AI Capital Investment Surpasses Dot-Com Boom
The investment surge in artificial intelligence has reached a scale that permeates the entire economy.
Current business spending on information processing equipment, software, data centers, and related sectors now represents nearly 6 percent of GDP, as noted in a recent analysis by Renaissance Macro’s Neil Dutta. This figure has already outstripped the investments seen during the dot-com era. Dutta highlights that this growth isn’t just confined to traditional AI companies; industrial firms like Caterpillar, Vertiv, Eaton, Cummins, and GE Vernova are closely aligning themselves with semiconductor stocks. He mentions, “their order book has become the order book for AI capital investment.”
Trade data further illustrates the situation. In March, the total value of imported capital goods, excluding autos, hit $120.7 billion and accumulated to $350 billion by the first quarter. Sales in computers, accessories, communications equipment, and semiconductors reached $71.8 billion in March alone, totaling $207.5 billion since the year’s start.
This surge in imports suggests a robust demand, with international supplies being attracted in significant quantities. Still, it’s worth noting that these imports do not contribute to U.S. production figures; hence, they don’t factor into GDP calculations.
Dutta argues that the method of calculating GDP may be too limited. He points out that the true concern should revolve around whether these investments yield substantial returns. While import leakage is a factor, it’s not the full story. Imported AI equipment strengthens global production cycles that can, in turn, benefit U.S. corporate profits and stock values, which subsequently bolster household wealth and state tax income.
On the domestic front, orders are also rising sharply. New orders for U.S. non-defense capital goods, excluding aircraft—essentially core capital goods—amounted to $83 billion in March and $241.6 billion year-to-date. This category encompasses construction equipment, industrial machinery, power tools, communication equipment, computers, and other physical infrastructure machinery.
The federal reserve’s capacity utilization data aligns with this trend. In April, utilization rates in the manufacturing of computers and peripherals hit 83.9 percent, a level not frequently observed since the late 1990s, aside from a temporary spike post-financial crisis. Previously noted increases had problems with the denominator; output was declining while measured capacity was dropping faster. Currently, output is on the rise, capacity is expanding, and utilization continues to increase.
An analysis of price data reveals real pressures. Companies are facing significant costs for the equipment and components used to build AI. However, this has not yet translated into widespread consumer technology inflation. The latest CPI report indicates a mere 2.3 percent year-over-year increase for computers and peripherals, while the larger information technology product category actually declined by 6.3 percent.
At this point, the pressure seems to be more upstream; businesses are making capital investments, while households are holding back. This situation differs from the broader consumer price concerns that might typically provoke alarm from the Fed.
Should the Fed Engage with the AI Investment Surge?
This situation recalls the late 1990s. Philadelphia Fed President Anna Paulson recently remarked on how, during the IT investment boom, Federal Open Market Committee officials expected inflationary pressures to necessitate interest rate hikes. Yet, those inflationary spikes never materialized. The Greenspan Fed’s patience proved worthwhile, resulting in strong growth, declining unemployment, and low inflation rates.
Today, the context is different from that of the late 1990s. Inflation has picked up, and in the ensuing decades, much of our tech production capacity has shifted abroad, increasing reliance on imports. However, it is still wise to consider these parallels. An investment boom with an eye toward enhanced productivity can boost current demand while also expanding future supply.
On some fronts, we might even be in a stronger position than we were in the 1990s. The unemployment rate has remained low for an extended period—lower than in the peak years of the dot-com boom. Although the number of employed individuals has risen, applications for unemployment insurance are two-thirds lower than they were back then. This creates an incentive for companies to invest in productivity technology, which enables wages to increase without adding inflationary pressure.
Moreover, current trade policies aiming to encourage domestic production, in contrast to the late 1990s strategies that promoted offshoring, provide incentives for firms to expand their domestic production capabilities—especially if the Fed avoids raising credit costs that would finance this expansion.
It’s possible the market is misinterpreting the AI boom, too quickly inferring that it will lead to higher prices. The demand for capital goods has risen steadily, necessitating abilities to meet that demand without pressuring consumer goods.
The pivotal question for the Fed revolves around whether investments in AI will spill into broader inflation or primarily generate capital expenditure cycles that enhance future production capacity. Current evidence seems to suggest the latter narrative is more accurate.
Essentially, we’re witnessing a supply-side boom rather than one driven by consumer demand—a scenario the Fed has historically found manageable rather than oppressive.
