AI’s economic boost isn’t showing up in the US GDP, Goldman says that’s a $115 billion blind spot
Summary
Goldman Sachs analysts estimate that AI has lifted US economic activity by about $160 billion since 2022, roughly 0.7% of GDP. Yet official statistics only record around $45 billion of that gain — leaving an estimated $115 billion uncounted. The discrepancy stems from how the US Bureau of Economic Analysis treats high-performance semiconductors (AI chips) as intermediate inputs and subtracts imports, so much AI-related investment and intangible value creation isn’t fully captured in GDP figures. Goldman also flags roughly $75 billion in cloud-based model and enterprise AI development not being recorded as investment, while recent import frontloading ahead of tariffs further muddies investment data.
Key Points
- Goldman estimates AI added about $160bn to US economic activity since 2022, but only ~$45bn shows in GDP.
- Approximately $115bn of AI-driven growth is potentially uncounted due to GDP measurement methods.
- The BEA treats semiconductors as intermediate inputs, so chips used to build AI systems often aren’t captured as final investment.
- About $75bn spent on developing AI models and cloud enterprise solutions may be missing from investment statistics.
- Import frontloading ahead of tariffs and the subtraction of imports from GDP complicate the picture and can mask genuine AI investment trends.
- Companies frequently mention AI but rarely quantify its current impact on earnings, making corporate reporting an imperfect signal of macro effects.
Context and relevance
This matters because GDP is a central gauge for policymakers, investors and economists. If a sizeable chunk of AI-driven capital formation and intangible value is undercounted, decisions on interest rates, investment incentives and trade policy may be based on an incomplete picture. The finding also highlights a broader measurement challenge: rapid technological change — especially when it relies on specialised imported hardware and creates intangible assets — can outpace statistical methods designed for older production models.
Why should I read this?
Because if you care about where the economy actually is (rather than what the headline numbers say), this is a neat wake-up call. Goldman reckons there’s a $115bn slice of AI value hiding from GDP — that could shift how you think about growth, investment and policy. It’s short, sharp and changes the way you read the official data.
Author’s take
Punchy and important: this isn’t just an academic quibble. The way statisticians classify chips and cloud work has real consequences for how the AI boom appears (or disappears) in the numbers. If you work in finance, policy or tech strategy, read the detail — it could change forecasts and planning assumptions.