AI By The Numbers
The Institute’s AI Information Terminal
The numbers defining the artificial intelligence era
A weekly updated view of AI’s scale, economics, capabilities, infrastructure and societal impact—built to separate what has happened from what is merely projected.
Scale & Adoption
How rapidly AI is entering ordinary life and workRegular workplace AI adoption
Share of U.S. employees using AI in their role a few times a week or more. Another 52% use it at least a few times per year; 15% use it daily.
Why it matters: This better tracks recurring economic use than a broad any-use measure—although it still does not prove productivity.
U.S. working-age adoption (2025 benchmark)
Share of U.S. adults ages 18–64 reporting any generative-AI use for work, outside work, or both in the 2025 survey comparison.
Why it matters: This shows exceptionally rapid diffusion—but it is not a global-population estimate or a measure of daily use.
U.S. AI diffusion (telemetry)
Microsoft telemetry-based estimate of U.S. population-level AI diffusion during the second half of 2025.
Why it matters: This uses a different methodology from the 53% survey benchmark; the two figures do not measure the same thing.
Annual value to U.S. consumers
Estimated annual consumer surplus from generative AI in the United States by early 2026—not company revenue, GDP, or measured productivity.
Why it matters: AI can create substantial user value even when little of that value appears in company revenue or conventional economic statistics.
Capital & Competition
Where money, capacity and national advantage are concentratingGlobal corporate AI investment
Corporate investment in AI during 2025, an increase of 130% from the preceding year.
Why it matters: The buildout has become a capital-allocation event large enough to shape markets, credit and industrial policy.
Global private AI investment
Private investment in AI during 2025, up 127.5% year over year.
Why it matters: Private capital is expanding the field while increasing the cost of any future slowdown or repricing.
U.S. versus China private AI investment
U.S. private AI investment was 23.1 times China’s reported total in 2025.
Why it matters: Capital leadership remains overwhelmingly American even as model-performance gaps narrow.
Capability & Control
What systems can do—and what remains visible to outsidersReal-world agent task success
Performance on Terminal-Bench, compared with approximately 20% one year earlier.
Why it matters: The leap signals a rapid transition from systems that answer questions to systems that execute multistep work.
Cybersecurity issue success
Share of evaluated cybersecurity issues solved by AI agents, compared with 15% in 2024.
Why it matters: Defensive capability is rising—but the same trajectory expands the offensive potential available to attackers.
Foundation-model transparency
Average Foundation Model Transparency Index score, down from 58 in the preceding edition.
Why it matters: Capability is advancing faster than public visibility into training, evaluation, risk and governance.
Infrastructure & Energy
The physical system beneath the intelligenceAI data-center power capacity
Estimated global power capacity devoted to AI data centers in the 2026 AI Index.
Why it matters: AI is becoming a power-system issue; this capacity is comparable to New York’s peak electricity demand.
Data-center electricity consumption
Global data-center electricity consumption in 2024—approximately 1.5% of worldwide electricity use.
Why it matters: AI demand is landing on grids already constrained by generation, transmission and equipment lead times.
Projected data-center electricity use
IEA base-case projection for global data-center electricity consumption by 2030—more than double the 2024 level.
Why it matters: If realized, data centers would consume slightly more electricity than Japan does today.
Terminal Standard
This is a living strategic-intelligence product, not a collection of disconnected statistics. The Institute will expand each signal into a permanent record so readers can see how the number—and its meaning—changes over time.
- Observed figures describe measured historical activity.
- Estimates depend on defined assumptions or modeled adoption.
- Benchmarks report performance in a specified evaluation environment, not universal real-world ability.
- Forecasts describe possible future conditions and remain visibly separate from historical results.