AI Statistics 2026: Investment, Adoption, Jobs and Energy

AI statistics are easy to quote and easy to misread. A company survey, a payment platform and a government business survey can all report different adoption rates without measuring the same population. This collection explains what each number actually describes.

Reviewed October 8, 2026. The 89 entries below include historical observations, company-reported figures, research estimates and forecasts. The reporting period belongs to each entry; this is not a table of 89 current-month measurements. Sources are linked so you can check the original definitions before reusing a figure.

Three comparisons to handle carefully

• Adoption: Ramp’s paying customers, Census businesses and Stanford’s survey respondents are different populations. Do not average their percentages.
• Market size: venture investment, a funding-round valuation, semiconductor revenue and annual spending forecasts are different quantities. Adding them double-counts activity.
• Performance: a score normalized to a human baseline is not a raw success rate. Likewise, a task horizon at 50% success is not a promise of reliable unattended work.

For definitions behind the figures, use our plain-English AI glossary.

AI market size and investment

Follow the money: investment, valuations, and capital spending.

• Global corporate AI investment hit $581.7 billion in 2025, up 130% from the prior year. (Full year 2025.) Source: Stanford HAI AI Index 2026
• Global private AI investment reached $344.7 billion in 2025, an increase of 127.5% from 2024. (Full year 2025.) Source: Stanford HAI AI Index 2026
• Stanford reports 2025 private AI investment of $285.9 billion in the U.S. and $12.4 billion in China, a 23.1-fold difference. These private-investment estimates do not comprehensively capture government funding. Source: Stanford HAI AI Index 2026
• Gartner forecasts worldwide AI spending will total $2.59 trillion in 2026, a 47% increase year over year. (2026 forecast (published May 2026).) Source: Gartner (reported by CIO Dive)
• Venture funding to AI startups reached $211 billion globally in 2025, up 85% year over year from $114 billion in 2024 and accounting for roughly half of all venture capital deployed. (Full year 2025.) Source: Crunchbase News
• Foundational AI startups raised $178 billion across just 24 deals in the first quarter of 2026 — double the $88.9 billion they raised across 66 deals in all of 2025. (Q1 2026 (as of March 31, 2026).) Source: Crunchbase News
• Crunchbase reported that OpenAI’s funding round had reached $122 billion by the end of Q1 2026. This is a dated financing snapshot, not revenue or annual spending. Source: Crunchbase News
• Anthropic raised a $30 billion Series G in early 2026 at a $380 billion post-money valuation, bringing its total raised to nearly $64 billion since its 2021 founding. (Q1 2026.) Source: Crunchbase News
• Elon Musk's xAI secured $20 billion in Series E funding in early 2026, taking its total reported debt and equity funding to $42.7 billion. (Q1 2026.) Source: Crunchbase News
• Bain’s December 2025 outlook projected $4.8 trillion in global M&A deal value for the year, up 36%. Its year-to-date tally of AI-related deals was $478 billion. Neither figure measures AI startup revenue. Source: Bain & Company

AI models and capabilities

What the frontier actually looks like in benchmarks and releases.

• Industry produced over 90% of notable AI models in 2025, continuing the shift of frontier model development away from academia. (2025.) Source: Stanford HAI AI Index 2026
• The United States produced 59 notable AI models in 2025, compared with China's 35. (2025.) Source: Stanford HAI AI Index 2026
• Stanford’s normalized SWE-bench Verified comparison rose from about 60% of the human baseline in 2024 to near 100% in 2025. This is performance relative to that baseline, not a claim that models resolved nearly every benchmark task. Source: Stanford AI Index 2026, Technical Performance chapter (normalized comparison)
• Stanford reports a 30-percentage-point gain on Humanity’s Last Exam over 2025. A benchmark improvement is not a measure of performance on every expert task. Source: Stanford HAI AI Index 2026
• The top U.S. and Chinese models were separated by 2.7% in Arena Elo ratings in March 2026. That rating gap is not a percentage difference in intelligence. Source: Stanford HAI AI Index 2026
• The leading closed and open-weight models had an Arena Elo rating gap of 3.3% in March 2026, compared with 0.5% in August 2024. Source: Stanford HAI AI Index 2026
• In Stanford’s March 2026 Arena snapshot, six of the top ten models were closed, and Anthropic led with an Elo rating of 1,503. Source: Stanford HAI AI Index 2026
• Stanford reports OSWorld computer-use benchmark performance rising from roughly 12% to 66.3% over 2025. The result applies to that evaluation, not arbitrary desktop work. Source: Stanford HAI AI Index 2026
• Gemini Deep Think scored 35 points in the 2025 International Mathematical Olympiad evaluation, reaching the gold-medal threshold. Google’s 2024 result used a different system. Source: Stanford HAI AI Index 2026
• METR’s January 2026 Time Horizon 1.1 update estimated a 320-minute, 50%-success task horizon for Claude Opus 4.5, with a 170–729-minute confidence interval. This covers its evaluated software tasks. The post-2023 fitted doubling time was about 131 days. Source: METR (Time Horizon 1.1)
• Meta announced a 10-million-token context window for Llama 4 Scout on April 5, 2025. Context capacity does not guarantee reliable retrieval of every detail in a long input. Source: Meta AI (official announcement)
• Epoch AI’s June 2024 analysis described training compute for notable models as doubling roughly every six months since 2010. Its prose and table give different exact annual growth factors, so no exact factor is reproduced here. Source: Epoch AI

Enterprise AI adoption

How businesses are really using AI beyond the press releases.

• In the survey data summarized by Stanford’s 2026 AI Index, 88% of respondent organizations reported AI use in at least one business function in 2025. This is not a census of all businesses. Source: Stanford HAI AI Index 2026 (Economy chapter, citing McKinsey survey)
• The 2026 AI Index reports generative AI use in at least one business function at 70% of surveyed organizations, using 2025 survey data. Source: Stanford HAI AI Index 2026 (Economy chapter)
• In survey data summarized by the 2026 AI Index, AI-agent deployment remained in single digits across nearly all business functions. The denominator is surveyed organizations, not all firms worldwide. Source: Stanford HAI AI Index 2026 (Economy chapter)
• Studies summarized in the 2026 AI Index reported productivity gains including 26% in software development, 14–15% in customer support and 50% in marketing output. These are results from different tasks and study settings, not a universal worker uplift. Source: Stanford HAI AI Index 2026 (Economy chapter)
• For the collection period ending May 3, 2026, the U.S. Census Bureau estimated that 19.8% of businesses used AI in any business function. The survey broadened its question in November 2025; earlier figures about producing goods or services are not directly comparable. Source: US Census Bureau, Business Trends and Outlook Survey
• AI adoption climbs sharply with company size: 37% of US firms with 250 or more employees use AI, versus under 20% of firms with four or fewer employees. (Collection period ending May 3, 2026.) Source: US Census Bureau, Business Trends and Outlook Survey
• Ramp measured paid AI adoption at 50.4% among businesses on its platform in March 2026, versus 35% a year earlier. This is a customer transaction sample, not an estimate for every U.S. business. Source: Ramp AI Index
• Average enterprise spending on large language models among surveyed Global 2000 companies has risen from about $4.5 million to about $7 million over two years, and executives expect it to grow another ~65% to roughly $11.6 million in 2026. (Survey published 2026, covering current spend and 2026 expectations.) Source: Andreessen Horowitz (a16z) enterprise AI spend survey of 100 Global 2000 VPs and C-level executives
• Menlo Ventures estimated enterprise generative-AI spending at $37 billion in 2025, versus $11.5 billion in 2024, and roughly 6% of the global SaaS market. These are market estimates. Source: Menlo Ventures, 2025: The State of Generative AI in the Enterprise
• Menlo Ventures estimated coding-related departmental AI spending at $4.0 billion in 2025, or 55% of its estimated $7.3 billion departmental total. Source: Menlo Ventures, 2025: The State of Generative AI in the Enterprise

Consumer AI usage

The user numbers behind the biggest consumer apps in history.

• ChatGPT reached 900 million weekly active users in February 2026, up from the 800 million OpenAI reported in October 2025, announced alongside a $110 billion funding round. (February 27, 2026.) Source: OpenAI announcement, reported by TechCrunch
• OpenAI reported more than 50 million paying consumer subscribers in its February 27, 2026 announcement, according to TechCrunch. Paid subscriptions and weekly active users measure different things. Source: OpenAI announcement reported by TechCrunch
• Google's Gemini app hit 950 million monthly active users, with daily active users tripling in the last year, per Alphabet's Q2 2026 earnings. (Q2 2026 (reported July 2026).) Source: Alphabet Q2 2026 earnings (Sundar Pichai, official Google blog)
• In Pew’s February 17–23, 2026 survey of 5,119 U.S. adults, 49% said they had ever used AI chatbots. “Ever used” does not mean regular or daily use. Source: Pew Research Center, Americans and AI 2026
• Pew’s February 2026 survey found that 44% of U.S. adults had ever used ChatGPT, compared with 24% for Gemini, 17% for Copilot and 14% for Meta AI. Respondents could have used multiple services. Source: Pew Research Center, Americans and AI 2026
• Roughly two-thirds of U.S. teens ages 13-17 (64%) say they ever use AI chatbots, and about three-in-ten use them every day, according to Pew Research Center's survey of 1,458 teens. (September 25 - October 9, 2025 survey (published December 9, 2025).) Source: Pew Research Center, Teens, Social Media and AI Chatbots 2025
• The NBER working paper “How People Use ChatGPT” reports 700 million weekly users sending 18 billion messages a week by July 2025. Its authors include OpenAI researchers; this is not an independent audit of the service. Source: NBER Working Paper 34255, "How People Use ChatGPT" (Chatterji et al.)
• The same paper reports roughly 2.63 billion daily consumer-plan messages in June 2025, versus 451 million in June 2024. Its classifier estimated that non-work messages accounted for 73%, up from 53%. Source: NBER Working Paper 34255, "How People Use ChatGPT" (Chatterji et al.)
• The NBER paper’s first-name proxy classified 48% of active users’ names as typically masculine by June 2025; this is not verified gender. Among adult messages with self-reported age available, about 46% came from users aged 18–25. Source: NBER Working Paper 34255, "How People Use ChatGPT" (Chatterji et al.)
• Appfigures estimated 318 million ChatGPT app downloads through mid-August 2025, 2.8 times the comparable 2024 period, and $2 billion in lifetime mobile consumer spending. These third-party estimates are not counts of unique users. Source: Appfigures, reported by 9to5Mac
• ChatGPT was the No. 1 most-downloaded free iPhone app in the U.S. in 2025 on Apple's official year-end chart — its first year at the top — ahead of Threads, Google, TikTok, and WhatsApp, with Google Gemini ranking tenth. (Full-year 2025 (published December 10, 2025).) Source: Apple App Store year-end charts, reported by TechCrunch
• Anthropic said Claude's paid subscriptions more than doubled in early 2026, with third-party credit-card panel data from Indagari showing record U.S. consumer subscriber growth between January and February 2026. (September 2025 through early March 2026 (published March 28, 2026).) Source: Anthropic spokesperson and Indagari data, reported by TechCrunch

AI energy and infrastructure

The physical footprint: power, chips, and data centers.

• The IEA estimates that all data centers consumed 485 TWh of electricity in 2025, up 17%, and projects about 950 TWh in 2030. The totals include non-AI workloads; they cannot be labeled AI-only electricity use. Source: IEA, Key Questions on Energy and AI
• The IEA estimates that electricity use in AI-focused data centers grew 50% in 2025 and projects a tripling between 2025 and 2030. The latter is a forecast, not an observed result. Source: IEA, Key Questions on Energy and AI
• Gartner’s April 8, 2026 forecast put worldwide semiconductor revenue above $1.3 trillion for 2026, up 64%. This covers the semiconductor market, not just AI chips. Source: Gartner, April 8 forecast
• AI semiconductors are expected to account for around 30% of total worldwide semiconductor revenue in 2026, with hyperscaler investment in AI infrastructure expected to increase by more than 50% during the year. (2026 forecast (Gartner release reported April 13, 2026).) Source: Gartner forecast (reported by Evertiq)
• Nvidia's data center revenue hit a record $75.2 billion in its fiscal Q1 2027 (quarter ended April 26, 2026), up 92% from a year earlier, out of $81.6 billion in total quarterly revenue (up 85% year over year). (Quarter ended April 26, 2026 (reported May 2026).) Source: NVIDIA official earnings press release
• Stanford’s 2026 AI Index reports global AI data-center power capacity of 29.6 gigawatts at the end of 2025. Power capacity and annual electricity consumption are different measures. Source: Stanford HAI AI Index 2026
• Stanford’s 2026 AI Index counts 5,427 U.S. data centers in its 2025 data. This includes data centers generally, not only facilities dedicated to AI. Source: Stanford HAI AI Index 2026
• Stanford reports AI compute capacity of 17.1 million H100-equivalents in 2025, growing 3.3-fold annually since 2022, with Nvidia hardware supplying over 60%. H100-equivalents normalize computing capacity; they are not a count of physical H100 cards. Source: Stanford HAI AI Index 2026
• Data Center Knowledge, citing Census data, reported U.S. private data-center construction at a $50.7 billion seasonally adjusted annual rate in April 2026, versus $43.8 billion for office construction. The annualized rate is not the amount spent during April. Source: US Census Bureau Value of Construction Put in Place survey (via Data Center Knowledge)
• In August 2025, Omdia forecast $207 billion in 2025 data-center GPU and AI-accelerator shipments, against $123 billion in 2024, and $286 billion by 2030. The 2025 figure was still a forecast when published. Source: Omdia (via RCR Wireless News)
• The EIA estimates that data-center servers used 7% of U.S. commercial-sector electricity in 2025. Its scenarios put that share at 22–33% in 2050. These are commercial-sector shares, not shares of all U.S. electricity. Source: US Energy Information Administration, Annual Energy Outlook 2026

AI jobs and the workforce

What AI is doing to hiring, wages, and job cuts.

• PwC’s 2026 analysis of more than one billion job ads across 27 countries and territories found a 62% advertised wage premium associated with AI skills. This association does not show that taking an AI course causes a 62% raise. Source: PwC 2026 Global AI Jobs Barometer (press release via PR Newswire)
• In PwC’s job-ad dataset, postings requiring specific AI skills grew 69%, versus 9% for postings overall. Those rates describe the analyzed advertisements, not net jobs created. Source: PwC 2026 Global AI Jobs Barometer (press release via PR Newswire)
• Challenger recorded AI as the stated reason for 101,743 announced U.S. job cuts in January–June 2026, about 23% of the total. These are employer announcements, not independently established AI-caused job losses. Source: Challenger, Gray & Christmas
• In June 2026, employers cited AI in 14,029 announced U.S. job cuts, or 31% of that month’s total in Challenger’s report. Source: Challenger, Gray & Christmas
• Technology employers announced 139,156 U.S. job cuts in the first half of 2026, up 83% from 76,214 a year earlier, per Challenger. This sector total includes reasons other than AI. Source: Challenger, Gray & Christmas
• Lightcast’s data for Stanford’s 2026 AI Index put AI-skill mentions at 2.5% of U.S. job postings in 2025, up 55% year on year and 297% over a decade. Different job-ad trackers use different samples and definitions. Source: Stanford AI Index 2026 labor-market chapter, powered by Lightcast
• The August 2026 revision of Stanford’s “Canaries in the Coal Mine?” reports a 19% relative employment decline for workers aged 22–25 in the most AI-exposed occupations, compared with less-exposed workers. The analysis points to reduced hiring; it does not establish that AI caused 19% unemployment. Source: Stanford Digital Economy Lab (Brynjolfsson, Chandar & Chen, "Canaries in the Coal Mine?"), based on payroll data from ADP
• The share of US job postings mentioning AI reached a record high of 4.2% at the end of 2025, with AI-mentioning postings 134% above February 2020 levels while total postings were just 6% above that baseline. (December 2025.) Source: Indeed Hiring Lab (Indeed AI Tracker)
• Indeed’s July 8, 2026 analysis found software-development postings up almost 15% since late February 2025, while overall postings fell 7%. Senior roles accounted for 71% of the May 2025–May 2026 increase. Timing alone does not prove a coding tool caused this growth. Source: Indeed Hiring Lab
• Microsoft’s 2026 Work Trend Index cites LinkedIn’s January 2026 report of at least 1.3 million AI-related job opportunities over the previous two years. “Opportunities” is not a net-employment measure. Source: LinkedIn 2026 Labor Market Report, cited in Microsoft 2026 Work Trend Index
• The World Economic Forum’s 2025 employer survey projects 170 million jobs created and 92 million displaced by 2030, a net 78 million. The forecast covers multiple economic, demographic and technological trends, not AI alone. Source: World Economic Forum

AI regulation and safety

The rules, penalties, and documented failures.

• Article 99 of the EU AI Act as adopted in 2024 sets a prohibited-practices ceiling of €35 million or 7% of worldwide annual turnover. The higher general ceiling has exceptions: for SMEs, the lower applicable ceiling governs. This is not an automatic fine. Source: Official EU AI Act, Article 99 (2024 adopted text)
• The Act’s original Article 99 also sets ceilings of €15 million or 3% for specified other infringements and €7.5 million or 1% for misleading information. SME exceptions, applicable provisions and dates matter; these numbers alone do not determine a company’s liability. Source: Official EU AI Act, Article 99 (2024 adopted text)
• The AI Incident Database recorded 362 incidents in 2025, versus 233 in 2024, according to Stanford. The roughly 55% rise measures documented database entries, not the true incidence of every AI-related harm. Source: Stanford HAI AI Index Report 2026 (Responsible AI chapter)
• After rising from 37 to 58 between 2023 and 2024, the average score on the Foundation Model Transparency Index fell to 40 in 2025, signaling declining disclosure by leading AI developers around training data, compute, and post-deployment impact. (2025.) Source: Stanford HAI AI Index Report 2026 (Responsible AI chapter)
• The share of surveyed businesses with no responsible AI policies in place fell sharply from 24% to 11% in 2025, while AI-specific governance roles grew 17%. (2025.) Source: Stanford HAI AI Index Report 2026 (Responsible AI chapter)
• In 2025, US state lawmakers in all 50 states introduced 1,208 AI-related bills, of which 145 were enacted into law. (Full-year 2025 state legislative sessions.) Source: MultiState AI Legislation Tracker
• As of March 2026, state lawmakers in 45 US states had already introduced 1,561 AI-related bills in the 2026 legislative session, outpacing the 2025 total with the session still underway. (As of March 2026.) Source: MultiState AI Legislation Tracker
• Half of US adults (50%) say the increased use of AI in daily life makes them feel more concerned than excited, while just 10% are more excited than concerned — concern is up from 37% when Pew first asked in 2021. (June 2025 survey (published March 2026).) Source: Pew Research Center
• Only 44% of Americans have a lot or some trust in the US government to regulate AI well, while 47% have not too much trust or none at all. (March 2025 survey.) Source: Pew Research Center
• Among more than 48,000 respondents across 47 countries in the University of Melbourne/KPMG study, 46% were willing to trust AI and 66% reported regular use. Fieldwork ran from November 2024 to January 2025. Source: KPMG / University of Melbourne, "Trust, attitudes and use of AI: A global study 2025"
• In that same survey, 70% supported AI regulation, 43% considered existing rules adequate and 87% wanted stronger laws against AI-generated misinformation. These percentages describe respondents, not a census of the world. Source: KPMG / University of Melbourne, "Trust, attitudes and use of AI: A global study 2025"
• Entrust’s 2026 Identity Fraud Report found deepfakes in one in five biometric fraud attempts in its dataset and reported a 58% increase in deepfake selfies. Provider-observed fraud does not measure all fraud worldwide. Source: Entrust Identity Fraud Report 2026

AI content and search

How AI is reshaping the web and search itself.

• Ahrefs classified 74.2% of 900,000 newly detected English-language pages in April 2025 as containing some AI text, including 2.5% classified as entirely AI-written. Its detector-based labels are estimates, not verified authorship; one page was sampled per domain. Source: Ahrefs (study of 900,000 new web pages)
• Graphite’s updated study of 55,400 English-language Common Crawl article URLs classified 49.9% of Q1 2026 articles as primarily AI-generated, versus 50.9% in Q4 2025. It used three detectors; classification errors and sample coverage limit the result. Source: Graphite updated study
• In Semrush’s tracked sample of more than 10 million keywords, AI Overviews appeared for 6.49% in January 2025, 24.61% in July and 15.69% in November. These are keyword-sample shares, not the percentage of all Google searches. Source: Semrush AI Overviews Study
• Semrush also found the intent mix of AI Overviews shifting toward commercial searches: 91.3% of queries triggering an AI Overview were informational in January 2025, but that share fell to 57.1% by October 2025. (January–October 2025.) Source: Semrush AI Overviews Study
• In Pew Research Center's tracking of 900 U.S. adults' real browsing behavior, 18% of all Google searches in the study generated an AI summary as part of the results. (Browsing data from March 1-31, 2025.) Source: Pew Research Center
• Pew observed traditional-result clicks on 8% of Google visits with an AI summary, versus 15% without one, in March 2025. This comparison is observational; it does not isolate the causal effect of adding a summary. Source: Pew Research Center
• Users who saw a Google AI summary clicked one of the source links inside the summary itself on only 1% of visits, per Pew Research Center's behavioral study. (Browsing data from March 2025.) Source: Pew Research Center
• SparkToro’s Similarweb-based U.S. browser panel estimated 68.01% zero-click Google searches in January–April 2026. It excludes the Google mobile app. The older comparison used a different panel, so it is not a same-user longitudinal study. Source: SparkToro / Similarweb, methodology and results
• Google CEO Sundar Pichai announced on the company's Q2 2025 earnings call that AI Overviews has over 2 billion monthly users across more than 200 countries and territories and 40 languages, up from 1.5 billion the prior quarter. (Announced July 23, 2025.) Source: Google Q2 2025 earnings call (via Digiday)
• Statcounter’s September 2026 worldwide search-engine share was 89.94% for Google and 5.29% for Bing. Its method measures search referrals in its website network, not every search query or chatbot interaction. Source: Statcounter Global Stats
• Similarweb traffic estimates reported in early January 2026 put ChatGPT at 64.5% of the measured generative-AI chatbot web traffic and Gemini at 21.5%. This is a dated web-traffic comparison, not total users, mobile-app use or API activity. Source: Similarweb (via Android Central / Yahoo Tech)

How to use these statistics responsibly

Save the source, observation date, denominator and measurement method alongside any figure you reuse. Keep “reported,” “estimated” and “forecast” in your sentence when the evidence requires them. Weekly users cannot be directly ranked against monthly users, and app downloads do not equal people.

This review compared the published claims with their linked research, company disclosures or identified reporting. It did not independently reproduce private datasets. Vendor statements and commercial research retain those limitations. AI-content detectors can misclassify both human and machine writing; their estimates should not be used as proof about an individual author.

FAQ

How big is the AI market in 2026?

There is no single interchangeable measure. Gartner’s May forecast put annual AI spending at $2.59 trillion. Stanford’s investment figures measure a different flow of money. Use the definition that matches your question.

How many people use AI chatbots?

The company reports cited above use different periods: ChatGPT reported 900 million weekly users in February 2026, while Google reported 950 million monthly Gemini app users for Q2. Those totals overlap and should not be added together.

Does the IEA’s 485 TWh figure measure AI alone?

No. It covers all data-center electricity consumption in 2025, including workloads unrelated to AI. The report separately discusses AI-focused facilities.

Do these figures show that AI caused job losses?

Employer-stated layoff reasons and observational employment comparisons offer evidence about different questions. Neither establishes a single economy-wide causal job-loss total.

Editorial correction, October 8, 2026: Reviewed the original 93 entries, removed a duplicate and three capex comparisons, corrected benchmark normalization and adoption definitions, and updated superseded employment, AI-content and search-share figures. Original publication date preserved. Report an error through our contact page.