Generative AI Statistics: Insights and Emerging Trends for 2026

• The short version
• How to read the numbers on this page
• Generative AI adoption statistics 2026
• Generative AI spending and investment statistics 2026
• Generative AI ROI statistics 2026
• Generative AI statistics for software development
• Generative AI workforce statistics 2026
• AI agent statistics 2026
• Generative AI risk and security statistics 2026
• AI regulation dates that matter in 2026
• Energy and infrastructure statistics 2026
• What to do with these numbers
• Frequently asked questions

The short version

• Worldwide AI spending will reach $2.59 trillion in 2026, a 47% increase over $1.765 trillion in 2025 (Gartner, May 2026).
• 88% of organizations report using AI in at least one business function, up from 78% the prior year, but only 6% are high performers attributing 5%+ of EBIT to AI (McKinsey, State of AI global survey, 2025).
• 52% of US employees now use AI at work at least a few times a year and 15% use it daily — the first time the figure has been above half (Gallup, Q2 2026, fielded May 2026).
• Only 19.8% of US firms report using AI in any business function as of May 2026, because the enterprise surveys everyone quotes are weighted toward large companies (US Census Bureau, Business Trends and Outlook Survey, May 2026).
• 90% of software developers now use AI in their work, a 14-point jump in one year, yet only 24% report high trust in AI-generated code (Google Cloud DORA, September 2025).
• Enterprise spend on generative AI applications and infrastructure hit $37 billion in 2025, up 3.2x from $11.5 billion in 2024 (Menlo Ventures, December 2025).
• One in four malicious breaches is now AI-enabled, and those breaches cost $6 million on average versus a $4.99 million global average (IBM, Cost of a Data Breach Report, July 2026).

How to read the numbers on this page

Every figure below carries a named source and a year. That matters more than usual in 2026, because "AI adoption" is measured at least four different ways and the answers range from 18% to 90% depending on who you ask and what you count.

Three things to keep straight. First, firm-level surveys (Census, Federal Reserve) count all businesses including small ones, and produce low numbers. Second, enterprise surveys (McKinsey, Deloitte) sample large organizations and produce high numbers. Third, worker-level surveys (Gallup) count individual usage, which runs ahead of official organizational adoption.

The McKinsey-versus-Census gap in particular is a sampling difference, not a difference in what the question counts. Since November 2025 the Census BTOS has asked firms about AI use in any business function — the same breadth McKinsey uses — so what separates 88% from 19.8% is which firms are in the sample and how they are weighted, not a narrow production-only measure on one side.

None of them are wrong. They answer different questions. We label which is which throughout.

Generative AI adoption statistics 2026

Enterprise adoption

• 88% of organizations report regular AI use in at least one business function, up from 78% the year before (McKinsey, State of AI global survey, 2025).
• 33% of organizations report scaling AI across the enterprise; roughly two-thirds have not begun scaling (McKinsey, State of AI global survey, 2025).
• Only 25% of organizations have moved 40% or more of their AI pilots into production, though 54% expect to within three to six months (Deloitte, State of AI in the Enterprise, January 2026).

The firm-level reality check

The numbers above come from surveys of large enterprises. Widen the sample to all US businesses and the picture changes sharply.

• 19.8% of US firms reported using AI in any business function as of May 3, 2026, with the rate hovering between 17% and 20% across the survey window (US Census Bureau, Business Trends and Outlook Survey, May 2026).
• 37% of US firms with 250+ employees use AI, versus under 20% of firms with fewer than 20 employees (US Census Bureau, BTOS, May 2026).
• By sector, current AI use runs 39.7% in Information, 33.9% in Finance and Insurance, and 14% in Retail Trade (US Census Bureau, BTOS, May 2026).
• About 18% of US firms had adopted AI as of year-end 2025 — but 78% of the US labor force works at a firm that uses AI, because adopters are large (Federal Reserve Board, FEDS Notes, April 2026).

The gap between 88% and 20% is not a measurement error. It is the difference between "large enterprises have a pilot" and "the median American business has changed how it works."

Worker and consumer adoption

• 52% of employed US adults use AI in their role at least a few times a year, 30% use it weekly or more, and 15% use it daily (Gallup, Q2 2026, fielded May 6–20, 2026, n=22,573).
• ChatGPT reached 900 million weekly active users and 50 million paying subscribers as of February 27, 2026, up from 800 million weekly users in October 2025 (OpenAI, via TechCrunch, February 2026).
• 59% of people globally said in 2025 that AI's benefits outweigh its drawbacks, up from 55% in 2024 (Stanford HAI, 2026 AI Index Report, via IEEE Spectrum, 2026).

Generative AI spending and investment statistics 2026

What organizations are spending

• Worldwide AI spending will total $2.59 trillion in 2026, a 47% increase over $1.765 trillion in 2025 (Gartner, May 2026).
• AI infrastructure accounts for $1.432 trillion of 2026 spending, the single largest category (Gartner, May 2026).
• AI services will reach $585.5 billion and AI software $453.2 billion in 2026 (Gartner, May 2026).
• Enterprise spending on generative AI reached $37 billion in 2025, up from $11.5 billion in 2024 and $1.7 billion in 2023 (Menlo Ventures, December 2025).
• Foundation model API spend accounted for $12.5 billion of 2025 enterprise generative AI spending (Menlo Ventures, December 2025).
• AI coding tools represented $4.0 billion in 2025, or 55% of all departmental generative AI spend (Menlo Ventures, December 2025).
• AI cybersecurity spending will reach $51.3 billion in 2026 (Gartner, May 2026).
• Spending on AI models will reach $32.6 billion in 2026, up 110% year over year — the fastest-growing segment (Gartner, May 2026).
• AI platforms will account for $29.9 billion of 2026 spending (Gartner, May 2026).

Investment

• Global AI private investment hit a record $581 billion in 2025, up from $253 billion in 2024 and above the prior 2021 record of $360 billion, with the United States capturing $344 billion of it (Stanford HAI, 2026 AI Index Report, via IEEE Spectrum, 2026).

We deliberately leave third-party "global AI market size to 2035" projections off this page. They are vendor-modelled, they vary by hundreds of billions between firms, and Gartner's spending forecast above measures something real. If you need a market-size figure for a deck, use the Gartner spend number and say what it counts.

Model vendor share

• Anthropic holds 40% of enterprise LLM usage share, up from 24% in 2024 (Menlo Ventures, December 2025).
• OpenAI holds 27% of enterprise LLM usage share, down from 50% in 2023 (Menlo Ventures, December 2025).
• Google holds 21% of enterprise LLM usage share, up from 7% in 2023 (Menlo Ventures, December 2025).

Generative AI ROI statistics 2026

This is the category where the numbers are least flattering and most useful.

• 39% of organizations attribute any level of EBIT impact to AI, and most of those cite less than 5% attribution (McKinsey, State of AI global survey, 2025).
• 6% of organizations qualify as high performers, attributing 5% or more of EBIT to AI (McKinsey, State of AI global survey, 2025).
• 80% of organizations cite efficiency as an objective of their AI initiatives (McKinsey, State of AI global survey, 2025).
• Roughly 95% of enterprise generative AI pilots showed no measurable P&L return, according to a widely cited MIT Media Lab NANDA study (MIT NANDA, August 2025).
• High performers are nearly three times as likely to have redesigned workflows around AI rather than layering it onto existing ones (McKinsey, State of AI global survey, 2025).
• Companies most exposed to AI grew labour productivity 34% between 2018 and 2025, versus 24% for the least exposed (PwC, Global AI Jobs Barometer, June 2026).

Where this claim is contestable

The "95% of pilots fail" figure deserves scrutiny. It comes from a preliminary MIT NANDA working paper, it has been criticised for its sample construction and its definition of failure, and it does not sit comfortably alongside McKinsey's finding that 39% of organizations see some EBIT impact. If you are using it in a board deck, use it as directional evidence that pilots rarely reach production, not as a precise failure rate.

The honest read across all the ROI data: the constraint is rarely the model. It is workflow redesign, data readiness, and the willingness to change how a process actually runs. McKinsey's own high-performer analysis points at workflow redesign, not model choice, as the differentiator. If your organization is not prepared to redo the process, the more defensible decision is to delay the AI investment rather than fund another pilot.

If your numbers look like the 88% rather than the 6%, the gap is usually workflow and data, not tooling. HatchWorks' AI Strategy & Roadmap engagement exists to find which of your processes will actually pay for the AI you are about to buy.

Generative AI statistics for software development

Developer adoption

• 90% of software development professionals now use AI in their work, up 14 percentage points in one year (Google Cloud DORA, State of AI-assisted Software Development, September 2025, n≈5,000).
• 51% of professional developers use AI tools daily (Stack Overflow Developer Survey, 2025).
• Nearly 80% of new GitHub developers use Copilot within their first week (GitHub Octoverse, October 2025).

What the productivity studies actually found

This is where the marketing numbers and the measured numbers diverge, and it is the most important section on this page for engineering leaders.

• Over 80% of developers report that AI has improved their productivity (Google Cloud DORA, September 2025).
• In a randomized controlled trial, 16 experienced open-source developers took 19% longer to complete 246 real issues when allowed to use AI tools — while estimating afterwards that AI had sped them up by 20% (METR, July 2025).
• Continuing that same experiment design through late 2025 — 57 developers, 143 repositories and 800+ tasks — produced two estimates, both negative: −18% (a slowdown, confidence interval −38% to +9%) for the 10 returning developers, and −4% (confidence interval −15% to +9%) for the 47 newly recruited developers (METR, February 2026).
• METR is changing the experiment design going forward because of selection effects, and says of the figures above that "these issues make it challenging to interpret our central estimate, and we believe it is likely a bad proxy for the real productivity impact of AI tools on these developers" (METR, February 2026).

Read those together. Self-reported productivity is high and consistent. Measured productivity, in the only rigorous randomized trials anyone has run on experienced developers, comes out negative in every cohort measured, with confidence intervals wide enough to cross zero. METR itself says the selection effects run downward — developers who most expect AI to help decline to take part, and 30–50% of participants reported holding back tasks they thought AI would speed up — so the true speedup could be higher than they measured. They are not claiming AI makes developers slower in general. They are claiming nobody has cleanly proven it makes them faster, and that there is no published result yet from the redesigned experiment.

For teams evaluating an AI coding rollout, that argues for measuring your own delivery metrics rather than trusting a vendor's speed claim. Our Generative Driven Development methodology is built around instrumenting that measurement rather than assuming it.

Trust in AI-generated code

• Only 24% of developers report high trust in AI-generated code, while 30% report low trust (Google Cloud DORA, September 2025).
• Only 3.1% of developers highly trust the accuracy of AI output; 46% actively distrust it (Stack Overflow Developer Survey, 2025).
• Developer trust in AI tools fell to 29% in 2025 from 40% in 2024, even as the share using or planning to use AI tools climbed to 84% (Stack Overflow, February 2026).
• 66% of developers name "AI solutions that are almost right, but not quite" as their biggest frustration (Stack Overflow Developer Survey, 2025).

If you don't use AI today, you won't deliver at the same capacity as your peers.

Fernando Manzo, Full Stack Engineer, HatchWorks AI

Generative AI workforce statistics 2026

• Early-career workers aged 22–25 in the most AI-exposed occupations saw a 16% relative decline in employment since widespread generative AI adoption (Stanford Digital Economy Lab, "Canaries in the Coal Mine?", November 2025).
• Employment for more experienced workers in the same occupations and for workers in less-exposed fields remained stable or continued to grow, and the adjustment shows up in employment levels rather than wages (Stanford Digital Economy Lab, November 2025).
• Entry-level software developer and customer support roles show notable decline, while mid-career and senior positions held steady or increased (Stanford HAI, 2026 AI Index Report, via IEEE Spectrum, 2026).
• Workers with AI skills command a 62% average wage premium, up from 57% the prior year, ranging from 16% in government to 118% in consumer markets (PwC, Global AI Jobs Barometer, June 2026, based on 1 billion+ job ads across 27 countries).
• AI-related job postings are growing 69% annually versus 9% for the overall job market (PwC, Global AI Jobs Barometer, June 2026).
• In the United States, entry-level postings in AI-exposed roles grew 35% since 2019, while other entry-level postings declined 10% (PwC, Global AI Jobs Barometer, June 2026, US-only data).
• Among AI-adopting organizations, 34% are hiring or expanding while 23% are reducing headcount (Gallup, Q1 2026, fielded February 2026, n=23,717).

The pattern across all of these: AI is compressing the bottom of the career ladder while raising the price of experience. That is the same squeeze we wrote about in our analysis of the software engineer shortage — the shortage did not disappear, it moved up-market.

AI agent statistics 2026

• Gartner predicts over 40% of agentic AI projects will be canceled by the end of 2027, citing escalating costs, unclear value, and inadequate risk controls (Gartner, June 2025).
• Gartner estimates only about 130 of the thousands of self-described agentic AI vendors offer genuine agentic capability; the rest are "agent washing" (Gartner, June 2025).
• Fewer than 1% of enterprise software applications included agentic AI in 2024; Gartner projects 33% by 2028 (Gartner, June 2025).
• 74% of organizations expect moderate or greater use of AI agents by 2027, and 23% expect extensive use (Deloitte, April 2026, n=3,235 across 24 countries).
• Only 21% of organizations report mature governance models for agentic AI (Deloitte, April 2026).
• 65% of data leaders expect many business processes to be augmented or replaced by agentic AI within two years (Cloudera and Harvard Business Review Analytic Services, March 2026).

The 40% cancellation forecast and the 21% governance-maturity figure are the same story told twice. Agents fail in production for operational reasons — unclear decision boundaries, no audit trail, no anomaly monitoring — not because the models cannot reason. We cover the patterns that survive contact with production in our guide to orchestrating AI agents, and build them as agentic AI automation engagements.

Generative AI risk and security statistics 2026

Security and breach cost

• The global average cost of a data breach was $4.99 million in 2026 (IBM, Cost of a Data Breach Report, July 2026, n=602 organizations, breaches March 2025–February 2026).
• AI-enabled breaches cost $6 million on average, roughly $1 million above the global average (IBM, July 2026).
• One in four malicious breaches is now AI-enabled, a 56% increase over the prior year (IBM, July 2026).
• More than 20% of organizations reported breaches that targeted their AI models or AI applications directly (IBM, July 2026).
• 38% of employees have shared confidential company data with unapproved AI systems (Cloud Security Alliance, 2025, cited by Stack Overflow, February 2026).

Reliability, incidents, and trust

• 51% of organizations have experienced negative consequences from AI use, with nearly one-third of all respondents citing consequences stemming from AI inaccuracy (McKinsey, State of AI global survey, 2025).
• The AI Incident Database passed 1,361 cumulative recorded incidents, adding 108 in the November 2025–January 2026 window alone, with deepfake-enabled fraud the largest single cluster (AI Incident Database, February 2026).
• Trust that government will regulate AI well ranges from 81% in Singapore to 31% in the United States (Stanford HAI, 2026 AI Index Report, via IEEE Spectrum, 2026).

Model-level failure modes are also getting better documented — reward hacking, emergent misalignment, and behavioural contagion through synthetic data are now named risks rather than research curiosities. We go deeper on those in our piece on AI model misbehavior.

Data readiness

• Only 7% of enterprises say their data is completely ready for AI, and 27% say it is not very ready or not at all ready (Cloudera and Harvard Business Review Analytic Services, March 2026, n=230+, surveyed October 2025).
• The top obstacles to AI-ready data are siloed data and integration issues (56%), lack of a clear data strategy (44%), data quality and bias issues (41%), and regulatory constraints (34%) (Cloudera and HBR Analytic Services, March 2026).

Seven percent is the number to put in front of anyone proposing an agent rollout this quarter. Data readiness is the rate limiter on nearly every stalled AI programme we see, which is why we run an AI Data Readiness & Governance Assessment before scoping delivery, and why the data engineering work usually precedes the model work.

AI regulation dates that matter in 2026

• EU AI Act Article 50 transparency obligations apply from August 2, 2026, with a four-month grace period to December 2, 2026 for watermarking of systems already on the market (EU Digital Omnibus agreement, via Gibson Dunn, May 2026).
• High-risk obligations were deferred: standalone systems under Annex III to December 2, 2027, and embedded systems under Annex I to August 2, 2028 (EU Digital Omnibus agreement, via Gibson Dunn, May 2026).
• General-purpose AI model obligations under Articles 51–56 took effect August 2, 2025 and were not changed by the omnibus (EU Digital Omnibus agreement, via Gibson Dunn, May 2026).

If you sell into the EU, the practical consequence is that transparency and disclosure obligations bind from August 2, 2026, while the heavier conformity-assessment work you may have scoped for 2026 has moved to late 2027. That is breathing room, not a reprieve.

Energy and infrastructure statistics 2026

• Data centres consumed approximately 485 TWh of electricity in 2025 (IEA, Key Questions on Energy and AI, 2026). Against global electricity demand of roughly 30,000 TWh, that is about 1.7% — our own arithmetic, because the IEA publishes a 2030 share but no 2025 share.
• Data centre electricity consumption is projected to reach roughly 950 TWh by 2030, about 3% of global electricity (IEA, 2026).
• Data centre electricity demand grew 17% in 2025, while AI-focused data centre demand surged 50% (IEA, 2026).

One historical figure worth keeping

• The cost of inference for a GPT-3.5-level model fell 280-fold between November 2022 and October 2024 (Stanford HAI, 2025 AI Index Report, April 2025 — historical figure, not 2026 data).

We flag this one explicitly because it is frequently mis-cited as a 2026 statistic. It is a 2025 AI Index finding covering a 2022–2024 window. It is still the cleanest single illustration of why unit economics stopped being the blocker, but do not date it to this year.

What to do with these numbers

Three conclusions survive the whole dataset.

Adoption is not a moat. At 88% enterprise adoption and 52% worker adoption, having AI is table stakes. The 6% EBIT figure is where the differentiation lives, and it correlates with workflow redesign, not with tooling choice.

Measure your own delivery, not the vendor's benchmark. The gap between DORA's 80% self-reported productivity gain and METR's measured 4% slowdown is the single most expensive misunderstanding in enterprise AI right now.

Data readiness gates everything else. Seven percent of enterprises call their data completely AI-ready, and 21% have mature agent governance. Those two numbers explain most of Gartner's projected 40% agentic project cancellation rate.

Frequently asked questions

What percentage of companies use generative AI in 2026?

It depends on company size. 88% of large organizations report using AI in at least one business function (McKinsey, 2025), but only 19.8% of all US firms report using AI in any business function (US Census Bureau, May 2026). Both figures measure the same breadth of AI use, so the gap is a sampling difference: the enterprise figure is the one usually quoted; the Census figure is the one that reflects the whole economy.

Is generative AI actually delivering ROI?

For most organizations, not yet at the level that shows up in financial statements. 39% of organizations report some EBIT impact from AI, but only 6% attribute 5% or more of EBIT to it (McKinsey, 2025). The organizations that do see returns are nearly three times as likely to have redesigned workflows rather than layering AI onto existing processes.

How much are companies spending on AI in 2026?

Worldwide AI spending will reach $2.59 trillion in 2026, up 47% from $1.765 trillion in 2025 (Gartner, May 2026). Of that, $1.432 trillion goes to AI infrastructure. Enterprise spending specifically on generative AI applications and infrastructure was $37 billion in 2025 (Menlo Ventures, December 2025).

Does AI make software developers faster?

Developers believe so, but the measured evidence does not show it. 90% of developers use AI and over 80% report productivity gains (Google Cloud DORA, September 2025), yet METR's randomized trial found experienced developers were 19% slower with AI in early 2025, and continuing that experiment through late 2025 measured a 4% slowdown for 47 newly recruited developers (confidence interval −15% to +9%) and an 18% slowdown for the 10 returning developers (confidence interval −38% to +9%) (METR, February 2026).

Is AI eliminating entry-level jobs?

The evidence points that way for the most exposed occupations. Workers aged 22–25 in the most AI-exposed jobs saw a 16% relative employment decline (Stanford Digital Economy Lab, November 2025), and entry-level software development and customer support roles are declining while mid-career and senior roles hold steady (Stanford HAI, 2026 AI Index). Wages have not adjusted — the change is showing up in headcount.

How this page is maintained

This page is refreshed annually. Every statistic is stored as a structured row with claim, value, source, and year, and the three interactive components read from that same dataset. The 2027 refresh replaces values in one place rather than rewriting sections. Figures older than the current cycle are either replaced or explicitly relabelled as historical, as with the 280-fold inference cost figure above.

The statistics that matter for your organization are the ones you generate yourself. HatchWorks' Forward Deployed Engineers embed with your teams to get AI into production and instrument the results — so next year you are quoting your own numbers, not Gartner's.

Sources

• AI Incident Database, "AI Incident Roundup – November and December 2025 and January 2026", February 2026, incidentdatabase.ai
• Cloudera and Harvard Business Review Analytic Services, "Only 7% of Enterprises Say Their Data Is Completely Ready for AI", March 2026, cloudera.com
• Deloitte, "From Ambition to Activation: State of AI in the Enterprise 2026", January 2026, deloitte.com
• Deloitte, "Agentic AI is scaling faster than guardrails", April 2026, deloitte.com
• Federal Reserve Board, "Monitoring AI Adoption in the U.S. Economy", FEDS Notes, April 2026, federalreserve.gov
• Gallup, "Rising AI Adoption Spurs Workforce Changes", Q1 2026, gallup.com
• Gallup, "Organizational AI Adoption Jumps Six Points", Q2 2026, gallup.com
• Gartner, "Gartner Forecasts Worldwide AI Spending to Grow 47% in 2026", May 2026, gartner.com
• Gartner, "Gartner Predicts Over 40% of Agentic AI Projects Will Be Canceled by End of 2027", June 2025, gartner.com
• Gibson Dunn, "EU AI Act Omnibus Agreement — Postponed High-Risk Deadlines and Other Key Changes", May 2026, gibsondunn.com
• GitHub, "Octoverse: A new developer joins GitHub every second as AI leads TypeScript to #1", October 2025, github.blog
• Google Cloud / DORA, "2025 DORA Report: State of AI-assisted Software Development", September 2025, blog.google
• IBM, "IBM Study: One in Four Malicious Breaches are AI-Enabled, Costing Companies $6 Million on Average", July 2026, newsroom.ibm.com
• IEA, "Key Questions on Energy and AI — Executive Summary", 2026, iea.org
• IEEE Spectrum, "Stanford's AI Index for 2026 Shows the State of AI", 2026, spectrum.ieee.org
• McKinsey & Company, "The State of AI: Global Survey", 2025, mckinsey.com
• Menlo Ventures, "2025: The State of Generative AI in the Enterprise", December 2025, menlovc.com
• METR, "Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity", July 2025, metr.org
• METR, "We are Changing our Developer Productivity Experiment Design", February 2026, metr.org
• MIT Media Lab NANDA, "The GenAI Divide: State of AI in Business 2025", August 2025, via Forbes coverage (the working paper itself is not publicly retrievable), forbes.com
• PwC, "AI reshapes global labour market into two distinct paths: 2026 Global AI Jobs Barometer", June 2026, pwc.com
• Stack Overflow, "2025 Developer Survey: AI", 2025, survey.stackoverflow.co
• Stack Overflow, "Mind the gap: Closing the AI trust gap for developers", February 2026, stackoverflow.blog
• Stanford Digital Economy Lab, "Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence", November 2025, digitaleconomy.stanford.edu
• Stanford HAI, "The 2026 AI Index Report", 2026, hai.stanford.edu
• Stanford HAI, "The 2025 AI Index Report", April 2025 (historical inference cost figure), via Tom's Hardware, tomshardware.com
• TechCrunch, "ChatGPT reaches 900M weekly active users", February 2026, techcrunch.com
• US Census Bureau, "Large Firms With at Least 20 Employees Biggest AI Users", Business Trends and Outlook Survey, May 2026, census.gov