January.
Control of the Fed by the executive branch would not be easy to attain. Fed Chairs, its staff, and virtually all appointed Governors and regional Reserve Bank Presidents cherish its independence. The FOMC typically reaches a solid consensus to convey confidence in its decisions and maintain its credibility. This was reflected in the FOMC’s last two policy votes, which were 11-1 and 10-2, despite the disparity in rate views. Presidents do have considerable power, but putting the Fed under the control of the executive branch would be very dangerous for the U.S. economy, strongly resisted by Fed officials, and possibly opposed by the judicial and legislative branches as well.
Markets Have Momentum, but Signs of a Bubble are Emerging
The S&P has managed a solid double-digit gain this year, despite the sharp selloff in the spring following the Liberation Day announcement of much higher and more widely imposed tariffs than had been expected. Since this year’s stellar performance comes on top of gains of well over 20% in both 2023 and 2024, concerns about a bubble are understandable.
Emerging technologies generate significant hype as they move from inception to widespread adoption, driven by expectations of transformative effects. The surge in optimism drives significant investment as businesses and investors don’t want to be left behind, pushing stock prices of many companies well beyond sustainable valuations. We have seen it all before, as far back as the UK’s railway mania in the 1860s and more recently the internet boom in the late 1990s. When it becomes apparent that many companies’ stock prices reflect unrealistically optimistic profit expectations, sentiment can shift sharply negative, triggering a severe correction.
It would not be surprising to see the stock market continue to rally over the next few months, given the current strength in corporate profits, the increase in energy production needed to implement AI technology, and the Fed’s recent rate cuts. The risks of a bubble, however, are growing. Current stock prices, in aggregate, look very expensive by valuation metrics such as P/E ratios, while the market value of AI-related stocks swamps that of the dot-com companies when that bubble burst in early 2000. The intellectual property around AI—the ability of companies to protect their innovations from being copied by competitors—is shaky, and barriers to entry are low. Moreover, while the benefits of AI could well end up being life-changing, the potential returns on the massive investment that is being made is far from clear. Tech bubbles are ripe for correction when value creation is clear but value capture is not. Generative AI is very expensive to provide, but the price of using it is currently very low, and no one knows whether there will be enough demand to justify the surge in prices that will be necessary to cover the massive costs that are being incurred.
While bubbles are painful in the short term, they can be beneficial over time. The bubble period creates excess capacity, which eventually lowers prices for new equipment and infrastructure, allowing more firms to adopt the latest technology. The internet bubble is a good example: it was very painful when it burst, but the increased availability of the infrastructure that was built enabled the U.S. to become a leader in the technology industry. The UK eventually built what was the world’s most extensive rail network in the late 19th century.
