Artificial intelligence is transforming corporate America, yet the boom remains understated in government growth statistics, according to Goldman Sachs.
Analysts at Goldman pointed to the scale of the boom in a Saturday note: “Revenue at US companies providing AI infrastructure has risen by $400bn since 2022, which at first glance seems to suggest that AI has been a meaningful driver of economic growth recently.”
But official numbers tell a different story.
AI technology has lifted real US economic activity by about $160 billion since 2022, or 0.7% of GDP, the analysts calculated. Yet only around $45 billion, or 0.2% of GDP, of AI-spurred growth has been recorded in official statistics. That leaves roughly $115 billion uncounted, according to the analysts.
That gap highlights the difference between what companies report and what the government measures due to the Commerce Department’s Bureau of Economic Analysis method for calculating growth.
“The measured impact of AI on GDP is likely much smaller because the BEA’s methodology for estimating GDP treats semiconductors as intermediate inputs, which are only counted towards final demand when the products (e.g., consumer laptops) that they enable are sold,” wrote the Goldman analysts.
So, high-performance semiconductors — the chips powering AI training — are classified as intermediate inputs. When they’re imported, the value is deducted from GDP, and their use in building AI systems doesn’t appear as investment.
However, the chips developed in recent years are being used for training and supporting AI models — essentially “building an intangible asset of which the ultimate output value has not been fully capitalized or measured in GDP,” the analysts wrote.
Related stories
Business Insider tells the innovative stories you want to know
Business Insider tells the innovative stories you want to know
Goldman’s analysts estimated that around $75 billion spent on developing AI models and enterprise solutions in the cloud has not been counted in investment statistics.
New import policies complicated the picture further.
In the first half of 2025, business investment in information-processing equipment appeared to jump, largely because companies rushed to import servers and networking gear ahead of President Donald Trump’s import tariffs.
The trend “probably reflects one-time frontloading ahead of tariffs and thus exaggerates normal AI investment demand,” the analysts wrote. Because imports are subtracted from GDP, the investment boost was partly offset.
AI’s impact is hard to pin down in other important indicators. Companies, too, are struggling to show it in their bottom lines.
While a record share of S&P 500 firms mentioned AI on earnings calls in the second quarter of the year, “the share of companies quantifying the impact of AI on earnings today remains limited,” according to a separate report from Goldman Sachs earlier this month.