AI Dark Output: The Visible Cost of Invisible Output
SemiAnalysis Twitter · SemiAnalysis (@SemiAnalysis_) · 2026-05-29
SemiAnalysis analysts Malcolm Spittler and Dylan Patel argue that AI is generating a massive and growing volume of economic value — termed "Dark Output" — that GDP and national accounting systems are structurally unable to measure, potentially dwarfing previous productivity mismeasurement episodes like the 1990s computing boom.
Appears in
Extraction
Topics: ai-economicsgdp-measurementproductivity-statisticsmacroeconomicsai-labor-displacement
Claims
- AI produces economic value that GDP and national accounts cannot currently capture, which SemiAnalysis calls 'Dark Output.'
- Dark Output has two forms: substitution (AI replacing human-performed tasks) and new output (AI performing work that was previously too costly to do).
- SemiAnalysis has identified approximately $1.5 trillion in tasks that current AI could substantially augment or automate.
- The AI measurement problem is larger in magnitude than the well-documented failure to capture the 1990s computing revolution in productivity statistics.
- Failure to measure AI output will create pressure to interpret AI spending as a bubble, because results will be invisible in official data.
Key quotes
AI output will be real before it is measurable.
We are at risk of having an event on the scale of the Industrial Revolution where most of the new output is invisible even as businesses spend increasingly large amounts on AI services.
Like the dark energy that makes up our universe, Dark Output will likely only be visible in its effects on other elements of the economy and not through direct observation.