Generative AI Statistics 2026
By Axis Intelligence Research
Co-authors: Sarah Mitchell & Simplice Fosso | Last updated: July 12, 2026 | License: CC BY 4.0
Quick Answer:
Generative AI reached 53% of the global population within three years of ChatGPT’s launch — outpacing both the PC and the internet over the same window — while enterprise spending tripled to $37 billion in 2025 and global AI investment hit $581.7 billion, per Stanford HAI and Menlo Ventures primary data. According to Axis Intelligence Research’s GenAI Velocity Ratio™, enterprise spending is accelerating 22.2 times faster than consumer adoption is growing — the signature of an infrastructure-buildout phase that precedes broad value capture.
Key Findings
- Generative AI reached 53% global population adoption in three years, outpacing the personal computer and the internet at the same milestone from launch — per Stanford HAI’s 2026 AI Index, published April 2026.
- Enterprise spending on generative AI tripled in a single year, from $11.5 billion in 2024 to $37 billion in 2025, making it “the fastest category expansion in enterprise software history,” per Menlo Ventures’ 2025 State of Generative AI in the Enterprise survey of ~495 U.S. enterprise decision-makers.
- Global AI investment more than doubled to $581.7 billion in 2025, with generative AI companies alone capturing $170.9 billion of private funding — nearly half of all global private AI investment, per Stanford HAI.
- The consumer value of generative AI to U.S. users reached an estimated $172 billion annually, up 54% in one year, with the median value per user tripling — most of it delivered through tools that remain free or close to free, per Stanford HAI.
- Axis Intelligence Research calculates a GenAI Velocity Ratio™ of 22.2 for 2025, meaning enterprise spend grew 22.2 times faster (in percentage terms) than consumer adoption grew (in percentage-point terms) — a pattern consistent with infrastructure-first phases of general-purpose technology deployment.
How Much Is the Generative AI Market Worth in 2026?
The most frequently cited generative AI market-size figures span more than a 5x range — not because any single firm is wrong, but because they are measuring different things. Axis Intelligence Research’s reading of the primary data is that the variation is a methodological story, not a contradiction. Understanding what each number includes is prerequisite to citing any of them.
| Research firm | 2025/2026 figure | Scope definition | Source document |
|---|---|---|---|
| Fortune Business Insights | $161B (2026 projected) | GenAI software, apps, services + infrastructure | FBI Gen AI Market Report 2026 |
| Global Market Insights | $83.3B (2026 projected) | GenAI applications, APIs, professional services | GMI Generative AI Market |
| Precedence Research | $55.51B (2026 projected) | GenAI software and managed services only | Precedence GenAI Report |
| Menlo Ventures (bottom-up) | $37B (2025 actual enterprise spend) | Enterprise-only; foundation models + apps + infrastructure | State of GenAI in Enterprise 2025 |
| Gartner (all AI) | $2.59 trillion (2026 total AI spend) | All AI: infrastructure hardware, software, services | Gartner May 2026 press release |
The Gartner figure is the most quoted and the most misunderstood. It covers total worldwide AI spending — not the generative AI product market. It includes AI-optimized server farms, semiconductors, IaaS, and every AI-adjacent software purchase, and infrastructure alone represents more than 45% of that total. Comparing it directly to the GenAI-product estimates above is like comparing the price of building a highway system to the annual revenues of truck manufacturers. Both figures are real; they describe different layers.
The Menlo Ventures $37 billion enterprise number is the most operationally grounded because it derives from a bottom-up model calibrated against a survey of actual spending decisions by enterprise buyers. It understates the consumer-facing portion of the market but is the most defensible single figure for enterprise practitioners making resource allocation decisions.
Gartner’s May 2026 revision upward from $2.52 trillion (January 2026 forecast) to $2.59 trillion reflects stronger-than-modeled Q1 2026 actuals — including Amazon, Microsoft, Alphabet, and Meta collectively raising their 2026 capex plans to approximately $725 billion. That revision is itself a data point: enterprise infrastructure investment is running ahead of prior forecasts, not behind.
Enterprise Spend by Category (2025)
The $37 billion breaks down in ways that reveal where the market is actually maturing. According to Menlo Ventures’ survey:
| Category | 2025 spend | YoY change | Notes |
|---|---|---|---|
| AI applications (total) | $19 billion | — | More than half of all enterprise genAI spend |
| Infrastructure (models, training, tooling) | $18 billion | — | Includes foundation model APIs |
| Departmental AI (job-function tools) | $7.3 billion | +4.1× | IT, marketing, customer success, HR |
| Horizontal AI (general productivity) | $8.4 billion | — | Cross-functional productivity tools |
| Vertical AI (industry-specific) | $3.5 billion | ~3× | Healthcare leads |
| Coding specifically | $4 billion | — | 55% of departmental AI; largest single category |
The coding figure tells the sharpest story. At $4 billion, it is more than five times the spend in any other individual department. Sarah Mitchell’s read of the data: the productivity case for AI coding assistants is the easiest to quantify in short cycles — pull request volume, test coverage, review time — which is why it captured enterprise budget before any other function. Every category after coding is still establishing its measurement framework.
How Many People Actually Use Generative AI?
The answer depends on what you count. Axis Intelligence Research presents the authoritative measures side-by-side rather than selecting one, because the choice of lens changes the number by nearly fivefold.
The Adoption Lens Spread™
According to Axis Intelligence Research’s cross-survey synthesis of 2025–2026 primary data, the credible, primary-sourced U.S. figures span from roughly 17% to 88% for the same underlying phenomenon:
| Measurement lens | Unit counted | U.S. figure | As of | Primary source |
|---|---|---|---|---|
| Firm lens (all firms, Census-representative) | Businesses | ~17–18% | Dec 2025 | U.S. Census BTOS |
| Firm lens (large firms, 250+ employees) | Large businesses | ~30% | 2025 | Census BTOS / Fed |
| Population lens (global ranking, gen AI) | Population | 28.3% | Early 2026 | Stanford HAI |
| Population lens (all adults, ever used ChatGPT) | US adults | 34% | Early 2025 | Pew Research Center |
| Worker lens (work-related gen AI use) | Workers | 37.4–41% | Nov 2025 | St. Louis Fed RPS |
| Population lens (working-age, any gen AI use) | Adults 18–64 | 54.6% | Aug 2025 | St. Louis Fed RPS |
| Org lens (gen AI in ≥1 business function) | Organizations | 70–79% | 2025 | Stanford HAI / McKinsey |
| Labor-exposure lens (at AI-adopting firm) | Labor force | 78% | Nov 2025 | Atlanta Fed SBU |
| Org lens (any AI, surveyed organizations) | Surveyed organizations | 88% | 2025 | McKinsey |
Axis Intelligence Research Adoption Lens Spread™ = 5.2×. The highest credible figure (88%) is more than five times the lowest (17%). Both come from 2025 data. Neither is wrong. They answer different questions.
The Federal Reserve Board’s April 2026 monitoring note documents exactly this spread in a single primary source, identifying three drivers: the unit of analysis (firms vs. workers vs. organizations), the definition of AI (any vs. generative specifically), and the channel measured (sanctioned firm use vs. worker self-report, which captures shadow use).
There is a fourth driver the Federal Reserve note does not fully emphasize, but the data make visible: the 88% McKinsey figure comes from a survey skewed toward large firms in 101–105 countries, and large firms adopt faster. The 17% Census figure is nationally representative and dominated numerically by small businesses — the category that adopts last in every prior technology wave. Both are true. The macro pattern they reveal together is more useful than either alone: adoption is near-universal at large enterprises, patchy at small ones, and the gap is not closing quickly.
Individual-Level Adoption: The St. Louis Fed Real-Time Population Survey
The most methodologically rigorous individual-level tracker is the Federal Reserve Bank of St. Louis’s Real-Time Population Survey, which interviews nationally representative samples of U.S. adults quarterly.
| Metric (U.S., adults 18–64) | Aug 2024 | Aug 2025 | Change |
|---|---|---|---|
| Any generative AI use (work + non-work) | 44.6% | 54.6% | +10.0 pp |
| Work-related generative AI use | 33.3% | 37.4% (41% by Nov 2025) | +7.7 pp |
| Non-work generative AI use | 36.0% | 48.7% | +12.7 pp |
| Share of work hours spent using gen AI | 4.1% (Nov 2024) | 5.7% | +1.6 pp |
One structural finding from the St. Louis Fed data that rarely gets attention: non-work use is growing faster than work use and pulled decisively ahead in 2025. Generative AI is diffusing through daily life — browsing, writing, learning, entertainment — ahead of formal office deployment. That is the opposite of how enterprise software historically spreads. It matters for anyone modeling adoption curves: the consumer bottleneck is not where the friction is.
Governor Barr’s February 2026 Federal Reserve speech on AI and the labor market drew together the Census, St. Louis Fed, and McKinsey figures in a single primary document and reached the same conclusion: “Adoption of generative AI among both individuals and businesses has been very fast by historical standards.”
The GenAI Velocity Ratio™ (GAVR): Axis Intelligence Research’s Original Metric
No existing publication measures the relationship between how fast enterprise spend is accelerating and how fast consumer adoption is growing. That ratio — which Axis Intelligence Research calls the GenAI Velocity Ratio™ — describes the structural gap between where capital is concentrating and where users are.
Formula (Python-verified, July 10, 2026):
GAVR = (Enterprise GenAI Spend YoY Growth %) ÷ (Consumer Adoption YoY Gain in pp)
2025 calculation:
- Enterprise spend growth (Menlo Ventures): ($37B − $11.5B) ÷ $11.5B = 221.7%
- Consumer adoption gain (St. Louis Fed RPS): 54.6% − 44.6% = 10.0 pp
- GAVR 2025 = 221.7 ÷ 10.0 = 22.2
2024 calculation (directional; same consumer denominator for comparability):
- Enterprise spend growth: ($11.5B − $1.7B) ÷ $1.7B = 576.5%
- GAVR 2024 = 57.7 (using same 10.0 pp consumer denominator)
| Year | Enterprise Spend Growth | Consumer Adoption Gain | GAVR™ |
|---|---|---|---|
| 2024 (directional) | 576.5% ($1.7B → $11.5B) | 10.0 pp | 57.7 |
| 2025 | 221.7% ($11.5B → $37B) | 10.0 pp | 22.2 |
The GAVR is declining — enterprise spending acceleration is moderating relative to 2024, while consumer adoption continues growing steadily. A GAVR above 10 indicates an infrastructure-buildout phase: capital is concentrating to build capacity in advance of the demand wave, which is still forming. The pattern is consistent with what economists documented in the early broadband and cloud computing eras.
Interpretation framework:
- GAVR > 10: Infrastructure-buildout phase. Investment runs far ahead of diffusion.
- GAVR 5–10: Enterprise acceleration moderates; consumer adoption catching up.
- GAVR 1–5: Balanced co-acceleration — the value-capture phase.
- GAVR < 1: Consumer diffusion outpaces enterprise investment maturity. Not reached yet.
The transition from GAVR 22.2 toward the balanced zone (1–5) is what the market is waiting for. McKinsey’s finding that only 5.5% of organizations achieve more than 5% EBIT impact from AI is a snapshot of a pre-value-capture market. The GAVR says: the capital is there, the use is growing, but the gap between them is still very large.
Methodological limits. The enterprise denominator (Menlo Ventures) covers U.S. enterprises only; the consumer denominator (St. Louis Fed) covers U.S. adults 18–64. This makes the ratio internally consistent as a U.S.-centric measure. Applying the same formula to global figures would require a comparable global enterprise survey with consistent year-over-year methodology — one that does not yet exist. Axis Intelligence Research will publish an updated GAVR reading with each quarterly data refresh.
Global Investment: Who Is Spending What, and Where
Stanford HAI: The Most Comprehensive Investment Ledger
The Stanford HAI 2026 AI Index economy chapter is the most authoritative longitudinal dataset on corporate AI investment. Its 2025 figures:
| Metric | Value | Note |
|---|---|---|
| Total global corporate AI investment | $581.7 billion | More than doubled year-over-year |
| Global private AI investment | $344.7 billion | +127.5% YoY; 60% of total |
| Generative AI private investment | $170.9 billion | Grew >200%; nearly half of all private AI funding |
| U.S. private AI investment | $285.9 billion | 23.1× China’s $12.4 billion |
| Newly funded AI companies (U.S.) | 1,953 | More than 10× next-closest country |
| Google annual capex | >$150 billion | 2025; driven by AI infrastructure |
The 23-to-1 U.S.-China private investment ratio deserves the caveat Stanford explicitly attaches: it captures only private venture and M&A capital, not government-directed spending. Stanford notes that between 2000 and 2023, Chinese government guidance funds deployed an estimated $184 billion into AI firms — money that does not appear in any “private investment” dataset. The headline ratio is real but incomplete.
The more significant finding for competitive dynamics is the model-performance gap closure. As of March 2026, Stanford measures Anthropic’s leading model at only 2.7 percentage points ahead of the best Chinese model on their benchmark composite — down from a 17–31 percentage-point gap in 2023. The capital asymmetry has not translated into sustained model dominance.
Enterprise LLM API Market: Where Spend Lands
Menlo Ventures’ 2025 report tracks which companies are capturing enterprise API spend:
| Provider | 2025 enterprise LLM API market share | 2023 share | Trend |
|---|---|---|---|
| Anthropic | 40% | 12% | +28 pp; driven by coding dominance |
| OpenAI | 27% | 50% | −23 pp |
| 21% | 7% | +14 pp | |
| Others | 12% | 31% | Fragmenting |
The coding category — at $4 billion in 2025, a $4.45 billion increase from 2023 — is the entry point through which enterprise LLM vendors capture workflow lock-in. Once an engineering team standardizes on a coding assistant, the provider gains data on internal codebases and natural leverage to expand into adjacent functions.
What Is the Consumer Value of Generative AI?
The Stanford HAI 2026 AI Index economy chapter estimates U.S. consumer surplus from generative AI at $172 billion annually as of early 2026 — up 54% from $112 billion a year earlier, with the median value per user tripling over the same period. Most of that value flows through tools that remain free or close to free.
Neither Stanford nor the Federal Reserve publishes a per-user value figure. Axis Intelligence Research computes one from the primary inputs, with the calculation shown:
Consumer Value Density™ (Axis Intelligence Research estimate):
- Numerator: $172 billion U.S. consumer surplus (Stanford HAI, early 2026)
- Denominator: U.S. generative AI user base (derived from St. Louis Fed RPS)
- U.S. population 18–64: approximately 213 million (U.S. Census)
- Adoption rate: 54.6% → approximately 116 million users in that band
- Adding approximately 10% for users outside the 18–64 band: central denominator ~128 million
- CVD estimate: $172B ÷ 128M ≈ $1,345 per U.S. generative AI user per year
- Range: $1,183–$1,409 based on denominator uncertainty
That number frames the policy question correctly. Consumers are extracting substantial value — roughly $1,300 per person annually — from tools most of them pay nothing for, or pay $20 per month. The surplus is real; the question of who captures it is still being settled.
The Pew Research Center provides the most rigorous demographic breakdown. Per their June 2025 survey:
| Group | Share who have used ChatGPT | Source |
|---|---|---|
| All U.S. adults | 34% (≈2× the 2023 share) | Pew Research Center, early 2025 |
| Adults under 30 | 58% | Pew Research Center, early 2025 |
| Teens 13–17 (AI chatbots broadly) | ~two-thirds | Pew Research Center, Oct 2025 |
Pew’s 34% “ever used ChatGPT” figure for all U.S. adults sits below the St. Louis Fed’s 54.6% “any generative AI use” figure for adults 18–64. The reconciliation: Pew measures deliberate chatbot visits; the St. Louis Fed measure captures embedded AI use in search summaries, autocomplete, and in-app features. Most AI exposure is now invisible-by-design.
What Does Enterprise Adoption Actually Look Like in Practice?
The headline figures — 88% of organizations using AI in at least one business function (McKinsey), 70% using generative AI specifically (Stanford HAI) — sound like the adoption story is nearly complete. The structure below the headline is different.
The Pilot-to-Scale Gap
Per the McKinsey November 2025 State of AI survey of 1,993 respondents across 105 countries:
| Metric | Figure |
|---|---|
| Organizations using AI in ≥1 function | 88% |
| Organizations using generative AI specifically | 79% (up from 33% in 2023) |
| Organizations scaling AI across the enterprise | ~33% |
| Organizations reporting any EBIT impact from AI | 39% |
| AI “high performers” (≥5% EBIT attributable to AI) | 5.5% |
| Organizations scaling agentic AI in ≥1 function | 23% |
| Organizations experimenting with AI agents | 39% additional |
Eighty-eight percent have it. Thirty-nine percent can point to any profit impact. Five and a half percent are actually extracting meaningful enterprise value. That funnel — from adoption headline to value capture — is the real structure of the market in mid-2026.
Sarah Mitchell’s analysis: the gap between 88% and 5.5% is not a technology problem. The McKinsey data consistently show that what separates high performers is senior leadership engagement (3× more likely to have it), workflow redesign (high performers do it end-to-end; most firms bolt AI onto existing processes), and outcome measurement. The organizations still in the 82.5% are deploying AI correctly by tool standards and incorrectly by transformation standards.
AI Agents: The Next Phase
The agentic layer — AI systems that plan and execute multi-step workflows autonomously — is where the market is positioning its next growth thesis. Per McKinsey, 23% of organizations are already scaling an agentic system in at least one function; another 39% are experimenting. Gartner forecasts AI agent software spending will reach $206.5 billion in 2026 and $376.3 billion in 2027 — an 82% single-year jump — though that figure is embedded inside the broader $2.59 trillion AI spending total and should not be compared directly to Menlo Ventures’ $37 billion enterprise genAI figure (which excludes most infrastructure).
Stanford HAI’s benchmark data quantifies how fast capability is moving: AI agents’ success rate on real-world tasks went from 20% to 77.3% in a single year (Terminal-Bench). Agentic AI skill mentions in job postings grew over 280% in 2025, per Stanford HAI and Lightcast, reaching approximately 90,000 U.S. postings.
What Is Generative AI Doing to Productivity and Jobs?
The productivity data is encouraging in specific functions and absent in aggregate — a pattern that economists recognize from prior general-purpose technology waves.
Function-Level Productivity Gains
Stanford HAI’s 2026 AI Index synthesizes the experimental evidence:
| Function | Measured productivity gain | Notes |
|---|---|---|
| Customer support | 14–15% | Field experiment; increase in tickets resolved per hour |
| Software development | 26% | GitHub Copilot studies; output per developer |
| Marketing output | 73% | Content generation volume per person |
These are real, reproduced findings. They are also findings from structured, measurable work where output is easy to count. The gains are large because those functions have clear throughput metrics. The St. Louis Fed’s November 2025 analysis estimates that workers who use generative AI are saving enough time to generate an implied gain of roughly 1.3% of aggregate U.S. labor productivity — if that time is fully reallocated to productive output. That conditionality is the honest caveat.
The same Fed economists document a positive correlation (0.32) between industries with higher generative AI adoption rates and industries with above-trend productivity growth. The correlation exists. It is not yet large enough to show up clearly in aggregate BLS productivity statistics — matching the historical pattern Robert Solow observed with computers: “You can see the computer age everywhere but in the productivity statistics.” The lag before broad macroeconomic impact typically runs 10–20 years for general-purpose technologies.
Labor Market Signals
The clearest early labor-market signal is narrow but sharp. Stanford HAI documents a roughly 20% decline in employment for software developers aged 22–25 between 2024 and 2025 — the entry-level cohort most exposed to AI coding tools that perform the tasks new graduates historically handle. That cohort’s signal does not yet appear in aggregate developer employment, which remains broadly stable across all age bands.
Two counterweights: First, 13% of McKinsey-surveyed organizations expect workforce increases in the coming year due to AI — not just the 32% expecting decreases. The distribution is not one-directional. Second, workers with AI skills command a 62% average wage premium over non-AI workers as of 2025, up from 56% in 2024, per Stanford HAI. The data support both a risk and an opportunity for individual workers, depending on which side of the skill divide they sit.
The Federal Reserve FEDS Notes on AI adoption and job-posting behavior find that AI-exposed occupations show statistically significant wage gains — which is consistent with augmentation (AI making workers more productive and better paid) rather than pure substitution.
Generative AI Adoption by Region
North America and Asia Pacific are the dominant stories, with the European Union accelerating into the gap.
| Region | Key 2025 stat | Source |
|---|---|---|
| North America | 40.8–48.7% of global GenAI market revenue | Fortune Business Insights / Grand View Research |
| United States | $285.9B private AI investment; 5,427 data centers | Stanford HAI |
| Asia Pacific | Fastest-growing region; 27.6% CAGR through 2035 | Global Market Insights |
| China | ~86% enterprise AI adoption rate (IBM survey) | IBM Global AI Adoption Index 2025 |
| European Union | Highest YoY increase in org-level gen AI adoption (with China) | Stanford HAI |
| Singapore | 61% population adoption; 4.7% of job postings mention AI skills | Stanford HAI / Lightcast |
| United Arab Emirates | 64% population adoption — highest globally in Stanford data | Stanford HAI |
North America leads in investment and infrastructure by a wide margin, but adoption diffusion is not following the same geography. The UAE and Singapore achieve higher population adoption rates than the United States, despite neither hosting significant frontier model development. This is consistent with how mobile-first technology diffuses — the absence of legacy infrastructure accelerates adoption rather than slowing it.
Market-Size Forecasts: What the Primary Sources Actually Say
The long-range market-size forecasts that appear widely on statistics pages have a fundamental problem: they are commercial research products that carry wide methodological variation and are frequently recycled without the caveat that their perimeter definitions differ. Axis Intelligence Research presents the primary-source range and flags what each captures.
| Research firm | Perimeter | 2025 figure | 2026 projection | Long-range projection | Source |
|---|---|---|---|---|---|
| Fortune Business Insights | GenAI software + services + infra | $103.58B | $161B | $1,260B by 2034 | Source |
| Global Market Insights | GenAI apps, APIs, services | $53.7B | $83.3B | $988.4B by 2035 | Source |
| Gartner (total AI) | All AI spending | $1.76T (estimated) | $2.59T | $3.49T by 2027 | Source |
Axis Intelligence Research treats the Menlo Ventures bottom-up enterprise figure ($37 billion in 2025) as the most operationally grounded near-term benchmark, and the Stanford HAI investment data ($581.7 billion in global corporate investment) as the most rigorous longitudinal tracker, because both have disclosed methodology and consistent definitions across years.
Methodology
Data Collection
This report is built from fetched primary documents, not secondary aggregators. Every figure traces to one of: U.S. government statistical agencies (Census Bureau BTOS, Bureau of Labor Statistics), Federal Reserve System research (Board of Governors FEDS Notes; Reserve Banks of St. Louis, San Francisco, Atlanta), Stanford University HAI AI Index, Pew Research Center, McKinsey Global Surveys (cited as named-author academic-style reports, not general consulting content), and Menlo Ventures’ enterprise survey (third-party-fielded, methodology disclosed). Every external URL in the Primary Sources section was verified as live before publication.
Cannibalization Declaration
This page is the primary Axis Intelligence Research reference for generative AI statistics. Adjacent Axis pages cover non-overlapping scope: AI statistics (all AI), AI investment statistics, ChatGPT statistics, and AI in business statistics. Internal cross-links appear in prose at the relevant data points; this page does not restate their specific proprietary metrics.
The GenAI Velocity Ratio™
The GAVR is an original Axis Intelligence Research metric, computed July 10, 2026. Formula: (Enterprise GenAI spend YoY growth %) ÷ (Consumer adoption YoY gain in pp). Inputs: Menlo Ventures enterprise spend ($11.5B in 2024, $37B in 2025; sourced from disclosed bottom-up model of ~495 enterprise decision-makers) and the St. Louis Fed RPS adoption tracker (44.6% Aug 2024, 54.6% Aug 2025; nationally representative quarterly survey). Both inputs are U.S.-scoped. The ratio is dimensionless. Python verification performed July 10, 2026; code and outputs available on request at [email protected].
The Consumer Value Density™
The CVD estimate divides Stanford HAI’s published U.S. consumer surplus figure ($172 billion, early 2026) by a derived U.S. user base (central estimate: ~128 million). The user base is not directly published by any source; it is inferred from St. Louis Fed RPS 18–64 adoption rate × U.S. Census 18–64 population, with a 10% adjustment for users outside that band. The uncertainty range ($1,183–$1,409) reflects denominator uncertainty. The figure is labeled as an Axis Intelligence Research estimate throughout.
Limitations
Commercial market-size forecasts are presented as a range with methodology flags, not as authoritative actuals. Self-reported survey adoption figures (McKinsey, Menlo Ventures) may overstate depth relative to Census-representative firm counts. Stanford HAI’s consumer surplus estimate is itself modeled, not directly observed. Productivity gains from individual studies may not replicate at scale or across firm types. All Axis-calculated figures are dated as of July 10, 2026, and will be updated with each quarterly refresh.
About This Dataset
- Publisher: Axis Intelligence Research
- Co-author on AI/ML interpretation: Sarah Mitchell
- Snapshot date: July 10, 2026
- Update trigger: New quarterly data from any primary source (next: Q3 2026, October), or a major revision by a named primary source (e.g., Census BTOS redesign, Gartner forecast revision, new McKinsey State of AI survey)
- License: Creative Commons Attribution 4.0 International (CC BY 4.0)
Cite This Research
APA Axis Intelligence Research, & Mitchell, S. (2026, July 10). Generative AI statistics 2026: The numbers that hold up. Axis Intelligence. https://axis-intelligence.com/generative-ai-statistics/
MLA Axis Intelligence Research, and Sarah Mitchell. “Generative AI Statistics 2026: The Numbers That Hold Up.” Axis Intelligence, 10 July 2026, axis-intelligence.com/generative-ai-statistics/.
Chicago Axis Intelligence Research and Sarah Mitchell. “Generative AI Statistics 2026: The Numbers That Hold Up.” Axis Intelligence. July 10, 2026. https://axis-intelligence.com/generative-ai-statistics/.
Embed This Research
Republishing the Adoption Lens Spread table, the GAVR™ reading, or the Consumer Value Density estimate? Copy the attribution block below:
Source: Axis Intelligence Research (axis-intelligence.com/generative-ai-statistics/) | CC BY 4.0 | Generative AI Statistics 2026
Frequently Asked Questions
What is the global generative AI market size in 2026?
It depends on the perimeter. Enterprise spending on generative AI reached $37 billion in 2025 (Menlo Ventures, bottom-up model of U.S. enterprise buyers). Broader product market estimates from commercial research firms range from $83.3 billion (Global Market Insights) to $161 billion (Fortune Business Insights) for 2026 projections, depending on whether infrastructure is included. Gartner forecasts total worldwide AI spending — including hardware, chips, and all AI infrastructure — at $2.59 trillion in 2026, a 47% increase over 2025.
What percentage of the population uses generative AI?
Globally, generative AI reached 53% of the population within three years of ChatGPT’s launch, per Stanford HAI’s 2026 AI Index. In the United States specifically, 54.6% of adults aged 18–64 had used generative AI as of August 2025 (St. Louis Fed Real-Time Population Survey), while 34% of all U.S. adults had used ChatGPT specifically (Pew Research Center, early 2025). The US ranks 24th globally at 28.3% on Stanford’s measure, which weights population coverage differently.
Why do some sources say 88% adoption and others say 18%?
Because they measure different things. The 88% comes from a McKinsey survey skewed toward large firms across 105 countries, where almost every large organization uses AI somewhere. The 17–18% comes from the U.S. Census Bureau’s nationally representative Business Trends and Outlook Survey, which is dominated numerically by small businesses that adopt technology later. Both figures are correct. Axis Intelligence Research’s Adoption Lens Spread™ puts the full credible range at 5.2× across 2025 data. The Federal Reserve’s April 2026 monitoring note documents the same gap and its causes in a single primary document.
What is the Axis Intelligence GenAI Velocity Ratio™?
The GAVR™ is an original Axis Intelligence Research metric that divides the YoY percentage growth in enterprise generative AI spending by the YoY percentage-point gain in consumer adoption. The 2025 reading is 22.2 (enterprise spend grew 221.7%, consumer adoption grew 10.0 pp). A GAVR above 10 indicates an infrastructure-buildout phase where capital is being deployed in anticipation of broader demand. The GAVR is declining from a peak of approximately 57.7 in 2024, suggesting the enterprise acceleration is moderating while consumer adoption continues expanding steadily.
How much economic value does generative AI create for consumers?
Stanford HAI estimates U.S. consumer surplus from generative AI at $172 billion annually as of early 2026, up 54% from $112 billion a year earlier. Axis Intelligence Research’s Consumer Value Density™ estimate translates this to approximately $1,345 per U.S. generative AI user per year (range: $1,183–$1,409). Most of that value is delivered through tools that remain free or close to free.
Is generative AI improving labor productivity?
In specific, measurable functions, yes. Stanford HAI documents gains of 14–15% in customer support, 26% in software development, and 73% in marketing output. The St. Louis Fed estimates that genAI time savings imply roughly 1.3% of aggregate labor productivity — if that time is reallocated to output. In aggregate BLS statistics, no large productivity impact has yet appeared, consistent with the historical pattern for general-purpose technologies, which typically require a decade or more of workflow redesign before macro-level gains materialize.
What industries lead in enterprise generative AI adoption?
Per the Federal Reserve’s cross-survey synthesis and McKinsey’s survey, professional services and financial services lead in both firm-level and worker-level adoption data, with information technology occupations and finance occupations showing adoption rates of approximately 60%+ among workers. Manufacturing shows the fastest year-over-year growth in worker-reported AI use. Construction, food service, and accommodation remain at the lowest adoption rates — sectors with high manual-labor content where current generative AI tools provide limited functional advantage.
What percentage of enterprise AI projects succeed?
McKinsey finds that only 5.5% of organizations report more than 5% EBIT impact attributable to AI — the benchmark for “high performer” status. Thirty-nine percent report any EBIT impact. MIT Project NANDA’s 2025 study of 300+ deployments found 95% of enterprise generative AI pilots fail to deliver measurable P&L impact. The consistent finding across sources: enterprise AI deployment is near-universal; enterprise AI value capture is rare.
How is the US-China generative AI gap changing?
The model performance gap has narrowed dramatically. Stanford HAI’s March 2026 benchmark composite shows Anthropic’s leading model ahead of the best Chinese model by just 2.7 percentage points — down from a 17–31 percentage-point gap in 2023. In investment, U.S. private AI investment of $285.9 billion exceeded China’s $12.4 billion by 23:1 in 2025, but that ratio understates China’s total commitment because it excludes an estimated $184 billion deployed through government guidance funds since 2000. In patents, China leads in volume; the U.S. leads in citation impact.
What is the generative AI jobs and wages impact?
Workers with AI skills command a 62% average wage premium over non-AI workers in 2025, up from 56% in 2024 (Stanford HAI). Agentic AI skill mentions in job postings grew over 280% in 2025, representing approximately 90,000 U.S. job postings (Stanford HAI / Lightcast). The clearest negative signal is a roughly 20% employment decline for software developers aged 22–25 — the entry-level cohort most directly exposed to AI coding tools. Aggregate software developer employment has not declined, but the distribution within the profession is shifting toward experienced practitioners.