AI Statistics 2026: Adoption, Spending, Jobs and Trust

AI statistics for 2026 show use racing ahead of payoff. 61.8% of US working-age adults used generative AI in the second quarter of 2026, up from 54.5% in the third quarter of 2025 (Real-Time Population Survey via FRED, Aug 2026). Yet only 37% of companies in McKinsey's global survey say AI has added anything to their operating profit, about the same share as a year earlier (McKinsey, Aug 2026).

The numbers below cover business and worker adoption, spending, models and computing power, electricity, jobs, trust and regulation. Each one links to the organisation that collected the data: statistical agencies, central banks, research labs, company filings and survey firms. Survey firms and vendors are named, so you can weigh them. Figures about AI assistants, chatbot traffic and AI search sit on our AI search statistics page, and the wider search picture is on the SEO statistics hub.

Key Takeaways

Worldwide spending on AI is forecast to reach $2.7 trillion in 2026, up 49.5% on 2025 (Gartner, Sep 2026).
About 56% of US workers used AI on the job for at least one of 11 tasks in March 2026 (U.S. Census Bureau, Aug 2026).
20.0% of EU businesses with 10 or more employees used AI in 2025, up from 13.5% in 2024 (Eurostat, Dec 2025).
OpenAI and Anthropic raised $217 billion in the first half of 2026, 43% of all startup funding worldwide (Crunchbase, Jul 2026).
Alphabet's free cash flow was minus $5.9 billion in Q2 2026, as it spent $44.9 billion on property and equipment in the quarter (Alphabet, Jul 2026).
Generative AI reached 53% population adoption within three years, faster than the personal computer or the internet (Stanford HAI, Apr 2026).
Training compute for frontier language models has grown 5 times a year since 2020 (Epoch AI, Feb 2026).
Data centre electricity use grew 17% in 2025, and use by AI-focused data centres grew 50% (IEA, Apr 2026).
US employers cited AI in 116,175 announced job cuts through August 2026, about 22% of all cuts (Challenger, Gray & Christmas, Sep 2026).
Employment of 22- to 25-year-olds in AI-exposed jobs is 19% below where it would be had it kept pace with less-exposed peers (Stanford Digital Economy Lab, Aug 2026).
Half of US adults say the rise of AI in daily life makes them more concerned than excited, and 10% are more excited (Pew Research Center, Oct 2025).
The AI Incident Database recorded 362 AI incidents in 2025, up from 233 in 2024 (Stanford HAI, AI Index 2026, Apr 2026).
US state legislatures passed 150 AI-related bills in 2025, up from fewer than 10 in 2020 (Stanford HAI, AI Index 2026, Apr 2026).

How many businesses use artificial intelligence?

Firm surveys agree on the direction and disagree on the level. Count firms and you get about one in five; count the people who work at firms using AI and you get most of the workforce, because big employers adopted first.

AI statistics: share of US firms and workers using AI by industry, Business Trends and Outlook Survey and Real-Time Population Survey, 2023 to 2025 — Figure 1: Professional services and finance pulled away from every other industry in 2025, while retail and manufacturing rose far more slowly, so "average" adoption hides a two-speed economy. Source: Federal Reserve Board, Monitoring AI Adoption in the US Economy, Apr 2026, using Census Bureau BTOS and Real-Time Population Survey data. Chart published by the source, captured 5 Oct 2026.

About 18% of US firms had adopted AI by the end of 2025, according to Census Bureau survey data analysed by the Federal Reserve (Federal Reserve Board, Apr 2026).

Before the Census Bureau revised its question in late 2025, the firm adoption rate grew by 68% in the year to September 2025 (Federal Reserve Board, Apr 2026).
Professional, scientific and technical services led at about 33% of firms, followed by finance at about 30% (Federal Reserve Board, Apr 2026).
Adoption averaged 37% in the information sector and 24% in real estate, against 13% in wholesale trade and 8% in accommodation and food services (Federal Reserve Board, Apr 2026).
Over 20% of firms expected to use AI in the first half of 2026 (Federal Reserve Board, Apr 2026).
Weighted by employment, about 78% of the US labour force works at firms that use AI, and about 54% at firms that use large language models, in the Survey of Business Uncertainty (Federal Reserve Board, Apr 2026).

The gap between the firm count and the employment-weighted figure is mostly arithmetic. About 57% of US firms have fewer than five employees, so a firm-weighted survey is dominated by very small businesses (Federal Reserve Board, Apr 2026). Small-business adoption gets its own section on our local SEO statistics page.

How does AI adoption compare across countries?

Roughly one firm in five uses AI across Europe, the OECD and Canada, with the Nordic countries far ahead and Eastern Europe far behind.

Denmark (42.0%), Finland (37.8%) and Sweden (35.0%) had the highest shares of AI-using enterprises in the EU in 2025, and Romania (5.2%), Poland (8.4%) and Bulgaria (8.5%) the lowest (Eurostat, Dec 2025).
Analysing written language was the most common business use of AI in the EU, at 11.8% of enterprises (Eurostat, Dec 2025).
Across OECD countries with data, 20.2% of firms used AI in 2025, up from 14.2% in 2024 and 8.7% in 2023 (OECD, Jan 2026).
52.0% of large OECD firms use AI, against 17.4% of small firms (OECD, Jan 2026).
57.3% of ICT firms across the OECD used AI in 2025, rising to 87.9% in Sweden (OECD, Jan 2026).
19.2% of Canadian businesses used AI to produce goods or deliver services in the 12 months to Q2 2026, triple the 6.1% of Q2 2024 (Statistics Canada, Jun 2026).
25.2% of Canadian businesses plan to use AI over the next 12 months, up from 14.5% a year earlier (Statistics Canada, Oct 2026).
26% of UK businesses used at least one type of AI in March 2026, up 8 points in a year, rising to 45% of businesses with 250 or more employees (Office for National Statistics, Apr 2026).

How far have large companies gone with AI?

Most large companies use AI somewhere, but fewer than half have scaled it across the business. 44% of McKinsey's respondents say AI is scaling across their enterprise, up from 38% a year earlier (McKinsey, Aug 2026).

Nearly nine in ten respondents report regular AI use in at least one business function (McKinsey, Aug 2026).
54% of respondents at companies with $1 billion or more in revenue report scaling AI across the enterprise, against one third at smaller companies (McKinsey, Aug 2026).
The share of large companies scaling AI agents rose from 27% to 40% in a year, while smaller companies stayed at 22% (McKinsey, Aug 2026).
32% of respondents say their company decided not to buy at least one software product because it could build the feature with AI coding tools (McKinsey, Aug 2026).
About 20% say AI operating costs, including token costs, have limited how much their company uses AI (McKinsey, Aug 2026).
AI high performers, which credit AI with at least 5% of operating profit, stayed at about 6% of respondents (McKinsey, Aug 2026).

How many workers use generative AI on the job?

Workers adopted AI faster than their employers did. Surveys of people find far higher use than surveys of firms, which means much of the AI used at work arrives without a company programme behind it.

Share of US workers by hours saved using AI at work in the last week, March 2026 — Figure 2: Most workers who used AI last week saved two hours or less, and one in seven saved nothing or lost time, which is a modest return next to the forecasts. Source: U.S. Census Bureau, About a Third of Workers Who Used AI in the Last Week Said They Completed Tasks One to Two Hours Faster, Mar 2026 survey data. Chart published by the source, captured 5 Oct 2026.

Between four and five in ten US workers use generative AI for work, depending on the survey and the time window it asks about. The Real-Time Population Survey put the work adoption rate among employed adults at 45.2% in Q2 2026, up from 37.4% in Q3 2025 (Real-Time Population Survey via FRED, Aug 2026).

52% of US employees say they use AI in their role, 30% use it a few times a week or more and 15% use it daily (Gallup, Jul 2026).
47% of US employees say their organisation has brought in AI tools, up from 41% a quarter earlier (Gallup, Jul 2026).
21% of US workers say at least some of their work is done with AI, up from 16% in 2024 (Pew Research Center, Mar 2026).
About 24% of workers who use AI on the job used it every day in the week before the survey (U.S. Census Bureau, Aug 2026).
75% of AI-using workers with a bachelor's degree used it at least one day that week, against 59% of those with a high school diploma or less (U.S. Census Bureau, Aug 2026).
30% of male AI users used it every day that week, against 17% of female users (U.S. Census Bureau, Aug 2026).

Pew asks whether AI does some of your work, Gallup whether you use it in your role, and the Census Bureau about 11 specific tasks. Narrow questions get lower answers.

How much time does AI save at work?

An hour or two a week for most users, not a day. 31% of workers who used AI the week before the survey said it saved them one to two hours, and another 25% said less than an hour (U.S. Census Bureau, Aug 2026).

15% said AI saved three to four hours and another 15% more than four hours (U.S. Census Bureau, Aug 2026).
10% said AI saved no time and 4% said it made their work take longer (U.S. Census Bureau, Aug 2026).
Generative AI assisted 6.3% of all US work hours by Q2 2026, up from 4.1%, and saved 2.2% of work hours, up from 1.6% (St. Louis Fed, FRED Blog, Aug 2026).
The share of employed adults who used generative AI for work in the past week rose from 28.2% in Q3 2024 to 39.2% in Q2 2026 (St. Louis Fed, FRED Blog, Aug 2026).
The most common uses at work were searching for information (37%), writing communications or documentation (32%) and generating ideas (32%) (U.S. Census Bureau, Aug 2026).
77% of employees who use AI for coding or process automation say it improved their productivity, the highest rating of any use (Gallup, Jul 2026).
In a study of 5,179 customer support agents, an AI assistant raised issues resolved per hour by 14% on average and by 34% for novice workers (NBER, Nov 2023).

How common is AI use among workers outside the US?

Use in Canada and Europe looks much like the US, with the same wide gaps by age, education and industry.

41.6% of Canadian workers used at least one AI or automation technology in their job in the year to March 2026, and 35.9% used generative AI tools (Statistics Canada, Jul 2026).
Generative AI use at work ranged from 65.6% of Canadian workers in professional, scientific and technical services to 16.3% in accommodation and food services (Statistics Canada, Jul 2026).
Only 11.6% of Canadian generative AI users use it for most or nearly all of their tasks (Statistics Canada, Jul 2026).
33% of people aged 16 to 74 in the EU used generative AI tools in the three months before the 2025 survey (Eurostat, Mar 2026).
64% of Europeans aged 16 to 24 used generative AI, against 7% of those aged 65 to 74 (Eurostat, Mar 2026).
41.1% of employed people across the OECD use generative AI tools, against 12.5% of retired and other inactive people (OECD, Jan 2026).
57% of employees in a 47-country study say they hide their use of AI and present AI-generated work as their own (KPMG and University of Melbourne, Apr 2025).

How much money is going into AI?

AI spending is now large enough to show up in national accounts. Most of it pays for data centres, and the companies building them are funding the bill largely from their own cash flow.

Quarterly capital spending at Amazon, Google, Meta, Microsoft and Oracle, and the same spending as a share of US GDP, 2021 to 2025 — Figure 3: Capital spending at the five largest cloud providers more than doubled between mid-2024 and the end of 2025 and reached about 1.3% of US GDP on an annualised basis, a buildout without recent precedent. Source: Federal Reserve Board, Monitoring AI Adoption in the US Economy, Apr 2026, Figure 1 top row. Chart published by the source, cropped, captured 5 Oct 2026.

How much will the world spend on AI in 2026?

Most of it goes on infrastructure. Gartner puts AI infrastructure spending at $1.48 trillion in 2026, more than half of its total AI forecast (Gartner, Sep 2026).

Gartner expects $576.5 billion of spending on AI services and $461.6 billion on AI software in 2026 (Gartner, Sep 2026).
Spending on generative AI models is forecast at $28.3 billion in 2026, a small slice of the total (Gartner, Sep 2026).
Gartner forecasts total AI spending of $3.64 trillion in 2027 (Gartner, Sep 2026).
In January, Gartner's forecast for 2026 was $2.52 trillion, so the estimate rose through the year (Gartner, Jan 2026).
Global corporate AI investment reached $581.7 billion in 2025, up 130% (Stanford HAI, Apr 2026).
US private AI investment of $285.9 billion in 2025 was 23.1 times that of China, at $12.4 billion (Stanford HAI, Apr 2026).
Epoch AI estimates that total AI capital spending will approach $1 trillion a year in 2026, close to 1% of world output (Epoch AI, May 2026).

Spending forecasts from different firms can differ widely, because they draw the line around "AI spending" in different places. Compare one firm's numbers over time rather than mixing firms.

How much venture money goes to AI startups?

More than 70% of global startup capital in Q2 2026 went to AI-focused companies, up from just under 50% a year earlier (Crunchbase, Jul 2026).

Global venture funding reached $510 billion in the first half of 2026, more than the $440 billion invested in all of 2025 (Crunchbase, Jul 2026).
Anthropic raised $65 billion in Q2 2026, close to a third of all venture funding that quarter (Crunchbase, Jul 2026).
16 companies raised rounds of $1 billion or more in Q2 2026, for $108.6 billion in total (Crunchbase, Jul 2026).
Private AI investment worldwide reached $344.7 billion in 2025, up 127.5% on 2024 (Stanford HAI, Apr 2026).
The number of newly funded AI companies rose 71% in 2025, and generative AI took nearly half of all private AI funding (Stanford HAI, AI Index 2026, Apr 2026).

How much are Big Tech companies spending on data centres?

Capital spending at the largest platforms is growing faster than their revenue. Amazon spent $54.2 billion on property and equipment in Q2 2026, against $32.2 billion a year earlier (Amazon, Jul 2026).

Amazon's purchases of property and equipment reached $173.0 billion in the 12 months to June 2026 (Amazon, Jul 2026).
Alphabet's purchases of property and equipment reached $132.4 billion in the 12 months to June 2026 (Alphabet, Jul 2026).
Meta expects 2026 capital spending of $130 billion to $145 billion (Meta, Jul 2026).
Capital spending by the largest technology companies exceeded $400 billion in 2025 and is expected to rise by another 75% in 2026 (IEA, Apr 2026).
On the revenue side, Amazon's AI business and its chips business each passed an annual run rate of $25 billion (Amazon, Jul 2026).
Google Cloud revenue grew 82% in Q2 2026, and its backlog reached $514 billion (Alphabet, Jul 2026).
Google's model APIs process about 22 billion tokens a minute, up from 16 billion a quarter earlier (Alphabet, Jul 2026).
Microsoft's Azure revenue passed $100 billion in a fiscal year for the first time (Microsoft, Jul 2026).

How fast are AI models and the computers behind them improving?

Scale still drives progress. Computing power, chips and power supply all grow several times faster than almost anything else in the economy, while the labs at the front say less and less about how their models are built.

Training compute of frontier language models, 2020 to 2025, growing about 5 times a year — Figure 4: Each dot is a frontier language model, and the compute used to train them has risen more than a thousandfold in five years, which is why access to chips and power now decides who can compete. Source: Epoch AI, Trends in AI, Feb 2026, CC BY. Chart published by the source, captured 5 Oct 2026.

Who builds the most capable AI models?

Companies, mostly American and Chinese. Industry produced over 90% of notable AI models in 2025, and the US produced 59 notable models to China's 35 (Stanford HAI, AI Index 2026, Apr 2026).

As of March 2026, the top US model led the top Chinese model by just 2.7% (Stanford HAI, Apr 2026).
The top closed model led the top open-weight model by 3.3% in March 2026, up from 0.5% in August 2024 (Stanford HAI, AI Index 2026, Apr 2026).
Four companies sat within 25 Elo points of each other on the Arena leaderboard in March 2026 (Stanford HAI, AI Index 2026, Apr 2026).
The average score on the Foundation Model Transparency Index fell from 58 to 40, and the most capable models often disclose the least (Stanford HAI, Apr 2026).
The number of AI scholars moving to the US has fallen 89% since 2017, and 80% in the last year alone (Stanford HAI, Apr 2026).
Frontier models gained 30 percentage points in a single year on Humanity's Last Exam, a test built to be hard for AI (Stanford HAI, AI Index 2026, Apr 2026).

How fast is AI computing power growing?

The world's installed stock of AI chip computing power grows about 3.4 times a year (Epoch AI, Feb 2026).

Global AI computing capacity reached 17.1 million H100-equivalents, growing 3.3 times a year since 2022, with Nvidia supplying over 60% (Stanford HAI, AI Index 2026, Apr 2026).
The performance per dollar of AI chips has improved 49% a year since 2023 (Epoch AI, Feb 2026).
The cost of training frontier models is climbing about 3.5 times a year (Epoch AI, Feb 2026).
OpenAI ended 2025 with 1.9 gigawatts of data centre capacity, up from 0.6 GW in 2024 and 0.2 GW in 2023 (Epoch AI, May 2026).
OpenAI used about 10% to 15% of the world's operational AI computing power at the end of 2025 (Epoch AI, May 2026).
Google and Meta together own around one third of the world's AI computing power (Epoch AI, May 2026).

How much electricity do AI data centres use?

Still a small share of the world's electricity, growing fast. Data centres used about 485 TWh in 2025, and the IEA expects that to roughly double to 950 TWh by 2030, around 3% of global demand (IEA, Apr 2026).

The energy used per AI task has fallen by at least an order of magnitude a year in recent years (IEA, Apr 2026).
Orders for gas turbines jumped 70% in 2025, one sign of strain in the energy supply chain (IEA, Apr 2026).
In the IEA's base case, electricity use by accelerated servers, mainly driven by AI, grows 30% a year to 2030 (IEA, Apr 2025).
US data centres used about 4.4% of US electricity in 2023 and could use 6.7% to 12% by 2028 (Lawrence Berkeley National Laboratory, Jan 2025).
US data centre electricity use rose from 58 TWh in 2014 to 176 TWh in 2023 (Lawrence Berkeley National Laboratory, Jan 2025).
AI data centre power capacity reached 29.6 GW, about the peak demand of New York State (Stanford HAI, Apr 2026).
Training Grok 4 produced an estimated 72,816 tonnes of CO2 equivalent (Stanford HAI, Apr 2026).

Is AI taking jobs?

Not across the whole economy, at least not yet. The damage so far is narrow: fewer entry-level hires in exposed occupations, and a growing share of layoffs that employers say are because of AI.

Employment of US workers aged 22 to 25 and 26 to 30 by AI exposure quintile, normalised to November 2022, through June 2026 — Figure 5: For 22- to 25-year-olds, the most AI-exposed occupations (darkest lines) fell below their late-2022 level while the least exposed kept growing, and the gap is smaller for workers just a few years older. Source: Stanford Digital Economy Lab, Canaries in the Coal Mine? (revised August 2026), Aug 2026, Figure 3. Screenshot of the source PDF, cropped, captured 5 Oct 2026.

How many layoffs are blamed on AI?

AI was the most cited reason for announced US job cuts for five months in a row, from March to July 2026 (Challenger, Gray & Christmas, Sep 2026).

In August 2026, AI fell to the fourth most cited reason, with 3,462 cuts, its lowest monthly total since December 2025 (Challenger, Gray & Christmas, Sep 2026).
In July 2026, AI was behind 10,970 announced cuts, 33% of the month's total (Challenger, Gray & Christmas, Aug 2026).
Since 2023, when Challenger began tracking it as a reason, AI had been cited in 184,538 job cut announcements by the end of July 2026 (Challenger, Gray & Christmas, Aug 2026).
Total announced US job cuts through August 2026 were 529,914, down 41% on the same period of 2025 (Challenger, Gray & Christmas, Sep 2026).
5% of UK businesses using AI said it had reduced their headcount, rising to 7% of those with 10 or more employees (Office for National Statistics, Apr 2026).
39% of McKinsey's respondents expect AI to reduce their company's total headcount in the coming year, up from 32% (McKinsey, Aug 2026).

Layoff announcements show what employers say, not what AI did. Some companies name AI to explain cuts they would have made anyway.

Are young workers losing jobs to AI?

In the most exposed occupations, yes. Employment of 22- to 25-year-olds in the two most AI-exposed groups of jobs fell about 11% between November 2022 and June 2026, while the same age group in the least exposed jobs grew about 10% (Stanford Digital Economy Lab, Aug 2026).

The gap comes mainly from companies hiring fewer young workers, not from firing more of them (Stanford Digital Economy Lab, Aug 2026).
Employment of US software developers aged 22 to 25 has fallen nearly 20% since 2024, even as older developers' headcount grows (Stanford HAI, Apr 2026).
Yale's Budget Lab finds that measures of AI usage show no connection to changes in employment or unemployment so far (Yale Budget Lab, Sep 2026).
In Denmark, linking AI chatbot surveys to payroll records, economists found no effect on earnings or hours worked, ruling out effects larger than 2% two years after ChatGPT launched (NBER, Mar 2026).

Will AI create more jobs than it destroys?

Employers expect a net gain. Companies surveyed by the World Economic Forum expect 170 million new roles and 92 million displaced by 2030, a net increase of 78 million (World Economic Forum, Jan 2025).

77% of employers plan to upskill workers in response to AI, and 41% plan to reduce their workforce as AI automates tasks (World Economic Forum, Jan 2025).
About 1 in 10 job postings in advanced economies asks for at least one new skill, and the incidence is about half that in emerging markets (IMF, Jan 2026).
Job postings that ask for new skills offer 3% to 3.4% higher wages in the US and the UK (IMF, Jan 2026).
Only 33% of Americans expect AI to make their jobs better, against a global average of 40% (Stanford HAI, Apr 2026).

Can people trust AI, and who is writing the rules?

Trust is not keeping up with use. People rely on AI every day while doubting what it tells them, and lawmakers are filling the gap one state and one deadline at a time.

Share of adults in 25 countries who say the increased use of AI in daily life makes them more concerned than excited, 2025 — Figure 6: Concern outweighs excitement in 23 of the 25 countries surveyed, with only South Korea and Israel leaning the other way, and the US, Australia and Canada are among the most worried, which matters for any business putting AI in front of customers. Source: Pew Research Center, Concern and excitement about AI around the world, Oct 2025, Spring 2025 Global Attitudes Survey. Chart published by the source, captured 5 Oct 2026.

How do people feel about artificial intelligence?

Warily. Fewer than half of people trust it: 46% of people across 47 countries say they are willing to trust AI systems (KPMG and University of Melbourne, Apr 2025).

A median of 34% of adults across 25 countries are more concerned than excited about AI in daily life, and 16% are more excited than concerned (Pew Research Center, Oct 2025).
When Pew first asked Americans in 2021, 37% said they were more concerned than excited (Pew Research Center, Mar 2026).
59% of people in a global survey feel optimistic about AI's benefits, up from 52%, while 52% say AI makes them nervous (Stanford HAI, Apr 2026).
66% of people rely on AI output without checking its accuracy, and 56% say they have made mistakes in their work because of AI (KPMG and University of Melbourne, Apr 2025).

How often does AI get things wrong?

Often enough that every answer needs checking. When journalists from 22 public broadcasters reviewed more than 3,000 answers from ChatGPT, Copilot, Gemini and Perplexity, 45% had at least one significant issue (EBU, Oct 2025).

31% of the answers had serious sourcing problems, and 20% had major accuracy issues such as invented details (EBU, Oct 2025).
Gemini had significant issues in 76% of its answers, more than double the other assistants (EBU, Oct 2025).
On a new accuracy benchmark, hallucination rates across 26 leading models ranged from 22% to 94% (Stanford HAI, AI Index 2026, Apr 2026).
The best model read analogue clocks correctly 50.6% of the time, against 90.1% for humans (Stanford HAI, AI Index 2026, Apr 2026).
Media reports of AI incidents and hazards rose from 92 to 324 a month on average between 2022 and 2025 (OECD, Feb 2026).
The share of reported AI incidents linked to cyberattacks and fraud grew 2.7 times over three and a half years (OECD, Feb 2026).

If AI assistants get the news wrong, they get brands wrong too. What they say about a company depends on which sources they find and trust.

How is AI being regulated?

Mostly by the EU and by US states. The EU AI Act became applicable on 2 August 2026, with rules for high-risk uses such as hiring, education and border control pushed back to 2 December 2027 (European Commission, Aug 2026).

Bans on prohibited AI practices applied from 2 February 2025, and obligations for general-purpose AI models from 2 August 2025 (European Commission, Aug 2026).
High-risk rules for AI built into regulated products such as machinery and medical devices now apply from 2 August 2028, after the AI Omnibus entered into force on 27 July 2026 (European Commission, Aug 2026).
California enacted 20 AI-related bills in 2025, Texas 12 and New York 10 (Stanford HAI, AI Index 2026, Apr 2026).
The US passed 25 AI-related laws between 2016 and 2025, the most of any G20 country, followed by South Korea with 17 (Stanford HAI, AI Index 2026, Apr 2026).
70% of people say AI regulation is needed, but only 43% think current laws are adequate (KPMG and University of Melbourne, Apr 2025).
31% of Americans trust their government to regulate AI, the lowest of the countries surveyed (Stanford HAI, Apr 2026).

Methodology and Sources

We included a number only after tracing it to the organisation that produced it and opening that page to confirm the figure and its date. Figures quoted in roundups, including competitor statistics pages, were followed back to the original study or dropped. Most numbers come from statistical agencies, central banks, international bodies, company filings and academic research; survey firms (McKinsey, Gallup, KPMG, Pew) and data vendors (Crunchbase, Gartner, Challenger, Epoch AI) are named in every citation. Where two sources measure different things, such as firm-weighted and employment-weighted adoption, we say so rather than blend them. This page is checked and refreshed every month.

Sources

Glossary

Generative AI: AI that produces text, images, code or audio in response to a prompt.
LLM: large language model, the type of model behind most generative AI tools.
AI agent: an AI system that carries out multi-step tasks on its own, such as running code or filling in forms.
BTOS: the Census Bureau's Business Trends and Outlook Survey of US employer businesses.
RPS: the Real-Time Population Survey, a survey of US adults run by economists Alexander Bick, Adam Blandin and David Deming.
Firm-weighted vs employment-weighted: counting each company once, or weighting companies by how many people they employ.
EBIT: earnings before interest and taxes, a measure of operating profit.
Capex: capital expenditure, spending on long-lived assets such as data centres and chips.
H100-equivalent (H100e): a unit that expresses computing power as a number of Nvidia H100 chips.
Training compute: the total computing work used to train a model, measured in floating-point operations (FLOP).
TWh / GW: terawatt-hours measure electricity used over time; gigawatts measure power capacity at a moment.
Hallucination: an AI answer that states something false as fact.
General-purpose AI (GPAI): in the EU AI Act, models that can perform a wide range of tasks, such as large language models.
OECD: the Organisation for Economic Co-operation and Development, a group of 38 mostly high-income countries.

140 AI Statistics October 2026