AI Spending Is Carrying US Growth. What Would a Pullback Mean?

On Thursday, the Commerce Department said the US economy grew at a 3.4% annual rate in the third quarter, up from 2.2% in the second. Investment, consumer spending and government spending all rose, and within business equipment the biggest gain came from information processing equipment, mainly computers, the category that holds the servers going into AI data centers.

Economists had flagged the pattern before the report. “AI and data centers are booming, the rest of the economy is not,” Brian Wesbury and Robert Stein, the chief and deputy chief economists at First Trust, wrote on Monday 26 October.

What the report showed

The 3.4% pace beat the 3.2% that economists had forecast, investingLive reported. It covers the quarter in which Alphabet’s and Amazon’s spending on property and equipment more than doubled from a year earlier, their results releases showed.

The report also showed the catch. Imports rose, and the biggest contributor was capital goods, the category that includes imported computers and chips. Imports are subtracted from GDP, so part of every dollar spent on AI equipment never shows up as US growth.

How much of the growth is AI

Working from construction data, First Trust’s economists estimated that nominal spending on building data centers rose at a 139% annual rate in the third quarter, even if September’s spending stayed at August’s level. “That is NOT a typo,” they wrote. An annual rate compounds one quarter’s change over a year; by our arithmetic, it means spending rose by about a quarter from the spring.

How much of overall growth that adds up to depends on how you count. James Knightley, chief international economist for the US at ING, estimated in August that technology investment accounts for about a third of US growth this year, after subtracting imported computers and chips, a finding the Wall Street Journal also reported.

The share is big because the spending is growing fast, not because it is large. AI-related investment, from data-center buildings to the computers and networking inside them, came to about 0.8% of GDP in the first quarter, the research group Epoch AI estimated.

What a pullback would mean

GDP growth counts the change in spending, not its level. AI investment adds to growth only while it keeps rising, so if budgets merely flattened at this year’s record, its contribution would fall toward zero, and the economy would lose the slice ING puts at about a third.

Even the most bullish forecasters expect the rise to slow. Goldman Sachs, whose forecasts sit above the Wall Street consensus, expects AI spending by the five biggest US spenders to grow 54% in 2027 and 12% in 2028, PYMNTS and Briefs reported.

The rest of the economy has less to fall back on. Employers added just 29,000 jobs in September, against 90,000 expected, the Labor Department said.

ING also found that business investment outside technology fell from a year earlier for six straight quarters through early 2026. “Excluding that, the economy is tepid,” the First Trust economists wrote of the AI boom.

A falling stock market would be a second hit. ING estimated that the extra stock-market wealth since ChatGPT’s launch adds consumer spending worth about 0.44% of GDP, mostly from richer households, and the Nasdaq Composite just had its worst month since March 2025.

What it means for the Fed

The Federal Reserve has noticed the strength. It held its key rate at 3.75% to 4.00% on Wednesday, but Governor Lisa Cook and Cleveland Fed President Beth Hammack voted for a hike, and the statement upgraded its description of the economy to “expanding at a strong pace,” from “solid” in September.

Traders put the odds of a December hike at about 85%, according to Bloomberg data cited by Yahoo Finance. The PCE price index, the Fed’s preferred inflation gauge, rose 3.6% in the year to September, the Commerce Department said.

Rates feed back into the build-out. Heavy borrowing by AI companies has added to the competition for capital and helped push Treasury yields up, NBC News reported on Friday. Higher borrowing costs could make debt-funded AI spending harder to justify, Reuters noted the same day; the 10-year yield closed at 4.93%, by our market data, after topping 5% in September for the first time since 2007.

So a pullback would cut both ways, in our view. Growth would slow, with little else in the economy growing fast enough to fill the gap, but some of the pressure on rates and yields would ease with it.

What we’ll learn next

• The October jobs report: Friday 6 November, 8:30 a.m. ET.
• The second estimate of third-quarter GDP: Wednesday 25 November, 8:30 a.m. ET, with the first look at corporate profits.
• The Fed’s next decision: 8 and 9 December.

We think the economy is more exposed to AI spending than its 0.8% share of GDP suggests, because growth, not size, is what’s at stake. A plateau would do the damage on its own; nobody has to cut.