AI boom creates analytical challenge for central banks as surging spending meets import offsets

As massive investments in artificial intelligence continue unabated, the contribution of these outlays to the US economy and their precise impact on productivity remain subjects of intense debate.

The AI boom is heavily reliant on huge imports of hardware that the United States does not manufacture domestically.

This dynamic creates an unusual disconnect between the sheer scale of investment activity and the way it is reflected in official statistical releases, particularly GDP.

In this context, not all capital expenditures are reflected equally in GDP metrics. The construction of AI infrastructure relies heavily on imported semiconductor chips, servers, and networking hardware.

Consequently, a significant portion of the investment surge is offset in government growth calculations.

Artificial intelligence does not possess its own distinct, comprehensive category within official statistical frameworks. This obliges economists to infer its contributions from baseline investment and trade classifications.

In a paper published this month, Federal Reserve economists Paul Soto, Mason Thieu, and Jeffrey Allen wrote:

“The lack of a dedicated category for AI-related investment in national accounts, combined with the high import content of the equipment driving infrastructure buildouts, presents a heightened challenge in precisely measuring how much recent GDP growth has been powered by AI-related capital spending.”

The US economy expanded at an annualized rate of 2.1% in the first quarter. Federal Reserve economists estimate that AI expenditures—comprising capital outlays in software, data centers, energy infrastructure, and computing hardware—contributed approximately 0.73 percentage points to GDP growth.

However, the estimated contribution of AI investments would have been substantially larger were it not for the counterbalancing drag generated by imported AI components.

Net imports of computers, peripheral equipment, and components alone subtracted 0.45 percentage points from headline growth.

US AI import shares expand papidly

According to research by Minneapolis Fed economist Michael Waugh, AI-related products accounted for 23% of total US imports in 2025, up from 15% in 2023.

Computer hardware constitutes roughly half of these incoming shipments.

The remainder consists of electrical equipment, networking hardware, and cooling systems essential to building and operating AI data centers.

The Trump administration largely exempted many of these AI-related inputs from broad-based tariffs.

The interplay between AI capital outlays and trade dynamics was particularly pronounced in late 2025. Corporate entities funneled enormous capital into AI hardware during the final three months of the year, but once imports were factored into the equation, the net boost to GDP was largely erased.

Federal Reserve economists noted that the net impact of AI investments “varied significantly across quarters, as sharp import spikes in certain periods produced a negative contribution from net exports of computers, peripherals, and parts that offset the bulk of gross capital expenditure.”

Net imports exert drag on headline growth

Federal Reserve economists estimate that AI infrastructure investments added just 0.14 percentage points to GDP, even though top-line expenditures in relevant categories contributed roughly 0.75 percentage points prior to accounting for trade flows.

Net imports of AI equipment reduced overall GDP growth by 0.61 percentage points over the period.

Economists project that the economy expanded at an annualized rate of 1.8% during the April-June quarter. The official report will be released today (July 30).

Goldman Sachs anticipates a rebound in consumer spending alongside continued strength in corporate capital outlays driven by AI equipment expenditures.

The AI expansion is simultaneously forcing economists to re-examine traditional methodologies for measuring the broader economy.

Because growth is increasingly anchored in imported hardware and intangible software assets, conventional indicators of broader economic activity are becoming harder to interpret.

The productivity question: Capital utilization over AI adoption

Artificial intelligence appears to be enhancing worker productivity across a range of industries, and the US economy as a whole has recorded a notable productivity gain over the past few years.

However, it is far from certain that the former is directly driving the latter. A controversial new analysis indicates that companies are extracting more hourly output per worker not through widespread AI adoption—at least for now—but through more efficient utilization of existing capital assets.

According to Ernie Tedeschi, chief economist at Stripe, AI advancements may be yielding meaningful micro-level gains in specific sectors, but they have not served as the primary catalyst behind one of the most significant macroeconomic trends of recent years.

After decades of sub-par gains, the recent acceleration in labor productivity has stood out as one of the most positive developments for the US economy over the past few years.

Tedeschi noted that output per hour worked increased by 2.5% over the past year, compared to an average annual growth rate of 1.6% over the preceding two decades.

While that difference might appear modest, if sustained over even a few years, it compounds into a powerful effect, driving per-worker income and output to substantially higher levels.

However, Tedeschi found that while labor productivity has risen, total factor productivity—which measures output per unit of combined labor and capital inputs—remained virtually unchanged.

Examining performance across sectors, he observed that industries with high AI adoption did exhibit higher productivity growth, but that trend actually predated the pandemic—well before advanced large language models (LLMs) entered widespread commercial deployment.

Instead, Tedeschi concluded that higher output has been driven by more intensive and efficient utilization of existing capital stock.

Tedeschi wrote:

“Consider factories already built running longer hours, fully amortized server racks and GPU clusters being driven harder, or existing hotel rooms achieving higher occupancy rates. Economists refer to this as ‘capital intensity’ or ‘utilization rate.’ Higher capital utilization represents real economic gains, but it is fundamentally distinct from micro-level technological efficiency.”

In statements to Axios, Tedeschi said the US is in a period of elevated productivity growth and that it is increasingly plausible AI plays a role, but added a note of caution:

“We must remain realistic about how AI fits into this process, because doing so helps us discern whether AI represents a temporary fluctuation along the growth trajectory or a more permanent, transformative shift.”

AI boom complicates central bank strategy

While central bankers operate under mandates to preserve price stability, foster full employment, and maintain financial stability, the AI boom introduces fresh complexity across all three domains.

A report by the Bank for International Settlements (BIS) indicates that AI is blurring the traditional signals central bankers rely on to formulate monetary policy.

According to the BIS, the technology is simultaneously influencing both the supply and demand sides of the economy, triggering structural and cyclical shifts at the same time.

In the United States and other primary hubs of AI innovation, the near-term investment surge has sparked sharp demand for semiconductors and data center infrastructure components.

Concurrently, the stock market rally has inflated paper wealth, further stoking consumer demand. However, concerns persist that a portion of this market valuation may be illusory, raising fears of an AI asset bubble that could eventually burst.

While AI presents a medium-term risk of large-scale labor displacement, empirical evidence that this process has begun remains inconclusive.

Conversely, a scenario in which AI advances unlock substantial productivity gains would represent a favorable supply shock, which would typically be expected to exert downward pressure on inflation.

BIS economists Iñaki Aldasoro, Leonardo Gambacorta, Enisse Kharroubi, and Matthias Rottner wrote:

“The vast uncertainty surrounding the ultimate impact of AI presents distinct challenges for monetary policy and financial stability. First, AI simultaneously impacts demand and supply from both cyclical and structural perspectives. Moreover, these effects vary widely across individual sectors, complicating any assessment of underlying economic trends. Elevated uncertainty heightens the risk of policy missteps.”

Central banks—including the Federal Reserve, which concludes its policy meeting today, as well as the Bank of England and the Bank of Japan, both scheduled to meet tomorrow—are forced to make real-time interest rate decisions based on limited clarity regarding the direction, magnitude, and timeframe of the AI boom’s macroeconomic impact.