AI Statistics (2026): 40 Sourced Figures

This page collects figures on artificial intelligence: how many firms and people use it, how much money goes into it, what it costs and what it consumes. Every number links to the body that first published it. Most come from government agencies, the Stanford AI Index, Pew Research Center and company filings with the US Securities and Exchange Commission. For teams that want to put AI into daily operations, the StatWharf page on AI workflow automation platforms compares the tools. Other StatWharf data pages are listed in the research directory.

Method

The data window runs from December 2024 to September 2026. StatWharf used primary sources only. These are the US Census Bureau, the US Department of Energy, the Federal Reserve Bank of St. Louis, Pew Research Center, the Stanford Institute for Human-Centered AI (HAI), company filings and vendor pricing pages. Statistics aggregators and blogs that re-quote other sources were excluded. Each figure was checked against its source in October 2026. StatWharf dropped each figure that did not appear on its primary source. No value on this page is modeled, estimated or calculated by StatWharf.

The 2026 AI Index was released by Stanford HAI in April 2026. It compiles data from outside providers. Where the AI Index relies on a named survey, the table says so. The organizational adoption figures, for example, come from a McKinsey & Company survey that the AI Index reproduces. Two figures from the International Energy Agency and from OpenAI’s price list were left out because their pages did not open at the check date.

AI adoption by US businesses

The US Census Bureau asks firms about AI use in its Business Trends and Outlook Survey (BTOS). The survey runs every two weeks. It asks whether a firm used AI to produce goods or services in recent weeks. The global figures from the AI Index use a wider definition. They count any regular use of AI in at least one business function.

The gap between 19.8% and 88% comes from method. It is not a contradiction. The Census Bureau samples all US employer businesses, and most of them are small. The McKinsey survey counts respondents from organizations that answer a business survey, which skews toward larger firms. The Census data shows the size effect directly. Firms with at least 250 employees use AI at 37%, close to twice the national rate. The Census Bureau also reports that AI use hovered between 17% and 20% from December 2025 to May 2026. In the same period, between 20% and 23% of businesses expected to use AI within six months.

How people and workers use AI

Three sources measure individual use. Pew asks US adults about chatbots. The St. Louis Fed runs the Real-Time Population Survey, which asks about generative AI at work and at home. The AI Index compares adoption across countries. Each uses its own question, so their rates differ.

Pew reports that chatbot use rose from a third of adults in 2024 to about half in 2026. In 2024, Pew asked the question only of adults who knew at least a little about chatbots. The two years are therefore not a perfect match. The St. Louis Fed measure is higher because it counts any use of generative AI, not chatbots alone. By its count, adult use rose from 45% in August 2024 to 62% in May 2026. Its occupation data draws on nearly 14,000 workers surveyed in four waves.

The time-saving figure is self-reported. The FRED Blog authors say such figures are approximate. Even so, the share of hours saved rose from 1.6% to 2.2% between Q3 2024 and Q2 2026.

AI investment and the AI market

Private investment figures in the AI Index come from Quid. They cover AI and machine learning companies with more than US$1.5 million in funding since 2013. Nvidia’s results are included because its data center segment sells most of the chips used to train and run AI models.

The US total of US$285.9 billion is more than 23 times the US$12.4 billion invested in China in 2025. The AI Index warns that private investment likely understates China’s total AI spending, because government guidance funds also invest. Generative AI companies drew US$170.9 billion, nearly half of global private AI funding. Global private investment grew 127.5% in the year. Nvidia’s data center revenue for fiscal 2026, which ended in January 2026, rose 68%. Its Q2 fiscal 2027 data center revenue was up 117% from a year earlier.

The question “how many AI companies are there” has no single official answer. No statistics agency keeps a register of AI firms. The count of newly funded companies is the closest dated measure. It counts only companies that raised private money in that year.

Data centers and energy

The largest cloud companies report capital spending in their annual reports. The line covers servers, data centers and other property. It is not limited to AI, but each company names AI infrastructure as a main driver.

Each company spent much more in its latest year than in the year before. Alphabet’s spending rose from US$52,535 million in 2024 to US$91,447 million in 2025. Meta’s rose from US$37,256 million to US$69,691 million. Microsoft’s rose from US$64,551 million in fiscal 2025 to US$115,948 million in fiscal 2026.

The United States has more than ten times as many data centers as any other country. Germany is next with 529, according to the AI Index. The Department of Energy report puts US data center electricity use at 176 TWh in 2023, up from 58 TWh in 2014. The AI Index compares the 29.6 GW of AI data center capacity to the peak demand of New York state.

AI in education

Students are among the heaviest users of AI tools. The AI Index and Pew both measure use for schoolwork. Pew surveyed 1,458 US teens and their parents from September 25 to October 9, 2025.

The AI Index also reports that only half of middle and high schools have AI policies. Use of the tools is far ahead of the rules that govern them. In the Pew teen survey, information search was the most common use of chatbots at 57%, just ahead of schoolwork.

Prices, risks and public views

AI prices are set by vendors, and no public body publishes an average. The rows below are list prices from three vendor pages. They are examples, not averages. API prices are quoted per million tokens. A token is a short piece of text, often part of a word.

Google lists its Gemini 3.8 Flash prices as promotional. They rise to US$1.50 input and US$7.50 output per million tokens from January 1, 2027. Anthropic’s Pro plan costs US$17 per month with annual billing. On both price lists, output tokens cost more than input tokens. A buyer who estimates the cost of an AI project therefore needs to know how much text the model writes, not only how much it reads.

Documented AI incidents rose from 233 in 2024 to 362 in 2025. The AI Incident Database records cases where AI systems caused harm or nearly caused harm. It depends on public reports, so it counts known incidents only. Public confidence in regulation is low. The share of adults with little or no confidence in government regulation of AI was 62% in 2024, according to the same Pew report.

What the figures show

Three points stand out. First, AI use is wide but uneven. About half of US adults use chatbots, yet only 19.8% of US businesses report using AI to produce goods or services. Large firms and information businesses lead. Small firms and retailers trail.

Second, money is moving faster than use. Global private investment more than doubled to US$344.7 billion in 2025. Each of the four largest cloud companies spent from US$69,691 million to US$131,819 million on property and equipment in its latest fiscal year. The Department of Energy projects that data center electricity demand grows with it.

Third, measured gains at work are still small. Workers report saving 2.2% of work hours through AI assistance. That number has trended upward since Q3 2024. A business that builds a case for AI spending has these public figures to test vendor claims against.

Where to go next

For teams that want to move from occasional chatbot use to automated, repeatable AI tasks, see the AI workflow automation platforms comparison. It sets out how each platform meters usage and what each publishes on pricing. It does not name a single best platform. More data pages are in the StatWharf research directory.

The next update follows three releases: the next Census BTOS AI estimates, Nvidia’s third quarter fiscal 2027 results and the 2027 AI Index. Send corrections to any figure to the editorial address below.

Factual corrections are free: editorial@statwharf.com.