State of AI, Year over Year: 2024 vs 2025 vs 2026
2024 vs 2025 vs 2026 to date, theme by theme: every comparable measure, how much it changed each year, and what the change means.
Thank you for reading this post, don't forget to subscribe!
Three years of data show AI accelerating in scale while several growth rates slow. ChatGPT’s weekly users went from 300 million in December 2024 to 800 million in October 2025 and more than 1 billion in August 2026, but grew only 11% in the last six months of that run. Organisational adoption, as McKinsey measures it, rose from 78% to 88% and then stalled at 89%. Individual use at work kept climbing: Gallup’s share of US employees using AI a few times a week or more went from 12% to 23% to 30%.
The money compounded. Big-four capital expenditure went from about $245 billion in 2024 to about $413 billion in 2025 and $508 billion in the first nine months of 2026. Venture funding to AI companies went from more than $100 billion to $202.3 billion and then about $488 billion in nine months, by Report AI’s calculation from Crunchbase. The combined run-rate revenue of OpenAI and Anthropic rose from about $6.5 billion in December 2024 to about $115 billion in the third quarter of 2026, about 18-fold.
Prices and volumes moved in opposite directions, and so did costs and rules. The cheapest GPT-4-level tokens got 58% cheaper between 2024 and 2026 while Google’s monthly token volume rose about 330-fold. AI-attributed US job cuts went from 12,742 to 54,836 to 121,401 in ten months of 2026, and incidents logged by the AI Incident Database from 233 to 362. Regulation moved from enactment to delay: the EU AI Act entered into force in 2024, applied its first obligations in 2025, and postponed its high-risk rules in 2026. Changes into 2026 compare a year to date or latest reading with a full year; each is marked.
Download the data (CSV) · 417 metrics with source, date and confidence for every value. Free to quote with attribution and a link.
Usage
- ChatGPT’s weekly users grew from 300 million (December 2024) to 800 million (October 2025) to more than 1 billion (August 2026) — roughly 3.3x in under two years (DERIVED: 1,000M / 300M), with growth decelerating in 2026.
- Organisational AI adoption rose from 78% (2024) to 88% (2025) to 89% (2026) in McKinsey’s surveys: the big move was 2023 to 2025, and the 2026 change is within noise.
- Individual use roughly doubled on every general-population measure: US employees using AI at work went from 27% (Q2 2024) to 46% (Q4 2025) to 52% (May 2026) in Gallup’s survey, and US adults aged 18 to 64 using generative AI went from 44.6% (August 2024) to 54.6% (August 2025) to 62% (May 2026) in the St. Louis Fed’s survey.
- Use shifted from work to everyday life: the non-work share of ChatGPT consumer messages rose from 53% (June 2024) to 73% (June 2025) in OpenAI’s NBER study, while AI’s share of website referral traffic stayed below half a percent (0.02% in 2024, 0.24% in 2025, 0.32% in 2026, SE Ranking).
Bars: 2024, 2025, 2026 (to date, blue). Each scaled to its own maximum.
| Platform | Metric | 2024 | 2025 | 2026 (to date) | Basis |
|---|---|---|---|---|---|
| ChatGPT | Weekly active | 300M (Dec) | 800M (Oct) | 1B+ (Aug) | Company disclosure |
| ChatGPT app | Monthly active | – | – | ~1B (May-Jun) | Reported; MEDIUM |
| Gemini app | Monthly active | – | 750M (Q4) | 1B+ (Oct) | Company disclosure |
| Meta AI | Monthly active | ~500M (Sep) | ~1B (May) | CONTESTED (~600M-1.2B) | Company; no 2026 total |
| Copilot apps | Monthly active | – | 150M (Q3) | Not disclosed | Company disclosure |
| Claude | Monthly (estimates) | – | – | CONTESTED (56M-245M) | Third-party estimates |
| Doubao | Monthly (China app) | ~56M (Nov) | 226M (Dec) | 419M (Sep) | QuestMobile panel |
| DeepSeek | Monthly (China app) | n/a (launched Jan 2025) | 135M (Dec) | 122.65M (Sep) | QuestMobile panel |
Notes: Compare only within a metric row group: weekly is never compared with monthly, and company disclosures are not the same kind of number as panel estimates. Chinese figures cover China only. Empty/contested cells: 10 of 24 (DERIVED count). Meta AI 2026: Report AI pages give 1.2B (page 970) and ~600M (page 2589); Meta disclosed no total in its Q3 2026 results. Gemini 2024: only a daily figure (9M, Oct 2024, court exhibit) exists.REPORTAI report-ai.org
| Measure | 2024 | 2025 | 2026 (to date) | 2024→25 | 2025→26 |
|---|---|---|---|---|---|
| ChatGPT weekly active users | 300M | 800M | 1B+ | +167% | +25% |
| ChatGPT paying subscribers | 15.5M | ~35M | 50M+ | +126% | +43% |
| Meta AI monthly active users (across Meta apps) | ~500M | ~1B | ~600M–1.2B | +100% | |
| ByteDance Doubao monthly active users (China app) | ~56M | 226M | 419.08M | +304% | +85% |
| Organisations using AI in at least one business function | 78% | 88% | 89% | +13% | +1% |
| US employees using AI at work at least a few times a year | 27% | 46% | 52% | +70% | +13% |
| US employees using AI at work a few times a week or more | 12% | 23% | 30% | +92% | +30% |
| US adults 18–64 using generative AI for work | 33.3% | 37.4% | 45% | +12% | +20% |
| US adults 18–64 using generative AI (any purpose) | 44.6% | 54.6% | 62% | +22% | +14% |
| US adults who have used ChatGPT | 23% | 34% | 44% | +48% | +29% |
| Share of ChatGPT consumer messages not related to work | 53% | 73% | 68.9% | +38% | -6% |
| Google AI Overviews monthly users | 1B+ | 2B+ | 2.5B | +100% | +25% |
| AI platforms’ share of all website referral traffic | 0.02% | 0.24% | 0.32% | +1,100% | +33% |
Sources, dates and confidence for each value: State of AI Usage 2026.
Spending
- Big-tech capex roughly doubled in two years: about $245 billion in 2024 and $413 billion in 2025 for Microsoft, Alphabet, Amazon and Meta, and $508 billion in the first nine months of 2026 alone (Report AI sums of reported figures).
- IDC’s measured AI infrastructure spending went from $153.9 billion (2024) to $318.7 billion (2025), with $521 billion forecast for 2026; the first two are actuals, the third is a forecast.
- Gartner’s worldwide AI spending series rose from about $988 billion (2024 estimate) to $1.79 trillion (2025 estimate) and a $2.67 trillion forecast for 2026, but category changes between editions mean part of each step is definitional.
- In enterprise LLM APIs the leader changed and then pulled away: Anthropic’s share of enterprise API spend went 24% (2024), 40% (2025), 55% (2026), while OpenAI went 34%, 27%, 31% (Menlo Ventures, an Anthropic investor).
Bars: 2024, 2025, 2026 (to date, blue). Each scaled to its own maximum.
| Measure | 2024 | 2025 | 2026 |
|---|---|---|---|
| Big-four capex (US$B) | ~245 | ~413 | ~508 Jan–Sep; ~730 guided |
| IDC AI infrastructure (US$B) | 153.9 | 318.7 | 521 (forecast) |
| Gartner AI spending (US$T) | 0.99 (est.) | 1.79 (est.) | 2.67 (forecast) |
| Menlo enterprise GenAI (US$B) | 11.5 (restated) | 37 | n/a: new measure |
| Anthropic API share (%) | 24 | 40 | 55 |
| Data-centre power (TWh) | ~415 | ~485 | not yet estimated |
| AI share of IT spend (%) | 19 | 32 | 42 (forecast) |
Notes: Big-four capex and AI share of IT spending are DERIVED. Gartner’s 2024 and 2025 values come from different editions with different categories. Menlo’s 2026 report uses an annualised run-rate ($176.8B) that is not a successor to the $37B. The 2026 column mixes year-to-date actuals and forecasts; read each cell’s label.REPORTAI report-ai.org
| Measure | 2024 | 2025 | 2026 (to date) | 2024→25 | 2025→26 |
|---|---|---|---|---|---|
| Worldwide AI spending (Gartner) | $988B | $1.79T | $2.67T (forecast) | +81% | +49% |
| AI infrastructure spending (IDC) | $153.9B | $318.7B | $521B (forecast) | +107% | +63% |
| AI spending as a share of worldwide IT spending (Gartner on Gartner) | 19% | 32% | 42% (forecast) | +66% | +31% |
| Anthropic share of enterprise LLM API spending (Menlo Ventures) | 24% | 40% | 55% | +67% | +38% |
| OpenAI share of enterprise LLM API spending (Menlo Ventures) | 34% | 27% | 31% | -21% | +15% |
| Google share of enterprise LLM API spending (Menlo Ventures) | ~12% | 21% | 9% | +75% | -57% |
| Microsoft capital expenditure, calendar year (incl. finance leases) | $75.6B | $118.0B | $118.1B (Jan–Sep, to date) | +56% | +0% |
| Alphabet capital expenditure, calendar year | $52.5B | $91.4B | $132.0B (Jan–Sep, to date) | +74% | +44% |
| Amazon capital expenditure, calendar year (cash, net of proceeds and incentives) | $77.7B | $131.8B | $172.2B (Jan–Sep, to date) | +70% | +31% |
| Meta capital expenditure, calendar year (incl. finance-lease principal) | $39.2B | $72.2B | ~$85.8B (Jan–Sep, to date) | +84% | +19% |
| Oracle capital expenditure, fiscal year ending 31 May (gross) | $21.2B (FY2025, Jun 2024–May 2025) | $55.7B (FY2026, Jun 2025–May 2026) | $28.5B (FY2027 Q1, Jun–Aug 2026, to date) | +163% | -49% |
| Combined capex of Microsoft, Alphabet, Amazon and Meta, calendar year | ~$245B | ~$413B | ~$508B (Jan–Sep, to date) | +69% | +23% |
Sources, dates and confidence for each value: State of AI Spending 2026.
Finance
- AI’s share of global venture funding rose from about a third in 2024 to about half in 2025 and about 72% in January–September 2026 (Crunchbase basis; 2026 is Report AI’s calculation from quarterly figures).
- OpenAI’s valuation rose from $157 billion (October 2024) to $500 billion (October 2025) to a reported $1.25 trillion (November 2026); Anthropic’s priced valuation went from about $18 billion (early 2024) to $183 billion (September 2025) to $965 billion (May 2026).
- Model-maker revenue grew far faster than capex: OpenAI’s recognised revenue went from about $3.7 billion (2024) to $13.07 billion (2025), and Anthropic’s from about $0.4 billion to $4.6 billion to $16.3 billion in H1 2026 alone, narrowing big-four capex per lab revenue dollar from about 60x to 23x to about 13x on 2026 guidance.
- Nvidia’s data-centre revenue went from $115.2 billion (fiscal year to January 2025) to $193.7 billion (to January 2026) to $164.2 billion in the first half of fiscal 2027 alone, while the financing of AI shifted from cash flow to bonds, equity and off-balance-sheet vehicles.
Bars: 2024, 2025, 2026 (to date, blue). Each scaled to its own maximum.
| Indicator | 2024 | 2025 | 2026 (to date) |
|---|---|---|---|
| AI venture funding (Crunchbase) | $100B+ | $202.3B | ~$488B (Jan–Sep) |
| AI share of global VC (Crunchbase) | ~34% | ~50% | ~72% (Jan–Sep) |
| AI share of US VC (PitchBook) | 46.4% | 65.4% | 82.7% (Jan–Sep) |
| Largest AI round | Databricks $10B | OpenAI $40B | OpenAI $122B |
| OpenAI valuation | $157B | $500B | $1.25T (Nov) |
| Anthropic valuation (priced) | ~$18.4B | $183B | $965B (May) |
| OpenAI revenue | ~$3.7B | ~$13.07B | ~$12.4B (H1) |
| Anthropic revenue | ~$0.4B | ~$4.6B | $16.3B (H1) |
| Nvidia data-centre revenue | $115.2B (FY25) | $193.7B (FY26) | $164.2B (H1 FY27) |
| Big-4 capex | ~$246B | ~$410B | $720–740B (guidance) |
| Capex per $ lab revenue | ~60x | ~23x | ~13x |
| Largest AI-related IPO | None of scale | CoreWeave $1.5B | SpaceX $75B |
Notes: 2026 values are to date with the period stated; half-year and nine-month figures are not comparable with full years. DERIVED: 2026 AI VC and share, capex ratio, Nvidia H1 FY27. CONTESTED: Anthropic 2025 valuation ($183B priced vs ~$350B implied), xAI 2025, Cursor 2026 ARR. 2026 capex is guidance, a forecast.REPORTAI report-ai.org
| Measure | 2024 | 2025 | 2026 (to date) | 2024→25 | 2025→26 |
|---|---|---|---|---|---|
| Global venture funding to AI companies (Crunchbase) | $100B+ | $202.3B | ~$488B (Jan–Sep, to date) | +102% | +141% |
| AI share of global venture funding (Crunchbase) | ~34% (about a third) | ~50% | ~72% (Jan–Sep, to date) | +47% | +44% |
| Total global venture funding, all sectors (Crunchbase) | $314B | $425B | $679B (Jan–Sep, to date) | +35% | +60% |
| AI share of US venture deal value (PitchBook-NVCA) | 46.4% of $209B | 65.4% of $339.4B | 82.7% of $515.8B (Jan–Sep, to date) | +41% | +26% |
| Largest AI venture round of the year | Databricks $10B at $62B | OpenAI $40B at $300B | OpenAI $122B at $852B | +300% | +205% |
| OpenAI valuation (latest of the year) | $157B | $500B | $1.25T (to date) | +218% | +150% |
| OpenAI revenue (recognised, calendar year) | ~$3.7B | ~$13.07B | ~$12.4B (H1 2026, to date) | +253% | -5% |
| OpenAI annualised revenue run-rate | ~$5.5B | ~$20B | ~$50B (to date) | +264% | +150% |
| OpenAI operating loss | ~$5B loss | $20.92B | ~$21.6B (H1 2026, to date) | +318% | +3% |
| Anthropic valuation (latest priced round of the year) | ~$18.4B | $183B | $965B (to date) | +895% | +427% |
| Anthropic revenue (recognised) | ~$0.4B | ~$4.6B | $16.3B (H1 2026, to date) | +1,050% | +254% |
| Anthropic annualised revenue run-rate | ~$1B | ~$9B | >$65B (to date) | +800% | +622% |
| Anthropic operating loss (profit) | $2.98B loss | $8.06B loss | $693M loss (H1); $974M profit in Q2 | +170% | -91% |
| xAI valuation | $45B–$50B | ~$200B (reported, denied) | $250B (in SpaceX merger) | +300% | +25% |
| Mistral AI valuation | ~€5.8B (~$6.2B) | €11.7B (~$14B) | >€21B (~$24B) | +126% | +71% |
| Nvidia data-centre revenue | $115.2B (FY2025) | $193.7B (FY2026) | $164.2B (H1 FY2027, to date) | +68% | -15% |
| Combined capex of Amazon, Alphabet, Microsoft and Meta | ~$246B | ~$410B | $720B–$740B (full-year guidance, not an outcome) | +67% | +78% |
| Big-four capex per dollar of OpenAI + Anthropic revenue | ~60x | ~23x | ~13x (guidance vs annualised H1 revenue) | -61% | -45% |
| Cursor (Anysphere) annualised revenue | ~$100M | ~$1B | $2.6B–$4B | +900% |
Sources, dates and confidence for each value: State of AI Finance 2026.
Work
- AI-attributed US job cuts rose from 12,742 in 2024 to 54,836 in 2025 and 121,401 in January–October 2026; as a share of all announced cuts, from about 1.7% to 4.5% to 19.7% (Report AI’s division of Challenger figures).
- The share of US Indeed postings mentioning AI rose from 2.71% at the end of 2024 to 4.44% at the end of 2025 and 7.34% on 30 September 2026, about 2.7 times in 21 months.
- US software-development postings sat about 32.5% below their February 2020 level at the end of both 2024 and 2025, then recovered to 18.4% below by 30 October 2026: AI did not stop developer hiring, but the recovery favours senior roles while early-career employment in exposed occupations kept falling behind.
- PwC’s AI skills wage premium in job advertisements rose from 25% (2023 ads) to 56% (2024 ads, restated 57%) to 62% (2025 ads), and daily AI use among US employees rose from about 4% (Q2 2024) to 12% (late 2025) to 15% (May 2026).
Bars: 2024, 2025, 2026 (to date, blue). Each scaled to its own maximum.
| Indicator | 2024 | 2025 | 2026 (to date) |
|---|---|---|---|
| AI-attributed US job cuts (Challenger) | 12,742 | 54,836 | 121,401 (Jan–Oct) |
| AI share of all announced cuts | 1.7% | 4.5% | 19.7% |
| US Indeed postings mentioning AI | 2.71% (Dec) | 4.44% (Dec) | 7.34% (Sep) |
| US software-development postings vs Feb 2020 | −32.5% (Dec) | −32.7% (Dec) | −18.4% (Oct) |
| US employees using AI daily (Gallup) | ~4% (Q2) | 12% (Q4) | 15% (May) |
| AI skills wage premium in job ads (PwC) | 56% | 62% | — |
| Early-career gap, AI-exposed jobs (Stanford) | — | 16% | 19% |
| Work hours saved by generative AI (St. Louis Fed) | 1.4% | ~2.0% | ~2.2% (Q1) |
| Hiring managers who rehired an AI-cut role | — | — | 32% |
Notes: Each row is one series; compare across columns, not across rows. Share-of-cuts row is DERIVED. Gallup changed its question in Q3 2025. PwC values are by year of ads analysed (2025 = 2026 edition); PwC restated 2024 as 57%. Stanford 2025 value is the Nov 2025 revision (RESTATED from 13%). St. Louis Fed 2025–26 values read from a secondary summary.REPORTAI report-ai.org
| Measure | 2024 | 2025 | 2026 (to date) | 2024→25 | 2025→26 |
|---|---|---|---|---|---|
| US job cuts attributed to AI (Challenger) | 12,742 | 54,836 | 121,401 (Jan–Oct, to date) | +330% | +121% |
| All announced US job cuts (Challenger) | 761,358 | 1,206,374 | 617,189 (Jan–Oct, to date) | +58% | -49% |
| AI-attributed cuts as a share of all US announced cuts | 1.7% | 4.5% | 19.7% (Jan–Oct, to date) | +165% | +338% |
| Cumulative US job cuts attributed to AI since 2023 (Challenger) | 16,989 | 71,825 | 193,226 (to date) | +323% | +169% |
| Share of US job postings on Indeed mentioning AI-related terms | 2.71% | 4.44% | 7.34% (to date) | +64% | +65% |
| Share of UK job postings on Indeed mentioning AI-related terms | 3.44% | 7.12% | 9.92% (to date) | +107% | +39% |
| US software-development job postings on Indeed vs 1 Feb 2020 | 67.5 (−32.5%) | 67.3 (−32.7%) | 81.6 (−18.4%, to date) | -0% | +21% |
| All US job postings on Indeed vs 1 Feb 2020 | 110.6 | 103.4 | 103.8 (to date) | -7% | +0% |
| US employees using AI at work daily (Gallup) | 4% | 12% | 15% (to date) | +200% | +25% |
| Share of all US work hours saved by generative AI (St. Louis Fed survey) | 1.4% | about 2.0% | about 2.2% (to date) | +43% | +8% |
Sources, dates and confidence for each value: State of AI Work 2026.
Models & Prices
- Coding: the top SWE-bench Verified score went from 71.7% (o3 preview, Dec 2024) to 80.9% (Claude Opus 4.5, Nov 2025) to 97.0% (Claude Opus 5, Sep 2026, independent harness) — saturated in under two years.
- Competition: the top US model’s Arena lead over China went 9.26% (Jan 2024) → 1.70% (Feb 2025) → 2.7% (Mar 2026), and the closed-over-open lead 8.04% → 1.70% → 3.3%: convergence, then a modest reopening.
- Frontier prices: Anthropic’s Opus-tier output price fell from $75 (2024) to $25 (2025) to $20 per million tokens (2026), and OpenAI’s newest flagship from $60 (o1) to $10 (GPT-5, GPT-6 Sol) — but a new top tier appeared above Opus at $50 (Claude Fable 5.1).
- Reasoning: AI went from a 28/42 silver at the 2024 IMO to a certified 35/42 gold in 2025 to reported perfect 42/42 scores in 2026.
Bars: 2024, 2025, 2026 (to date, blue). Each scaled to its own maximum.
| Indicator | 2024 | 2025 | 2026 |
|---|---|---|---|
| SWE-bench Verified, top | 71.7% | 80.9% | 97.0% |
| GPQA Diamond, top | 87.7% | 93.8% | 94.3% |
| ARC-AGI, top verified | 75.7% (v1) | 54% (v2) | 84.6% (v2) |
| IMO, best AI (of 42) | 28 | 35 | 42 (reported) |
| US lead over China | 9.26% | 1.70% | 2.7% |
| Closed lead over open | 8.04% | 1.70% | 3.3% |
| Notable models US / China | 40 / 15 | 50 / 30 (CONTESTED) | not yet published |
| Opus-tier output $/M | $75 | $25 | $20 |
| OpenAI flagship output $/M | $60 | $10 | $10 |
| Haiku-tier output $/M | $4 | $5 | $0.50 |
| Largest training run | ~3.8×10^25 FLOP | ~5×10^26 FLOP | not disclosed |
| Flagship context (median) | 200K | 400K | ~1M |
| Energy per query | ~0.3 Wh (est.) | 0.24 Wh (Google) | 0.31 Wh (MSR) |
Notes: Years are the year the reading describes. US–China and closed–open gaps are Arena readings from Jan 2024, Feb 2025 and Mar 2026. ARC-AGI changed version between 2024 and 2025. Flagship context is the median of OpenAI, Anthropic and Google flagships (DERIVED). Energy readings use different methods each year. 2026 SWE-bench uses an independent harness.REPORTAI report-ai.org
| Measure | 2024 | 2025 | 2026 (to date) | 2024→25 | 2025→26 |
|---|---|---|---|---|---|
| SWE-bench Verified, top reported score | 71.7% | 80.9% | 97.0% (to date) | +13% | +20% |
| GPQA Diamond (graduate-level science questions), top reported score | 87.7% | 93.8% | 94.3% (to date) | +7% | +1% |
| Humanity’s Last Exam, top reported score | ~9% | 41.0% (no tools) | 54.8% (to date) | +356% | +34% |
| ARC-AGI, top verified score (version changes: ARC-AGI-1 in 2024, ARC-AGI-2 from 2025) | 75.7% (ARC-AGI-1) | 54% (ARC-AGI-2) | 84.6% (ARC-AGI-2, to date) | -29% | +57% |
| MMLU, frontier score (saturation marker) | 92.3% | >92% (saturated; labs report MMLU-Pro instead) | >92% for every frontier system | ||
| Lead of the top US model over the best Chinese model, Chatbot Arena / LMArena | 9.26% | 1.70% | 2.7% | -82% | +59% |
| Lead of the top closed-weight model over the best open-weight model, Chatbot Arena / LMArena | 8.04% | 1.70% | 3.3% | -79% | +94% |
| Rate of price decline at a fixed capability level | ~10× per year | ~50× per year median; ~200× per year since Jan 2024 | ~47% per quarter (≈13× per year) | ||
| Anthropic Opus-tier list price (input / output) | $15 / $75 (Claude 3 Opus) | $5 / $25 (Claude Opus 4.5) | $4 / $20 (Claude Opus 5.5) | -67% | -20% |
| OpenAI newest flagship, list price at launch (input / output) | $15 / $60 (o1) | $1.25 / $10 (GPT-5) | $2 / $10 (GPT-6 Sol) | -83% | +0% |
| Anthropic Haiku-tier (smallest model) list price (input / output) | $0.80 / $4 (Claude 3.5 Haiku) | $1 / $5 (Claude Haiku 4.5) | $0.10 / $0.50 (Claude Haiku 5.5) | +25% | -90% |
| DeepSeek flagship list price, open-weight price floor (input / output) | $0.27 / $1.10 (DeepSeek-V3) | $0.55 / $2.19 (DeepSeek-R1) | ~$0.87 output (DeepSeek V4-Pro) | +99% | -60% |
| Context window, median of OpenAI, Anthropic and Google flagships | 200K | 400K | ~1M (to date) | +100% | +162% |
| Best AI result at the International Mathematical Olympiad | 28/42 (silver standard) | 35/42 (gold, officially certified) | 42/42 (perfect scores reported) | +25% | +20% |
| Energy per typical AI text query | ~0.3 Wh (GPT-4o, estimate) | 0.24 Wh (median Gemini text prompt) | 0.31 Wh median; ~3.9 Wh for long reasoning responses | -20% | +29% |
| Notable frontier model releases | Claude 3 (Mar); GPT-4o (May); Claude 3.5 Sonnet (Jun); Llama 3.1 405B (Jul); o1-preview (Sep) and o1 (Dec); Gemini 2.0 Flash (Dec); DeepSeek-V3 (26 Dec) | DeepSeek-R1 (20 Jan); Claude 3.7 Sonnet, GPT-4.5 (Feb); Gemini 2.5 Pro (Mar); Llama 4, o3 (Apr); Claude 4 (May); Grok 4 (Jul); GPT-5 (7 Aug); Gemini 3 Pro (18 Nov); Claude Opus 4.5 (24 Nov); Mistral Large 3 (Dec) | Gemini 3.1 Pro (Feb); Meta Muse Spark (8 Apr; closed weights); DeepSeek V4-Pro (24 Apr); Gemini 3.5 Flash (19 May); Claude Opus 4.8 (28 May); Claude Fable 5 (9 Jun; suspended 12 Jun–1 Jul under export controls); Grok 4.5 (8 Jul); GPT-5.6 Sol (9 Jul); Qwen3.8-27B (14 Aug); Claude Fable 5.1 and Mythos 5.1 (1 Sep); GPT-6 Astra (3 Sep); GPT-6 Sol/Luna and Claude Opus 5.5 (22 Sep); Claude Haiku 5.5 (7 Oct); Gemini 4 Argon GA (13 Oct); Meta Muse Spark Open (15 Oct); Mistral Large 4 weights (29 Oct) |
Sources, dates and confidence for each value: State of AI Models & Prices 2026.
Tokens
- Tokens per dollar at GPT-4 quality rose from about 2.1 million (2024) to 3.2 million (2025) to 5.0 million (2026), against 27,000 for GPT-4 in March 2023, so most of the gain came before 2024. At OpenAI’s flagship the figure barely moved: 229,000, then 291,000, then 250,000.
- Google’s monthly token volume went from 9.7 trillion (spring 2024) to 480 trillion (May 2025) to 3.2 quadrillion (May 2026); Doubao’s daily volume went from 4 trillion (Dec 2024) to 50 trillion (Dec 2025) to 180 trillion (Jun 2026).
- Combined OpenAI and Anthropic run-rate revenue went from about $6.5 billion (Dec 2024) to $29 billion (end 2025) to $115 billion (Q3 2026): about 18 times in under two years, an order of magnitude less than token volume. Menlo’s enterprise spend estimates ($11.5 billion for 2024; $37 billion, restated to $33 billion, for 2025; an annualised $176.8 billion in Sep 2026 on a new definition) are not comparable across that break.
- Reported business return improved only slightly. Wharton’s positive-ROI share went from about 65% to 74% to 80%, McKinsey’s EBIT-contribution share held near 37-39%, and the share reporting strong or full returns stayed below one in three.
Bars: 2024, 2025, 2026 (to date, blue). Each scaled to its own maximum.
| Indicator | 2022 | 2023 | 2024 | 2025 | 2026 |
|---|---|---|---|---|---|
| Tokens per $1, GPT-3.5 level | 50,000 | 800,000 | 14.3M | – | – |
| Tokens per $1, GPT-4 level | no such model | 66,667 | 2.09M | 3.17M | 5.0M |
| Tokens per $1, OpenAI flagship | 50,000 | 66,667 | 228,571 | 290,909 | 250,000 |
| OpenAI API tokens per minute | – | ~300M | – | 6B+ | – |
| Google tokens per month | – | – | 9.7T | 1.3 quadrillion | 3.2 quadrillion |
| Doubao tokens per day | not launched | not launched | 4T+ | ~50T | 180T+ |
| OpenRouter tokens per week | not founded | – | ~10B | ~7T | ~194T |
| Enterprise GenAI spend (Menlo) | – | $1.7B (restated from $2.3B) | $11.5B | $37B (restated $33B) | $176.8B annualised, new definition (not comparable) |
| OpenAI + Anthropic revenue | ~$28M (OpenAI recognised; different basis) | ~$1.7B run rate | ~$6.5B | ~$29B | ~$115B |
| Leaders reporting positive ROI (Wharton) | – | – | ~65% | 74% | 80% |
Notes: Each cell is the latest reading in that calendar year; months differ by row (see data file). Tokens per dollar, run rates, OpenRouter 2025 and Wharton 2024 are DERIVED. Units differ by row. ‘-‘ means no sourced value. 2022 revenue is OpenAI’s full-year recognised revenue and is not comparable with the later run rates. Menlo’s 2023 figure is on its 2025-report basis (RESTATED from $2.3B); its 2026 figure uses a new definition and is not comparable with earlier totals (CONTESTED); no multiple is computed from that row. 2026 values are to date.REPORTAI report-ai.org
| Measure | 2024 | 2025 | 2026 (to date) | 2024→25 | 2025→26 |
|---|---|---|---|---|---|
| Tokens per $1 at GPT-4 (March 2023) quality, cheapest named model | 2.09M | 3.17M | 5.0M | +52% | +58% |
| Tokens per $1 at OpenAI’s flagship list price | 228,571 | 290,909 | 250,000 | +27% | -14% |
| Tokens per $1 at Anthropic’s flagship list price (Opus tier from 2024) | 33,333 | 100,000 | 125,000 | +200% | +25% |
| Google: tokens processed per month across all products | 9.7T | 1.3 quadrillion | 3.2 quadrillion | +13,302% | +146% |
| OpenRouter: tokens routed per week | ~10B | ~7T | ~194T | +69,900% | +2,671% |
| ByteDance Doubao models: tokens processed per day | 4T+ | ~50T | 180T+ | +1,150% | +260% |
| China: average daily token calls, national | ~0.1T (100B) | ~100T | 140T+ | +99,900% | +40% |
| Disclosed tokens per month, Google plus ByteDance (cross-platform floor) | ~130T | ~2.8 quadrillion | ~8.6 quadrillion | +2,059% | +207% |
| OpenAI plus Anthropic revenue (combined annualised run rate; 2022 recognised revenue) | ~$6.5B | ~$29B | ~$115B | +346% | +297% |
| Jevons check: token volume index (Google monthly tokens, May each year), 2024 = 100 | 100 | 4,948 | 32,990 | +4,848% | +567% |
| Jevons check: revenue index, OpenAI + Anthropic run-rate revenue, 2024 = 100 | 100 | 446 | 1,769 | +346% | +297% |
| Enterprise leaders reporting positive ROI from generative AI (Wharton / GBK) | ~65% | 74% | 80% (28% strongly positive) | +14% | +8% |
Sources, dates and confidence for each value: State of AI Tokens 2026.
Infrastructure & Energy
- Nvidia’s data-centre revenue went from $115.2 billion (FY ending Jan 2025) to $193.7 billion (FY ending Jan 2026) to $164.2 billion in the first half of fiscal 2027 alone; custom silicon grew faster, with Broadcom’s AI revenue going from $12.2 billion to about $20 billion to $35.9 billion in three quarters.
- North American primary-market vacancy fell from 2.8% (H1 2024) to 1.6% (H1 2025) to 1.4% (H1 2026) while capacity under construction nearly doubled from 3,872 MW to 7,481 MW: building faster has not loosened the market.
- The scale of power deals grew roughly eightfold in two years, from Microsoft’s 835 MW reactor restart (2024) to Amazon’s 1,920 MW Talen contract (2025) to Meta’s 6.6 GW nuclear package (2026), and PJM capacity prices rose from $269.92 to the $325–333 cap.
- H100 rental prices fell from about $8 an hour (2023) to about $2 (late 2024 and December 2025), then rose to about $2.88 (October 2026), a reversal that marks the shift from training-led to inference-led demand.
Bars: 2024, 2025, 2026 (to date, blue). Each scaled to its own maximum.
| Indicator | 2024 | 2025 | 2026 to date |
|---|---|---|---|
| Nvidia data-centre revenue | $115.2B (FY) | $193.7B (FY) | $164.2B (Feb–Jul) |
| Broadcom AI revenue | $12.2B (FY) | ~$20B (FY) | $35.9B (Q1–Q3) |
| AMD Data Center revenue | $12.6B | $16.6B | $21.0B (Q1–Q3) |
| OpenAI compute, year end | 0.6 GW | 1.9 GW | 10 GW contracted (Apr) |
| N. America vacancy (H1) | 2.8% | 1.6% | 1.4% |
| N. America under construction (H1) | 3,872 MW | 5,243 MW | 7,481 MW |
| Data-centre electricity, world | ~415 TWh (IEA) | ~485 TWh (IEA) / 447 (Gartner) | 565 TWh (Gartner forecast) |
| Largest tech nuclear deal | 835 MW | 1,920 MW | 6.6 GW |
| PJM capacity price (auction year) | $269.92 | $329.17 | $325 (cap) |
| H100 rental, $/GPU-hr | ~$2 (Oct, resale) | $2.00 (Dec, index) | $2.88 (Oct, index) |
| Energy per text prompt | ~2.9 Wh (estimate) | 0.24 Wh (Google) | 0.31 Wh (MSR median) |
Notes: Not every row is like-for-like across years. OpenAI 2026 is contracted capacity, not operating (no operating figure published). World electricity 2025 is CONTESTED and 2026 is a forecast from a different source. H100 2024 is a resale-market reading, 2025–26 the Silicon Data index. Energy per prompt uses different methods each year. 2026 vendor revenue values are DERIVED sums of reported quarters.REPORTAI report-ai.org
| Measure | 2024 | 2025 | 2026 (to date) | 2024→25 | 2025→26 |
|---|---|---|---|---|---|
| Nvidia data-centre revenue | $115.2B | $193.7B | $164.2B (to date) | +68% | -15% |
| Nvidia total revenue | $130.5B | $215.9B | $177.8B (to date) | +65% | -18% |
| AMD Data Center segment revenue (EPYC CPUs + Instinct GPUs) | $12.6B | $16.6B | $21.0B (Q1–Q3, to date) | +32% | +27% |
| Broadcom AI semiconductor revenue (custom XPUs + AI networking) | $12.2B | $20B | $35.9B (Q1–Q3, to date) | +64% | +80% |
| North America primary-market data-centre vacancy (H1) | 2.8% | 1.6% | 1.4% | -43% | -13% |
| North America primary-market data-centre capacity under construction (H1) | 3,871.8 MW | 5,242.5 MW | 7,481 MW | +35% | +43% |
| Share of North America primary-market construction already preleased (H1) | ~80% | 74.3% | 80.4% | -7% | +8% |
| Largest tech-company nuclear power deal announced in the year | 835 MW (Microsoft–Constellation, Crane/TMI-1) | 1,920 MW (Amazon–Talen, Susquehanna) | 6.6 GW (Meta–Vistra, TerraPower, Oklo) | +130% | +244% |
| PJM capacity auction clearing price (RTO), by year auction was held | $269.92/MW-day (2025/26 delivery) | $329.17/MW-day (2026/27 delivery, at cap) | $325/MW-day (2028/29 delivery, at cap) | +22% | -1% |
| Nvidia H100 rental price (neocloud / market) | ~$2/GPU-hr (resale market) | $2.00/GPU-hr (low) | $2.88/GPU-hr | +0% | +44% |
Sources, dates and confidence for each value: State of AI Infrastructure & Energy 2026.
Geography
- US private AI investment rose from $109.1 billion in 2024 to $285.9 billion in 2025 while China’s rose from $9.3 billion to $12.4 billion, widening the ratio from about 12x to 23x (Stanford HAI).
- The US share of newly added AI compute climbed from about 60% (2024) to 72% (2025) to 78% (H1 2026), according to Epoch AI, even as China’s share of AI patents and open-model downloads rose.
- The US-China top-model gap fell from 9.26% (Jan 2024) to 1.70% (Feb 2025) and then held near parity at 2.7% (Mar 2026) on Chatbot Arena/LMArena (Stanford HAI).
- China’s generative AI user base grew from 249 million (Dec 2024) to 602 million (Dec 2025) to more than 700 million (Jun 2026), according to CNNIC, making it the largest national user base by that survey measure.
Bars: 2024, 2025, 2026 (to date, blue). Each scaled to its own maximum.
| Indicator | 2024 | 2025 | 2026 to date |
|---|---|---|---|
| US private AI investment | $109.1B | $285.9B | n/a until 2027 |
| China private AI investment | $9.3B | $12.4B | n/a until 2027 |
| US share of AI start-up funding | 76% | 79% | ~88% |
| US share of new AI compute | ~60% | 72% | 78% (H1) |
| US-China Arena gap | 9.26% (Jan) | 1.70% (Feb) | 2.7% (Mar) |
| Notable models, US : China | 40 : 15 | 50 : 30 or 59 : 35 | n/a |
| China gen-AI users (CNNIC) | 249M (Dec) | 602M (Dec) | 700M+ (Jun) |
| Europe AI start-up funding | ~$10B | $17.5B | $23B (H1) |
| Nvidia revenue from China | ~19% (FY25) | 9.1% (FY26) | ~8.2% (Q2 FY27) |
| World working-age AI use | n/a | 16.3% (H2) | 18.8% (Jun) |
| China gen-AI filings (CAC) | 302 | 748 | 1,112 (Aug) |
Notes: Each cell is the latest reading available for that year; 2026 values are to date and often half-year. Notable models 2025 CONTESTED (two Stanford versions). Nvidia China share uses a customer-headquarters count; by billing location FY2025 was 13% (CONTESTED basis). Europe 2024 CONTESTED (Dealroom $12.8B). Rows measure different things and should be read across, not down.REPORTAI report-ai.org
| Measure | 2024 | 2025 | 2026 (to date) | 2024→25 | 2025→26 |
|---|---|---|---|---|---|
| US share of global AI start-up funding | 76% | 79% | ~88% | +4% | +11% |
| European AI start-up venture funding | ~$10B (Crunchbase) / $12.8B (Dealroom) | $17.5B | $23B (H1) | +75% | +31% |
| Mistral AI valuation (Europe’s most valuable AI lab) | €5.8B | €11.7B | €21B+ | +102% | +79% |
| US share of new AI compute added worldwide | ~60% | 72% | 78% (H1) | +20% | +8% |
| Nvidia revenue from China (customer HQ basis) | 19.2% (HQ basis) / 13% (billing basis) | 9.1% | ~8.2% | -53% | -10% |
| Performance gap, top US vs top Chinese model (Chatbot Arena / LMArena) | 9.26% | 1.70% | 2.7% | -82% | +59% |
| Chinese open-weight models’ share of tokens on OpenRouter | 1.2% | ~13% (average; peaks near 30%) | ~61% | +983% | +369% |
| Generative AI users in China (CNNIC) | 249M | 602M | 700M+ | +142% | +16% |
| Doubao (ByteDance) monthly active users | ~60M | 157M | 419.08M | +162% | +167% |
| Generative AI services filed with China’s Cyberspace Administration (cumulative) | 302 | 748 | 1,112 | +148% | +49% |
Sources, dates and confidence for each value: State of AI Geography 2026.
Industries
- Paid business adoption on Ramp rose from 36.4% (Nov 2024) to 46.6% (Dec 2025) to 55.9% (Sep 2026); finance and insurance moved from 51.2% to 58.1% to 73.7% (Aug 2026), the steepest climb among large sectors.
- FDA AI device authorizations went from 235 (2024) to a record 335 (2025, contested range 295-335) to roughly 167-178 in January-June 2026; the cumulative list grew from 1,016 to 1,614 entries between December 2024 and September 2026.
- Practitioner adoption converged at a high level: physicians from 66% (Nov 2024) to 81% (early 2026), UK undergraduates from 66% to 92% to 95% across three academic years, legal professionals from 79% to 79% to 85% (Clio).
- The measures themselves changed: Census BTOS switched questions in November 2025, Pew widened its question from ChatGPT to all chatbots, Ramp expanded its sample in June 2026, and Gartner restated 2025 spending from $1.48T to about $1.79T, so several apparent jumps are partly definitional.
Bars: 2024, 2025, 2026 (to date, blue). Each scaled to its own maximum.
| Sector | Firms paying for AI (Ramp, Aug 2026) | Firms using AI (Census, May 2026) | Practitioner use (latest) | Spend signal | Measured outcome |
|---|---|---|---|---|---|
| Software & IT | 80.9% | 39.7% | ~83% of developers (2026, DERIVED) | Coding ~55% of departmental gen-AI spend (2025) | Experienced devs 19% slower (METR RCT, 2025) |
| Finance & insurance | 73.7% | 33.9% | 27% of employees use AI daily (May 2026) | Banking ~$31.3B AI spend (IDC, 2024) | Vendor-reported only (e.g. +20% fraud detection) |
| Manufacturing | 61.1% | n/a | 34% of operations AI-augmented (2026) | Not disclosed | No controlled evidence found |
| Retail | 48.8% | ~14% | n/a | ~$25B AI spend (IDC, 2024) | AI-influenced holiday sales $262B (attribution, 2025) |
| Healthcare | 43.4% | n/a | 81% of physicians (Jan-Feb 2026) | $1.5B vertical gen AI (Menlo, 2025) | Scribes save ~16 min/day (JAMA, 2026); +29% cancers detected (MASAI) |
| Education | n/a | n/a | 95% of UK undergraduates (2025-26) | n/a | Weekly teacher users save 5.9 h/week (self-reported, 2025) |
| Legal | n/a | n/a | 85% at firms with AI (Clio, 2026) | $650M vertical gen AI (Menlo, 2025) | 17-33% hallucination in research tools (Stanford, 2024) |
| Marketing & media | n/a | n/a | 22.4% of marketing activities (2026) | $360M creator tools (Menlo, 2025) | n/a |
| US federal government | n/a | n/a | 3,611 use cases (2025 inventory) | $5.6B AI contracts FY2022-24 (GAO) | n/a |
Notes: Read down the columns, not across: each column uses one source and its own definition. Ramp’s ‘technology’ sector is compared with Census’s ‘information’ sector. ‘Practitioner use’ rows come from different surveys with different populations and are not ranked. Outcome entries are single studies or company figures, not sector-wide estimates. Sources for figures not in the data file: Gallup daily use (May 2026); Rockwell (May 2026); Menlo coding share and creator tools (site page 376); MASAI (site page 863); Stanford (site page 472); GAO (site page 374).REPORTAI report-ai.org
| Measure | 2024 | 2025 | 2026 (to date) | 2024→25 | 2025→26 |
|---|---|---|---|---|---|
| US businesses paying for AI tools (Ramp AI Index) | 36.4% | 46.6% | 55.9% | +28% | +20% |
| Finance and insurance businesses paying for AI tools (Ramp) | 51.2% | 58.1% | 73.7% | +13% | +27% |
| Organizations using AI in at least one business function (McKinsey) | 78% | 88% | 89% | +13% | +1% |
| FDA AI-enabled medical device authorizations per year | 235 | 335 | ~167-178 (Jan-Jun) | +43% | |
| Cumulative FDA-authorized AI-enabled medical devices | 1,016 | 1,430 | 1,614 | +41% | +13% |
| UK full-time undergraduates using AI (HEPI) | 66% | 92% | 95% | +39% | +3% |
| UK undergraduates whose university provides generative-AI tools (HEPI) | 9% | 23% | 38% | +156% | +65% |
| Legal professionals using AI (Clio Legal Trends) | 79% | 79% | 85% | +0% | +8% |
| Developers currently using AI tools in their development process (Stack Overflow) | 62% | 78% | ~83% | +26% | +6% |
| Bank of America Erica cumulative client interactions | 2.5B+ | 3.2B+ | 3.4B+ | +28% | +6% |
| Share of marketing activities using generative AI (The CMO Survey) | 7.0% | 15.1% | 22.4% | +116% | +48% |
Sources, dates and confidence for each value: State of AI Industries 2026.
Risk
- The record deepfake-style fraud loss stood at about $25 million (Arup) at the end of 2024 and 2025, then rose to €95 million (~$108 million, Fideuram) in September 2026; each record case used less synthetic media than the last.
- Chatbot reliability on news worsened and partly recovered: NewsGuard measured false-claim repetition at 18% (August 2024), 35% (August 2025) and 28.8% (January 2026).
- Official measurement arrived only in the middle year: the FBI had no AI category for 2024, counted 22,364 AI-related complaints and $893 million in losses for 2025, and will not report 2026 until April 2027 — so 2026 comparisons rely on vendor, watchdog and case data.
- Chatbot-harm litigation against OpenAI went from no ChatGPT injury or death suits at the end of 2024 to eight filings in 2025 and roughly two dozen coordinated cases by August 2026, alongside the first state attorney-general suit naming a CEO personally.
Bars: 2024, 2025, 2026 (to date, blue). Each scaled to its own maximum.
| Indicator | 2024 | 2025 | 2026 (to date) |
|---|---|---|---|
| AI incidents (AIID) | 233 | 362 | — |
| FBI AI-linked losses | — | $893M | — (Apr 2027) |
| Record single loss | ~$25M (Arup) | ~$25M | €95M (~$108M) |
| Phishing w/ AI (KnowBe4) | 80% | 84% | 86% |
| Chatbot false claims | 18% (Aug) | 35% (Aug) | 28.8% (Jan) |
| AI news sites (NewsGuard) | 1,121 (Nov) | 1,271 (May) | 3,749 (Jun) |
| AI share new articles | ~48% | 50.9% (Q4) | 49.9% (Q1) |
| IWF AI abuse videos | 13 | 3,443 | — |
| Agent-run intrusions | 0 | 1 | 1 |
| ChatGPT harm suits | 0 | 8 | ~24 (Aug) |
| WEF misinfo rank | #1 | #1 | #2 |
Notes: Each row compares like with like within its source. ‘—’ = no comparable value. KnowBe4 row is CONTESTED (80% and 84% are KnowBe4’s restated comparisons; the widely quoted 82.6% covers September 2024–February 2025). Graphite 2024 is approximate. Agent-intrusion and suit counts are DERIVED tallies. Deepfake-tracker counts are excluded here because attribution and coverage changed between years.REPORTAI report-ai.org
| Measure | 2024 | 2025 | 2026 (to date) | 2024→25 | 2025→26 |
|---|---|---|---|---|---|
| Largest publicly reported single deepfake / AI-impersonation fraud loss (record at year end) | ~$25M (Arup, HK$200M) | ~$25M (Arup record unchanged) | €95M (~$108M, Fideuram) | +0% | +322% |
| Deepfake incidents/attacks logged by public deepfake-incident trackers | 150 | >1,500 | 821 (H1) | +900% | -45% |
| Phishing showing any AI involvement (KnowBe4) | 80% | 82.6% | 86% | +3% | +4% |
| Rate at which leading chatbots repeat false news claims (NewsGuard audits) | 18% | 35% | 28.8% | +94% | -18% |
| AI content-farm ‘news’ sites tracked by NewsGuard | 1,121 | 1,271 | 3,749 | +13% | +195% |
| Share of new web articles written primarily by AI (Graphite) | ~48% | 50.9% | 49.9% | +6% | -2% |
| Rank of mis- and disinformation among short-term (2-year) global risks, WEF Global Risks Report | #1 | #1 | #2 | +0% | +100% |
| Publicly documented intrusions in which AI agents did most of the operational work | 0 | 1 | 1 | +0% | |
| Personal-injury and wrongful-death lawsuits against OpenAI over ChatGPT (cumulative) | 0 | 8 | ~24 | +200% | |
| Best overall grade any major AI developer received, FLI AI Safety Index | C (Anthropic) | C+ (Anthropic) | C+ (Anthropic, 2.66/4) |
Sources, dates and confidence for each value: State of AI Risk 2026.
Regulation
- US state AI lawmaking roughly doubled each year on MultiState’s count from 2024 to 2025 (635 to 1,208 bills introduced; 99 to 145 enacted), and NCSL’s 2026 year-to-date enactments (280) already exceed its 2025 total (227). Different trackers give different totals, so compare within one tracker only.
- EU AI Act deadlines moved later each year: in 2024 stand-alone high-risk duties were due on 2 August 2026, a delay to December 2027 was proposed in 2025, and the Omnibus made it law in 2026.
- US federal policy went from agency rule-making in 2024 (59 regulations) to executive orders in 2025 (eight by our count, ending with EO 14365 against state AI laws) and to proposals in 2026 (EO 14409, the Great American AI Act draft and the AI Kill Switch Act), with no comprehensive statute in any of the three years.
- China’s filed generative-AI services went from 302 at the end of 2024 to 748 at the end of 2025 and 1,112 by 31 August 2026. Its binding AI-specific rules went from three to four, with labelling added in September 2025.
Bars: 2024, 2025, 2026 (to date, blue). Each scaled to its own maximum.
| Dimension | European Union | United States | China |
|---|---|---|---|
| Main instrument | AI Act (Reg. 2024/1689) | No federal statute; state laws + EOs | CAC administrative rules (4) |
| Approach | Risk tiers, rights-based | Market-led, deregulatory | State control of content |
| Maximum penalty | €35M or 7% of turnover | No federal cap; TX $200k/violation | Takedown, suspension |
| Frontier-model trigger | 10^25 FLOP (GPAI) | CA SB 53 (10^26 FLOP) | Filing + security review |
| AI content labelling | Art. 50, from Aug 2026 | State laws; TAKE IT DOWN | Mandatory since Sep 2025 |
| 2026 direction | Delay (Omnibus) | Preemption push vs state growth | More filings; no AI law |
| Enforcement record | 0 AI Act fines (Sep 2026) | FTC, FCC, state AG suits | 1,112 services filed (Aug 2026) |
Notes: Penalty figures are statutory maximums, not fines imposed. US state penalties vary (Texas TRAIGA up to $200,000 per violation; Utah $2,500, page 1101). China’s sanctions run through existing cybersecurity and content law. EU, US and China counts are as of the dates shown.REPORTAI report-ai.org
| Measure | 2024 | 2025 | 2026 (to date) | 2024→25 | 2025→26 |
|---|---|---|---|---|---|
| EU AI Act: obligations applicable by year end | In force 1 Aug 2024; no obligations applicable yet | Prohibitions + AI literacy (2 Feb 2025); GPAI-model duties + governance (2 Aug 2025) | Omnibus Reg. (EU) 2026/1744 in force 27 Jul 2026; Art. 50 transparency and AI Office GPAI enforcement from 2 Aug 2026; stand-alone high-risk moved to 2 Dec 2027; two new Art. 5 bans from 2 Dec 2026 | ||
| EU AI Act maximum fine (prohibited practices) | €35M or 7% of turnover (enacted; not yet applicable) | €35M or 7% of turnover (applicable from 2 Aug 2025) | €35M or 7% of turnover (unchanged by the Omnibus) | +0% | +0% |
| EU AI Act stand-alone high-risk (Annex III) application date, as scheduled at year end | 2 Aug 2026 | 2 Aug 2026 (delay to Dec 2027 proposed 19 Nov 2025) | 2 Dec 2027 (product-embedded high-risk: 2 Aug 2028) | ||
| Fines published under the EU AI Act | 0 | 0 | 0 (to date) | ||
| US state AI-related bills introduced (MultiState count) | 635 | 1,208 | 1,561 (to date) | +90% | +29% |
| US states with election-deepfake laws (Public Citizen tracker) | 21 | 25+ | 31 (to date) | +19% | +24% |
| Colorado AI Act effective date, as scheduled at year end | 1 Feb 2026 (SB 24-205, signed May 2024) | 30 Jun 2026 (delayed) | Repealed and replaced by SB 26-189 (signed 14 May 2026), effective 1 Jan 2027 | ||
| AI-focused US presidential executive orders signed | 0 | 8 | 1+ (EO 14409, 2 Jun 2026) | -88% | |
| US federal preemption of state AI law: status at year end | None proposed in law; no federal AI statute | 10-year state-law moratorium struck from budget bill 99–1 (Jul 2025); EO 14365 (11 Dec 2025) orders DOJ litigation task force and BEAD funding conditions | No preemption statute; DOJ AI Litigation Task Force set up 9 Jan 2026; House ‘Great American AI Act’ (3-year preemption) still a discussion draft | ||
| US federal AI enforcement: headline action of the year | FCC $6M forfeiture over AI-cloned Biden robocall | TAKE IT DOWN Act signed 19 May 2025 (federal crime to publish nonconsensual intimate deepfakes) | FTC enforcing TAKE IT DOWN 48-hour removal duty from 19 May 2026; civil penalty $53,088 per violation | ||
| China: generative-AI services filed with the CAC (cumulative) | 302 | 748 | 1,112 (to date) | +148% | +49% |
| China: generative-AI apps/features registered (cumulative) | 105 | 435 | 731 (to date) | +314% | +68% |
| China: binding national AI-specific administrative rules in force | 3 | 4 | 4; no national AI law | +33% | +0% |
| AI-specific national statutes in force among the nine jurisdictions Report AI maps | 1 (EU) | 2 (EU, Japan) | 3 (EU, Japan, South Korea) | +100% | +50% |
| United Kingdom: AI regulatory approach at year end | No AI statute; principles-based, sector regulators; AI Safety Institute | No AI statute; institute renamed AI Security Institute (Feb 2025); UK did not sign Paris summit statement | No AI statute; frontier-AI risk package incl. AI incident reporting under the Cyber Security and Resilience Bill (12 Oct 2026); £700m sovereign-AI fund (Budget, 28 Oct 2026) | ||
| Largest EU data-protection fine on a generative-AI chatbot provider | €15M (Italy Garante v. OpenAI) | €5M (Italy Garante v. Luka/Replika) | €158,000 (Italy Garante v. Character.AI) | -67% | -97% |
Sources, dates and confidence for each value: State of AI Regulation 2026.
Frequently asked
What grew fastest in AI between 2024 and 2026?
Token volume: Google’s monthly token processing rose about 330-fold, from about 9.7 trillion in spring 2024 to more than 3.2 quadrillion in May 2026. Combined OpenAI and Anthropic run-rate revenue rose about 18-fold over a similar period.
Did AI adoption keep growing in 2026?
Organisational adoption plateaued: McKinsey found 78% of organisations using AI in 2024, 88% in 2025 and 89% in 2026. Individual use kept growing: Gallup found the share of US employees using AI a few times a week or more rose from 12% to 23% to 30%.
Are AI prices falling?
At fixed capability, yes: the cheapest model of GPT-4 quality cost about $0.48 per million tokens in 2024 and about $0.20 in 2026 on a 3:1 input:output blend. At the frontier, list prices barely moved: OpenAI’s flagship bought about 229,000 tokens per dollar in 2024 and about 250,000 in 2026 (Report AI calculations).
Are AI job cuts rising?
Annual totals rose from 12,742 in 2024 to 54,836 in 2025 and 121,401 in January to October 2026, according to Challenger, Gray & Christmas. The monthly count fell to 1,265 in October 2026, the lowest of the year.
How are the year-over-year changes calculated?
Report AI computes each change only where both years have a comparable value on the same definition; these are marked DERIVED. 2026 values are year to date or the latest reading, so a change into 2026 compares a part year with a full year where noted.
Figures are compiled from company disclosures, regulators, research firms and reputable reporting, and cross-checked against Report AI’s own index pages. Each value carries its as-of date, source and a confidence rating: HIGH where the primary source was read, MEDIUM where it rests on consistent reputable secondary reporting (most figures, because many primary pages cannot be opened from our research environment), DERIVED where it is our own arithmetic, and CONTESTED where credible sources disagree. A year means the calendar year the figure describes; 2026 values are the latest available and marked “to date”. Weekly and monthly user counts, and forecasts and outcomes, are never compared as if they were the same measure. Percentage changes are Report AI arithmetic (DERIVED) and are shown only where both years have a comparable numeric value; 2026 values are to date, so changes into 2026 compare a part-year or latest reading with a full year where noted. These reports compile and verify secondary data with original analysis; they are not original surveys. See our editorial standards, corrections log and fact-check register.
State of AI Usage · State of AI Spending · State of AI Finance · State of AI Work · State of AI Models & Prices · State of AI Tokens · State of AI Infrastructure & Energy · State of AI Geography · State of AI Industries · State of AI Risk · State of AI Regulation · Year over year · AI Market Pulse · AI statistics that are wrong