AI Is Reshaping the Economy. Measuring It Is Another Story.

Artificial intelligence is fundamentally transforming the US economy, yet economists and statisticians are struggling to capture its impact in official data. There is no standalone line item for AI in the U.S. national accounts, meaning that billions of dollars in AI-related activity are scattered across existing categories like software investment, semiconductor manufacturing, and research and development. As The New York Times reports, this measurement gap has profound implications for policymakers, investors, and workers trying to navigate the AI era.

The Measurement Problem

The core challenge is structural. AI does not appear as its own category in the U.S. national accounts. Instead, AI-related economic activity is dispersed: AI software appears inside “software investment,” AI research appears inside “R&D,” AI chips appear inside “semiconductor manufacturing,” and AI cloud services appear inside “computer and data services.” As the New Space Economy analysis notes, “AI is already counted where it appears through existing product and industry categories, but the AI portion usually cannot be separated from the non-AI portion.”

This means GDP can include significant AI-related production without ever showing an AI line item. The Bureau of Economic Analysis (BEA) is acutely aware of the problem. In June 2026, BEA Director Vipin Arora announced that the agency will release “groundbreaking experimental statistics on the size of the American AI economy” later this year, calling it “a big step towards improving AI measurement and broadening our understanding of its impacts.”

The Productivity Paradox

Perhaps the most striking finding is the disconnect between AI investment and measured productivity gains. Top AI companies have forecast spending $710 billion on data centers across North America in 2026 alone. Yet Goldman Sachs chief economist Jan Hatzius stated that AI contributed “close to 0%” to US GDP growth in 2025, with only about 20 basis points of contribution over the last three to four years.

This paradox echoes the “Solow Paradox” of the 1980s and 1990s, when economist Robert Solow famously observed, “You can see the computer age everywhere but in the productivity statistics.” It took years for measurement methodologies to catch up with the IT revolution, and AI may present an even greater challenge due to its embedded nature.

However, early evidence from the BEA tells a more nuanced story. A February 2026 working paper by economists Tina Highfill and Jon D. Samuels found that industries with high AI intensity experienced total factor productivity growth about 2 percentage points higher per year. A subsequent April 2026 analysis found that AI intensity was associated with lower output price growth and lower labor and materials cost contributions. These findings suggest AI may be delivering productivity gains that official statistics are not yet capturing cleanly.

The Labor Market Signal

The most concrete evidence of AI’s impact is showing up in early-career hiring. The Stanford Digital Economy Lab’s AI Economic Indicators project, a collaboration with ADP Research, found that workers aged 22 to 25 in the most AI-exposed occupations experienced a 16% relative decline in employment after the spread of generative AI. More experienced workers in the same occupations did not suffer the same decline, and employment held steady in entry-level jobs with low AI exposure.

As MIT Technology Review reported, “Artificial intelligence has not so far produced a clean story of mass unemployment. Aggregate employment in developed countries remains broadly stable… But a troubling change may be hiding beneath the surface: the quiet weakening of the first rung of the career ladder.” The Federal Reserve Bank of New York adds to the concern, reporting that recent college graduate unemployment rose to 5.6% in Q4 2025, with underemployment reaching 42.5%.

Why This Matters

The measurement challenge has profound implications. Without accurate data, policymakers cannot effectively respond to AI-driven economic changes. Fiscal and monetary policy decisions rely on accurate GDP, productivity, and employment data. The disconnect between massive AI investment ($710 billion in data center spending) and near-zero measured productivity gains raises questions about whether current AI valuations are sustainable.

As the NBER working paper “Making AI Count: The Next Measurement Frontier” argues, current statistical frameworks “would likely struggle to capture [AI’s] full economic impact.” The paper calls for more granular, task-based, and outcome-focused approaches to ensure economic statistics remain relevant in an increasingly AI-driven economy.

What’s Next

The BEA is actively working on experimental statistics for the American AI economy, with a potential AI satellite account on the horizon — similar to existing accounts for the digital economy and outdoor recreation. The Stanford Digital Economy Lab is expanding its AI Economic Indicators dashboards. And the White House’s “America’s AI Action Plan” has directed federal statistical agencies to study AI’s labor-market impact using already collected data.

But until better data, definitions, and methodologies are in place, policymakers, investors, and workers will be navigating the AI transformation with incomplete information. The stakes are high: getting the measurement right is a prerequisite for getting the policy response right. As Erik Brynjolfsson, director of the Stanford Digital Economy Lab, put it, the goal is to “turn scattered signals into a shared, evidence-based understanding of how AI is reshaping work, productivity, and prosperity.”