Goldman Economists Say AI Added Basically Zero to U.S. GDP in 2025
America spent $450 billion on AI last year. Goldman Sachs economists say it contributed nearly nothing to U.S. economic growth — exposing a stark gap between corporate hype and hard numbers.
America spent $450 billion on artificial intelligence last year. Goldman Sachs economists say that investment contributed basically zero to U.S. economic growth — a stunning verdict on the technology sector's most expensive boom in history.
Jan Hatzius, Goldman Sachs chief economist, told the Atlantic Council in February that widespread media coverage overstated AI's economic impact. "We don't actually view AI investment as strongly growth positive," Hatzius said. "I think there's a lot of misreporting, actually, of the impact AI investment had on U.S. GDP growth in 2025, and it's much smaller than is often perceived."
The reason lies in the arithmetic of GDP itself. Imports are subtracted from domestic investment, and roughly 75 percent of AI data center costs come from imported components. "A lot of the AI investment that we're seeing in the U.S. adds to Taiwanese GDP, and it adds to Korean GDP but not really that much to U.S. GDP," Hatzius explained. American companies are writing enormous checks — and much of that money flows straight overseas.
That reality check contrasts sharply with earlier, more optimistic readings. Harvard economist Jason Furman posted in October that investment in information processing equipment and software accounted for 92 percent of GDP growth in the first half of 2025. The Federal Reserve Bank of St. Louis later estimated AI-related spending contributed 39 percent of economic growth through the first nine months of the year.
Goldman's analysis suggests both figures failed to account for the import offset. Independent economist Joseph Politano calculates AI contributed about 0.2 percentage points to the nation's 2.2 percent GDP growth in 2025. "AI is a special case, because so much is imported," Politano said. The headline numbers, it turns out, were telling a very different story than the underlying data.
The gap between corporate enthusiasm and quantifiable returns runs deep. Ronnie Walker, a senior U.S. economist at Goldman, found that while 70 percent of S&P 500 management teams discussed AI on quarterly earnings calls, only 1 percent quantified any actual impact on earnings. Executives talked about AI constantly — but almost none could point to the bottom line and show what it bought them.
The same disconnect holds for productivity. "We still do not find a meaningful relationship between productivity and AI adoption at the economy-wide level," Walker said in March. Goldman researchers identified gains in only two specific use cases — customer support and software development — where median productivity improvements reached 30 percent. Everywhere else, the needle has barely moved.
Adoption itself remains thin. U.S. Census data shows fewer than 20 percent of establishments currently use AI for business functions — meaning the technology has yet to reach the critical mass required to shift macroeconomic indicators.
Joseph Briggs, joint-lead of Goldman's Global Economics Investment Research, offered a frank explanation for why earlier estimates missed the mark so badly. "It was a very intuitive story," Briggs told The Washington Post. "That maybe prevented or limited the need to actually dig deeper into what was happening." A compelling narrative, it turns out, can be its own obstacle to rigorous analysis.
Goldman's findings align with its original 2023 forecast, which predicted measurable AI impact would not arrive until 2027. The firm's baseline scenario still projects 6 to 7 percent of U.S. workers — roughly 11 million jobs — may face displacement by AI over the long term.
Other economists share the skepticism. Austan Goolsbee, president of the Federal Reserve Bank of Chicago, noted that "much AI investment is buying imported goods, so we need to subtract that to get the overall GDP effect." The structural problem is not unique to Goldman's model — it is baked into how the U.S. economy actually works.
Whether 2026 delivers different results remains the central question. The top five technology companies project $700 billion in AI infrastructure spending this year, a substantial jump from last year's total.
Imports from Taiwan have surged 670 percent over three years, and nearly $42 billion in computer and parts imports entered the U.S. tariff-free in November 2025. The import problem has not been resolved — and until American manufacturers can supply what the AI build-out demands, the returns may keep flowing east while the bills stay home.