AI Trends 2027
By Axis Intelligence Research
Co-author: Sarah Mitchell | Last updated: November 2, 2026 | License: CC BY 4.0
Axis Intelligence Research projects about 4.0 billion monthly users of generative AI platforms by October 2027, up roughly 23% from 3.24 billion in October 2026. Growth is slowing from last year’s 57.2%, and the reason is not demand: it is that the most-measured markets are filling up while the next billion users sit in markets where AI arrives on a phone before it arrives on a desk.
Quick Answer
Axis Intelligence Research’s base case puts monthly generative AI users at about 4.0 billion in October 2027 (range 3.75–4.23 billion), measured in the same scope as DataReportal’s 3.24 billion for October 2026. No other firm publishes a 2027 user forecast. The variable that moves the number most is how fast emerging markets convert internet users into AI users. On the technology side, the 2027 lists from Gartner, Forrester and IDC agree on eight shifts: world models, speech reasoning, governed agents, cheaper open inference, sovereign AI, a spending squeeze, content provenance and a split in regulation. For 2027 itself, voice and on-device distribution are the ones most likely to change how many people use AI.
Key Findings
- Axis Intelligence Research projects about 4.0 billion monthly generative AI users worldwide in October 2027, a 23.5% rise from DataReportal’s 3.24 billion in October 2026.
- Global counts of AI users disagree by about 4x, according to an Axis Intelligence Research calculation: roughly 1.0 billion working-age users in Microsoft’s telemetry versus more than 4.1 billion people aged 16+ using any AI in DataReportal’s estimate.
- The two main global estimates disagree on growth, not just level: Menlo Ventures puts 2026 user growth at 11%, DataReportal at 57.2%.
- Anthropic’s revenue for the first half of 2026 ($16.29 billion) was 3.55 times its revenue for all of 2025, according to its October 26, 2026 Form S-1, an Axis Intelligence Research calculation.
- 64% of US teens aged 13 to 17 use AI chatbots, per Pew Research Center, and California’s SB 1119 imposes most of its child-safety duties on companion chatbots from July 1, 2027.
- Only three of Gartner’s ten technology trends for 2027 carry a forecast in its October 20, 2026 announcement, and all three are about control rather than new capability: up to 80% higher agentic AI accuracy from semantic data by 2027, AI agent abuse behind 25% of enterprise security breaches by 2028, and digital sovereignty strategies at more than 75% of non-US organizations by 2030 (an Axis Intelligence Research count).
Which AI Technologies Will Shape 2027?
“AI trends 2027” has two meanings. One is technological: what the systems will be able to do. The other is economic: how many people will use them and who will be paid. Most lists answer only the first. This page leads with the second because it is measurable, but the two are linked: a new capability only becomes a trend once it reaches users. Below are the eight technology shifts that analyst firms name for 2027, cross-checked against what is already measurable.
Gartner’s Top Strategic Technology Trends for 2027, announced on October 20, 2026 by Gene Alvarez, Distinguished VP Analyst, lists ten trends that it says will drive disruption “within the next five years.” In the Symposium presentation, as reported by Network World, Gartner sorted them into three horizons: “leverage the nontextual AI frontier” (one to three years), “govern the autonomous agent platform” (three to five years) and “architect durable agility” (five years and more). Forrester’s Predictions 2027 (October 21, 2026) and IDC’s FutureScape 2027 (October 27–29, 2026) cover the same ground from the spending and deployment side.
| # | Shift | Who names it | Published prediction | Already measurable | Go deeper |
|---|---|---|---|---|---|
| 1 | World models and quantitative models: AI that simulates space, time and cause and effect, or that is trained on numbers rather than text | Gartner (world models; large quantitative models) | 40% of robotics applications will use world models by 2030, from less than 1% in 2025; quantitative models will replace LLMs in 25% of strategic forecasting and operational modeling tasks by 2030 (Symposium, via Network World) | AMI Labs raised a $1.03 billion seed round at a $3.5 billion valuation (March 10, 2026); World Labs raised $1 billion (February 18, 2026) | This page |
| 2 | Speech reasoning models: voice AI that reasons on audio directly instead of transcribing it first | Gartner | More than 70% of enterprise voice interactions handled autonomously by speech models by 2030, up from less than 10% in 2026 (Symposium, via Network World) | — | This page |
| 3 | Governed agents: context platforms, agent identity and closed-loop coding harnesses | Gartner, Forrester, IDC | Gartner: organizations that prioritize semantics in AI-ready data will raise agentic AI accuracy by up to 80% and cut costs by up to 60% by 2027; agent abuse behind 25% of enterprise security breaches by 2028; autonomous coding agents generating 20% of all public commits by late 2026 (Symposium, via Network World). Forrester: agentic buying will play a role in 75% of B2B purchases | IDC: enterprises expected to go from 23.5 to 43.2 types of AI agents in production between April and the end of 2026 | AI agents trends 2027, AI cybersecurity trends 2027 |
| 4 | Cheaper, more open inference: open-weight models and alternatives to Nvidia | Forrester | A quarter of AI infrastructure RFPs will require non-Nvidia accelerator options | Open-weight models: 75% of OpenRouter tokens (Microsoft); Gemini API tokens 3.2x across three quarterly disclosures (Alphabet) | AI energy consumption forecast |
| 5 | Sovereign AI: models, data and infrastructure that give a credible exit from a foreign provider | Gartner, Forrester | More than 75% of non-US organizations will adopt digital sovereignty strategies by 2030 (Gartner) | 147-economy adoption map below | AI adoption by country |
| 6 | The spending squeeze: budgets rise while every project must show a return | Forrester | AI will account for 20% of business technology spending in 2027, within $6.1 trillion of global technology spending; rising AI costs will force B2C leaders to scale back a third of AI initiatives | Anthropic’s compute bill fell from 2.01x to 0.55x revenue (S-1) | Revenue section below |
| 7 | Content provenance: proving where text, images and video come from | Forrester, EU AI Act | A third of consumers will reject content that cannot prove its source (Forrester) | EU AI Act transparency duties apply from August 2, 2026; watermarking grace period ends December 2, 2026 | AI scam trends 2027 |
| 8 | Regulation splits: Europe delays, the US preempts, minors become the exception everyone agrees on | — | EU high-risk rules for stand-alone systems moved to December 2, 2027 | EU Digital Omnibus on AI in force since July 27, 2026; US Executive Order 14365 (December 11, 2025) | AI regulation tracker, minors section below |
Sources: Gartner press release (trend names and the agent-accuracy, agent-identity and sovereignty predictions); Gartner predictions marked “via Network World” were presented at the Symposium and do not appear in the press release; Forrester and Forrester Predictions 2027; IDC; AMI Labs funding as reported by Silicon Republic; World Labs funding as reported by Reuters; EUR-Lex, Regulation (EU) 2026/1744. The other three Gartner trends (autonomous exposure remediation, disposable applications, wetware as a service) are mainly enterprise or long-horizon and are covered by our security and future-of-work outlooks.
What the trend lists leave out
Every list above is written for enterprise buyers. Read against the user data on this page, three things stand out.
The near-term shift is voice and distribution, not world models. Gartner’s own horizon puts world models in the one-to-three-year band, but its benchmark year is 2030 and its base is less than 1% of robotics applications. For 2027, the capability that reaches the most people is the one already shipped inside phones and messaging apps. The Gemini app grew from 750 million to more than 1 billion monthly users in nine months largely because it sits on Android.
Agents are an enterprise story with a consumer test. IDC expects enterprises to nearly double the types of agents they run in production in 2026, and Forrester expects agentic buying to touch 75% of B2B purchases. On the consumer side, the test is advertising: ChatGPT ads already run in 89 countries, and an agent that shops for you decides which ads ever get seen.
Regulation is diverging by audience. The EU’s Digital Omnibus on AI (Regulation (EU) 2026/1744, adopted July 8, 2026) postponed high-risk obligations for stand-alone systems from August 2, 2026 to December 2, 2027, and for AI in regulated products to August 2, 2028, as summarized by Orrick and Gibson Dunn. In the US, Executive Order 14365 created an AI Litigation Task Force to challenge state AI laws, and the White House’s National Policy Framework of March 20, 2026 asks Congress to preempt state AI laws while leaving child-protection laws to the states, as summarized by Ropes & Gray. Both moves loosen rules for general AI while rules for minors tighten, which is why the under-18 section below has more 2027 dates than any other.
Sarah Mitchell’s read: Treat each trend list as a statement about where vendors will put their sales effort. The test for 2027 is simpler: does a capability change how many people use AI, or how much they pay? Voice and on-device distribution pass that test next year. World models will pass it later, if they pass it at all.
Source: Axis Intelligence Research, AI Trends 2027. Data: download the CSV (CC BY 4.0).
How Many People Will Use AI in 2027?
About 4.0 billion monthly users of generative AI platforms by October 2027, in Axis Intelligence Research’s base case. That is a projection, built on DataReportal’s October 2026 count of 3.24 billion monthly active users of the top LLM platforms, equal to 38.9% of the world’s population.
DataReportal itself warns that this figure may not represent unique individuals, because one person can run a personal and a work account. Read it the way you read “social media user identities”: a consistent series that tracks direction well, not a census of humans.
The Axis Base Case: how the 4.0 billion is built
Most “AI users” projections apply a growth rate to a headline total. That breaks quickly, because a 57.2% growth rate cannot hold against a ceiling. The Axis Base Case models AI users as a share of internet users that closes part of its remaining gap to a ceiling each year:
GenAI users 2027 = internet users 2027 × [share 2026 + k × (ceiling − share 2026)]
- Share of internet users using GenAI platforms. 33.8% in October 2025 (implied: 2.061 billion GenAI users ÷ 6.096 billion internet users) and 52.7% in October 2026 (3.24 billion ÷ 6.15 billion). Both are Axis Intelligence Research calculations from DataReportal figures.
- Ceiling = 83.5%. DataReportal’s partner Manochi estimates that 83.5% of internet users aged 16 and above use some form of AI each month, including search summaries and AI inside office software. Standalone GenAI use should not exceed overall AI use, so this is the cap.
- k = 0.38. Over the past year the GenAI share closed 38% of its gap to the ceiling: (52.7 − 33.8) ÷ (83.5 − 33.8) = 0.380. The base case assumes the same pace.
- Internet users 2027 = 6.21 billion. An Axis assumption of +1.0% on DataReportal’s 6.15 billion. DataReportal reports growth of 54 million users in the past year and says delays in reporting probably understate the true figure.
Base case: 6.21 billion × [52.7% + 0.38 × (83.5% − 52.7%)] = 6.21 billion × 64.4% = about 4.0 billion.
Base, upside and downside scenarios for October 2027
| Scenario | Monthly GenAI users (Axis projection) | Share of internet users | Gap closed (k) | What separates it |
|---|---|---|---|---|
| Downside | about 3.75 billion | 60.4% | 0.25 | US and European adoption flat; Global South growth slows as the North–South gap keeps widening |
| Base | about 4.0 billion | 64.4% | 0.38 | Last year’s conversion pace holds |
| Upside | about 4.23 billion | 68.1% | 0.50 | Assistants built into Android, WhatsApp and search convert passive users into standalone GenAI users |
Source: Axis Intelligence Research calculation using DataReportal (Digital 2027, October 21, 2026). Every 0.05 change in k moves the 2027 figure by about 96 million users; a half-point change in internet-user growth moves it by about 20 million. The conversion pace, not connectivity, carries the forecast.
Sarah Mitchell’s read: The base case is deliberately boring. It assumes 2027 converts users at the rate 2026 did, against a cap that the survey data already shows. What would make it wrong is a definitional shift, not a demand shock. If DataReportal starts counting assistants that ship inside an operating system, the series jumps without a single new person picking up a chatbot.
Why Do Estimates of AI Users Differ So Much?
Because they count different things. Four credible global sources, all published in 2026, put the number of AI users between about 1.0 billion and more than 4.1 billion. That spread of roughly 4x is an Axis Intelligence Research calculation, and it is almost entirely scope.
| Source | Latest value | What it counts | As of | Method |
|---|---|---|---|---|
| Microsoft AI Economy Institute | 18.8% of people aged 15–64 (about 1.0 billion, Axis calculation) | Working-age people who used a generative AI product in the period | June 2026 | Anonymized Microsoft telemetry, adjusted for device share, internet access and population |
| Menlo Ventures | 2.0 billion | People using AI tools | 2026 | US survey of 5,067 adults, extrapolated globally |
| DataReportal (Kepios/Manochi) | 3.24 billion | Monthly active users of top LLM platforms; may include duplicate accounts | October 2026 | Platform data plus survey cross-checks |
| DataReportal (Manochi, using GWI) | More than 4.1 billion | People aged 16+ using any form of AI monthly, including AI Overviews and embedded tools | October 2026 | GWI survey of online adults, extrapolated |
The 1.0 billion figure applies Microsoft’s 18.8% to the World Bank’s 2025 count of 5,355,513,094 people aged 15 to 64 (UN World Population Prospects). Microsoft does not publish a headcount; the conversion is ours.
Which way is AI use growing? The growth rates disagree too
| Series | Change | Period |
|---|---|---|
| DataReportal monthly GenAI users | +57.2% | Oct 2025 → Oct 2026 |
| Menlo Ventures global AI users | +11% (1.8 billion → 2.0 billion) | 2025 → 2026 |
| Microsoft working-age share | 16.3% → 18.8% | Q4 2025 → Q2 2026 |
| CNNIC, China | +36% or more (515 million → more than 700 million) | June 2025 → first half of 2026 |
| ChatGPT weekly users (company-reported) | +33% (900 million → 1.2 billion) | Feb 27 → Sep 29, 2026 |
| Gemini app monthly users (company-reported) | +33% (750 million → 1 billion+) | Feb 4 → Oct 28, 2026 |
Sources: DataReportal; Menlo Ventures; Microsoft; CNNIC via ECNS and The Star; OpenAI; Alphabet. Growth rates for China, ChatGPT and Gemini are Axis Intelligence Research calculations from the reported endpoints.
Menlo’s 11% and DataReportal’s 57.2% can both be right. Menlo’s global figure is extrapolated from American behavior, and American adoption is close to its plateau. DataReportal’s figure is built from platform totals, which grow fastest where new phones and new accounts are arriving. We will not average them: one is a survey of people, the other a count of accounts, and a blend would describe neither.
The track record argues for caution in both directions. DataReportal’s April 2026 count was 2.42 billion; six months later it was 3.24 billion, a 34% jump in half a year that coincides with Similarweb introducing a new methodology for its app data. A year earlier, DataReportal would commit only to “more than 1 billion” people using LLMs each month.
Which AI Chatbot Has the Most Users in 2027?
ChatGPT still leads, but its share of the audience is falling. OpenAI’s DevDay recap of September 29, 2026 describes ChatGPT as reaching “our collective 1.2B weekly users.” Alphabet said on October 28, 2026 that the Gemini app has more than 1 billion monthly active users. Each company reports a different metric, so the two numbers cannot be ranked against each other directly.
| Platform | Latest reported figure | Metric | As of | Source type |
|---|---|---|---|---|
| ChatGPT | 1.2 billion | Collective weekly users | Sep 29, 2026 | Company-reported (OpenAI DevDay recap) |
| Gemini app | 1 billion+ | Monthly active users | Oct 28, 2026 | Company-reported (Alphabet) |
| Meta AI | More than 1 billion | Monthly actives | Oct 29, 2025 [older data] | Company-reported on Meta’s Q3 2025 call; Meta’s Q3 2026 earnings release (Oct 21, 2026) gives no Meta AI user count |
| Microsoft 365 Copilot | More than 30 million | Paid seats | Fiscal 2026 | Microsoft proxy statement |
| GitHub Copilot | 50 million | Users | Fiscal 2026 | Microsoft proxy statement |
| Claude | Not disclosed | — | — | Anthropic publishes no user count |
| All GenAI apps, China | More than 700 million | Users | First half of 2026 | CNNIC (via ECNS) |
Independent measurement points the same way. Sensor Tower estimates that ChatGPT’s True Audience share, which counts unique users across apps and the web, fell below 50% for the first time in March 2026. By May 2026, ChatGPT held 46% worldwide, Google Gemini 28% and Claude 10%, across 25 markets, with people who use several assistants counted once for each. Similarweb data in DataReportal’s report shows Gemini’s web audience at more than 60% of ChatGPT’s in August 2026, while ChatGPT’s mobile app still had more than three times Gemini’s monthly users outside China.
Claude is the fastest-growing challenger in both datasets. Sensor Tower estimates Claude’s True Audience rose 452% year on year in May 2026 and its US share went from 4.4% to nearly 14%; it was still 5% in December 2025, so most of that gain came in five months. Similarweb’s app data shows Claude’s monthly users in August 2026 at 17 times the level of August 2025, from a much smaller base.
Underneath the chatbot brands, the model layer is opening up. Microsoft’s Q2 2026 AI Diffusion Report finds that open-weight models now account for 75% of token usage on OpenRouter, up from roughly one-third in late 2025. Microsoft cautions that OpenRouter users experiment across providers more than most, so this is not a global market share; it is a signal that developers price-shop models the way cloud buyers price-shop instances.
Sarah Mitchell’s read: The leaderboard question is the wrong one for 2027. Gemini’s growth comes from distribution: it sits on Android and inside Google’s services, so its monthly figure includes people who never chose a chatbot. ChatGPT’s weekly figure measures a habit. When a company reports weekly users, it is telling you retention is its strong suit; when it reports monthly, reach is. Watch the metric each company picks, because it changes when the story stops flattering them.
How Much Time Do People Spend Using AI?
Less than headlines suggest. Sensor Tower estimates ChatGPT users spent an average of 215 minutes a month in the app in Q1 2026 (iOS and Android phones, worldwide), roughly 7.1 minutes a day by Axis Intelligence Research’s calculation (215 ÷ 30.4), and notes that ChatGPT engagement softened in each of the past three quarters. The challengers are deepening instead: Claude rose from 40 to 120 minutes a month and Gemini from 14 to 100 between Q1 2025 and Q1 2026.
Total time is rising quickly anyway, because the user base is. Sensor Tower projects 36 billion hours spent in generative AI apps in the first half of 2026 (iOS and Google Play, projected through the end of June), up from 17.2 billion hours a year earlier, about 2.1 times as much, and 965 billion sessions. The web adds its own layer: generative AI websites drew more than 67 billion visits and nearly 23 billion hours in Q1 2026, with visits up 28% and time up 41% year on year, per Sensor Tower Web Insights (56 markets).
For context, Similarweb data puts TikTok’s average at 1 hour 55 minutes per user per day in August 2026. We do not compute a ratio between the two: the figures come from different firms with different panels. The direction is plain enough without one. AI is wide and shallow; short video is narrower and deep.
Depth varies sharply between users. In Menlo Ventures’ July 2026 survey of US adults, 29% of AI users spend less than 15 minutes a day with AI, while 10% spend more than two hours. Daily use among Americans rose from 19% in 2025 to 25% in 2026.
What do people use AI for?
Finding information is the top reason worldwide, cited by 60.4% of active AI users in GWI’s Q2 2026 survey reported by DataReportal. Getting advice on problems follows at 40.6%, and learning or improving skills at 38.4%.
Shopping research is the fastest-moving use worth watching for 2027. The share of online adults using AI to research brands rose from 24.4% in Q1 2026 to 26.0% in Q2, per GWI. Search engines still lead that stage, used by 45.3%.
Which Generation Uses AI the Most?
The youngest adults, by a wide margin, in every dataset. Across 53 countries, 93.3% of online adults aged 16 to 24 used AI in the past month, against 52.2% of those 65 and older, according to GWI data in DataReportal’s report.
| Age group (US adults) | Use AI chatbots |
|---|---|
| 18–29 | 66% |
| 30–49 | 61% |
| 50–64 | 42% |
| 65+ | 23% |
| All US adults | 49% |
Source: Pew Research Center, survey conducted February 17–23, 2026. Separately, 44% of US adults say they use ChatGPT specifically.
Heavy use does not mean high trust. Only 34.0% of internet users aged 16 to 24 believe AI benefits society, per GWI, though that is twice the 16.8% among those 65 and over. The generation most fluent with these tools is also the most likely to hold them at arm’s length.
Teenagers sit ahead of most adults. Pew Research Center found that 64% of US teens aged 13 to 17 use AI chatbots, with about three in ten using them daily, in a survey fielded September 25 to October 9, 2025. Only 51% of their parents think their teen uses chatbots.
Which Countries Use AI the Most?
The United Arab Emirates, at 73.3% of its working-age population in June 2026, followed by Singapore at 64.3%, according to Microsoft’s Global AI Diffusion Report. The global average is 18.8%.
| Economy or group | Working-age population using GenAI | Change noted |
|---|---|---|
| United Arab Emirates | 73.3% | Rank 1 |
| Singapore | 64.3% | Rank 2 |
| South Korea | 40.6% | Rank 12, up from 16; +3.5 points, largest absolute gain |
| United States | 33.0% | Rank 21; up from 31.3% in Q1 |
| Saudi Arabia | 31.4% | Rank 25, up from 30 |
| Global North | 28.8% | — |
| Japan | 24.7% (Axis calculation) | +2.2 points from 22.5% in Q1, largest relative gain |
| Brazil | 20.2% | — |
| Global average | 18.8% | Up from 17.8% in Q1 |
| India | 18.5% | — |
| China | 17.5% | Up from 16.4% in Q1; Microsoft says its measure misses widely used domestic tools |
| Global South | 16.2% | — |
| Indonesia | 14.8% | — |
| Nigeria | 10.4% | — |
| Kenya | 8.9% | — |
Source: Microsoft AI Economy Institute, Global AI Diffusion Report Q2 2026, published September 21, 2026, data for June 2026. The dataset carries Microsoft’s full table for all 147 economies, with Q1 and Q2 2026 readings.
The gap between the Global North and South stood at 12.6 percentage points in June 2026, up from 12.1 in Q1 2026 and 10.6 in the second half of 2025, per Microsoft. The South is not using AI less seriously: in Microsoft’s consumer Copilot data, self-directed learning takes a 29.2% larger share of conversations in the Global South than globally. That is the single biggest risk to the 2027 base case: the users still to come are disproportionately in the South.
Kenya shows the measurement problem at country scale. Microsoft puts generative AI use at 8.9% of Kenya’s working-age population; GWI finds that 97.3% of Kenya’s online adults used some form of AI in the past month, the highest rate in its 53-country survey. Nigeria reads 10.4% in Microsoft’s data and 96.6% in GWI’s. Both can be true: GWI’s denominator is internet users aged 16 and above, and internet penetration is 40.6% in Kenya and 47.5% in Nigeria. In Africa’s connected cities AI is already mainstream; the constraint is the connection. Ghana, at 35.8%, is the lowest in GWI’s survey.
China is the large market Microsoft’s telemetry under-sees: 17.5% in Microsoft’s measure, against more than 700 million generative AI users and penetration above 50% in the first half of 2026 according to CNNIC. CNNIC finds that 76.0% of Chinese users ask AI to answer questions. Microsoft says its next report will add tools currently missing, with the largest effect in China, where several widely used domestic tools are not yet covered, and that the increase will reflect better measurement rather than a sudden change in adoption. For anyone tracking the 18.8% series, that means a break in the line in 2027.
Traffic tells a different story from adoption rates. India was the largest market for generative AI websites in Q1 2026, with more than 13 billion visits, ahead of the United States with over 8 billion, according to Sensor Tower Web Insights. Supply is far more concentrated than demand: US publishers took nearly 60% of global generative AI app downloads and 85% of in-app purchase revenue between June 2025 and May 2026, per Sensor Tower.
How Much Money Will AI Companies Make in 2027?
The 2027 question is no longer whether people use AI but who pays for it. Three revenue engines are now visible in primary disclosures: enterprise consumption, consumer subscriptions and advertising.
Enterprise consumption, measured by an S-1. Anthropic’s Form S-1, filed October 26, 2026, gives the first audited revenue history for a frontier AI lab.
| Anthropic, PBC | Revenue | Compute and infrastructure expense |
|---|---|---|
| FY2023 | $41 million | — |
| FY2024 | $386 million | — |
| FY2025 | $4,585 million | — |
| H1 2025 | $1,163 million | $2,334 million |
| H1 2026 | $16,286 million | $8,945 million |
Source: Anthropic, PBC Form S-1, filed October 26, 2026. Annual figures audited; half-year figures unaudited.
Two Axis Intelligence Research calculations from those rows matter for 2027. First-half 2026 revenue was 14.0 times the first half of 2025. And the compute bill fell from 2.01 times revenue in H1 2025 to 0.55 times revenue in H1 2026. Inference at scale is starting to cost less than it earns. The S-1 also reports net revenue retention of approximately 900% as of September 30, 2026, which means existing customers’ spending, excluding consumer Pro and Max subscribers, grew about ninefold in a year. More than 11,900 customers paid $100,000 or more over the trailing twelve months.
Usage intensity is rising across the industry, not only at one lab. Alphabet’s first-party model APIs processed more than 10 billion tokens per minute at the start of 2026, more than 16 billion in Q1, 22 billion in Q2 and nearly 32 billion by its October 28 report, a 3.2x rise across three quarterly disclosures (Axis calculation). Google Cloud revenue grew 103% year on year in Q3 2026.
Consumer subscriptions. OpenAI reported more than 50 million consumer subscribers and more than 9 million paying business users on February 27, 2026. Menlo Ventures estimates global consumer AI spending at $40 billion in 2026, up from $12 billion in 2025, and finds that 55% of US AI users pay for at least one AI product.
Advertising. OpenAI says ChatGPT Ads reached $1 billion in annualized revenue run rate less than 200 days after launch, and that more than 100,000 advertisers have run ads since then (October 21, 2026). The same day, OpenAI extended ads to 89 countries, as reported by PPC Land. In a US Android sample, Sensor Tower estimates both the number of ChatGPT ads and the number of users seeing them rose roughly 7x between the first week of March and the end of May 2026. Willingness to pay is also moving: Sensor Tower estimates Claude’s average monthly revenue per US app user rose from less than $0.50 in September 2025 to $2.76 in May 2026. OpenAI’s CFO said the company passed $20 billion in annualized revenue at the end of 2025, as reported by Reuters; annualized revenue is a run-rate, not reported annual revenue.
Sarah Mitchell’s read: The cost line is the one to hold onto. A year ago, Anthropic spent two dollars on compute for every dollar it booked; by the first half of 2026 it spent 55 cents. That is the gap between a research lab subsidizing usage and a software business with unit economics. For 2027, the consumer side is the open question: 1.2 billion weekly ChatGPT users against 50 million subscribers is why ads arrived, and why they are now in 89 countries.
Is AI Legal for Kids? The Under-18 Rules Coming in 2027
Yes, but access is narrowing. No major market bans minors from AI chatbots outright. What is spreading is a set of age-assurance and safety duties aimed at companion chatbots, products that simulate friendship or emotional support, and 2027 is when the strictest ones take effect.
| Jurisdiction | Rule | Status | Date |
|---|---|---|---|
| United States (federal) | FTC Section 6(b) study orders to seven AI chatbot companies on effects on children and teens | Study; no enforcement | Issued Sep 11, 2025 |
| Italy | Law 132/2025, art. 4: parental consent required for children under 14 to access AI; 14–17-year-olds may consent themselves | In force | Oct 10, 2025 |
| New York | AI Companion Models law: disclosure and suicide-prevention protocols | In force | Nov 5, 2025 |
| California | SB 243: companion chatbot disclosure, break reminders every three hours for known minors; annual reporting to the Office of Suicide Prevention from Jul 1, 2027 | In force | Jan 1, 2026 |
| Australia | Age-Restricted Material Codes: AI companion chatbots able to generate sexually explicit, high-impact violence or self-harm material must confirm a user is 18+ before access to it; civil penalties up to A$49.5 million | In force | Mar 9, 2026 |
| California | SB 1119 (Adam’s Law): risk assessment before offering a companion chatbot to children, independent child safety audits every two years, ban on cross-context behavioral advertising to under-18s | Signed Sep 10, 2026 | Operative Jul 1, 2027 |
| New York | Safe by Design Act (S.9008-C, Part Y): age assurance before offering an integrated AI companion on a covered platform; disabled by default for minors, parent can override | Enacted | Jan 1, 2027 |
Sources: Federal Trade Commission; Gazzetta Ufficiale n. 223 (September 25, 2025); California SB 243 and SB 1119 bill texts; Office of the Governor of California; eSafety Commissioner; New York S.9008-C bill text. In-force dates for Italy, New York’s 2025 law and SB 243 as summarized by Squire Patton Boggs, Morrison Foerster and Orrick. This table describes the rules; it is not legal advice.
The demand side explains the urgency. In a survey of 1,950 Australian children aged 10 to 17, 79% had used an AI companion or AI assistant and 8% had used an AI companion specifically, which the eSafety Commissioner estimates at around 200,000 children (March 24, 2026). The rules already changed products: after eSafety’s transparency notices, Character.AI introduced age assurance for Australian users and removed the chat function from its under-18 experience, and Chub AI withdrew its service from Australia.
The FTC study has produced no published findings as of November 1, 2026. When it does, it becomes the first federal dataset on how AI companies treat minors, and the obvious template for a federal rule.
What Could Change This Outlook?
| Signal | Date | Pushes toward | Why it matters |
|---|---|---|---|
| New York Safe by Design Act takes effect | Jan 1, 2027 | Downside for companion apps in the US | First US age-assurance gate before AI companion access |
| Alphabet Q4 2026 results | Early February 2027 (expected) | Upside if Gemini users and tokens keep compounding | Gemini app is the fastest-growing reported audience |
| Anthropic listing and first quarterly report | Pending S-1 effectiveness | Either | Public quarterly revenue and cost data for a frontier lab |
| Microsoft’s next Global AI Diffusion Report with expanded measure | Not yet dated | Upside in measured China usage | Breaks the 18.8% series; compare like with like |
| California SB 1119 duties apply | Jul 1, 2027 | Downside for teen engagement with companion chatbots | Risk assessments and ad limits for child users |
| EU AI Act watermarking grace period ends; ban on AI nudifier tools applies | Dec 2, 2026 | Downside for unlabeled synthetic media in the EU | First EU AI Act deadlines after the Digital Omnibus |
| Gartner’s “20% of public commits by autonomous coding agents by late 2026” | Checkable early 2027 | Upside for agent adoption if met | The nearest-dated agent prediction on any 2027 list |
| DataReportal Digital 2028 Global Overview | October 2027 (expected) | Grade-back | Tests the 4.0 billion base case directly |
Methodology
What the base case measures. Monthly active users of top generative AI platforms, in DataReportal’s scope, for October 2027. It is not a count of unique people; DataReportal says the series may include duplicate accounts.
Inputs. DataReportal’s Digital 2027 Global Overview Report and its AI deep-dive (both October 21, 2026) for GenAI users, internet users and the 83.5% any-AI ceiling. The October 2025 base (2.061 billion) is implied from DataReportal’s 57.2% growth rate. Internet-user growth of 1.0% is an Axis assumption.
Model. Share of internet users using GenAI platforms closes a fixed fraction k of its gap to the ceiling each year. The base case k (0.38) equals last year’s observed value. Downside and upside use 0.25 and 0.50.
Limitations. The ceiling applies to internet users aged 16 and above, while the GenAI count includes younger users; the cap is therefore approximate. The ceiling itself may rise as AI is built into more products. One year of history sets k. A change in DataReportal’s method, such as counting operating-system assistants, would move the series without any change in behavior.
Reconciliation choices. Microsoft’s share was converted to a headcount using the World Bank’s 2025 population aged 15 to 64; Microsoft does not publish one. We did not average any two user estimates, and we did not compute time-spent ratios across firms with different panels.
What was excluded. Statista Market Insights’ generative AI user forecast is excluded. Its current edition (most recent update May 2026, horizon 2032) shows values only to subscribers, and figures circulating on secondary sites (about 374 million users in 2025, 1.2 billion by 2031) come from an earlier edition whose scale differs sharply from the current chart. We do not quote numbers we cannot read at the source, and we do not read values off a chart image. Counterpoint Research’s insight of December 17, 2025 forecasts that monthly active users of AI chatbot platforms will surpass 5 billion by 2030 (the same page elsewhere says they will grow by nearly 5 billion); it counts platform users and is not directly comparable to the base case. Its companion forecast of $699 billion in consumer generative AI spending by 2030, up from $225 billion in 2023, is mostly hardware. Meta AI’s user figure is shown as older data because Meta has not disclosed a newer absolute number.
Technology trends. Trend names and three Gartner predictions come from Gartner’s October 20, 2026 press release. The world model, speech reasoning, quantitative model and coding-agent predictions are not in that release; they were presented at Gartner IT Symposium/Xpo and are taken from Network World’s report of the same day, cross-checked against Forbes, and marked non-primary in the dataset. We report only the predictions as Gartner phrases them and add no probability of our own. The eight shifts group Gartner’s ten trends with Forrester’s and IDC’s predictions; the grouping is an Axis editorial choice.
Source labels. Company figures are labeled company-reported; Sensor Tower, Similarweb, GWI, Menlo Ventures and Counterpoint figures are estimates. Where a primary page could not be retrieved, the relaying publisher is named and the row is marked non-primary in the dataset.
Grade-back commitment. We will compare these projections with DataReportal’s October 2027 count in our November 2027 edition, and that edition will open with the comparison, right or wrong.
About This Dataset
The file ai-trends.csv contains 669 rows: the base-case inputs, scenarios and formulas, the four-source reconciliation of AI user counts, platform user figures, time-spent estimates, age and country breakdowns (including Microsoft’s 147-economy AI diffusion table), revenue disclosures, and the under-18 and EU and US regulatory dates, and the 2027 technology-trend predictions of Gartner, Forrester and IDC cited on this page. Each row carries its source organization, document, URL, retrieval date, a primary-source flag and, for calculated rows, the formula. License: CC BY 4.0.
Cite as: Axis Intelligence Research, “AI Trends 2027: 4 Billion Users, 8 Technology Shifts and the Race to Get Paid,” 2026.
APA: Axis Intelligence Research, & Mitchell, S. (2026, November 2). AI trends 2027: 4 billion users, 8 technology shifts and the race to get paid. Axis Intelligence. https://axis-intelligence.com/ai-trends/
MLA: Axis Intelligence Research, and Sarah Mitchell. “AI Trends 2027: 4 Billion Users, 8 Technology Shifts and the Race to Get Paid.” Axis Intelligence, 2 Nov. 2026, axis-intelligence.com/ai-trends/.
Chicago: Axis Intelligence Research, and Sarah Mitchell. “AI Trends 2027: 4 Billion Users, 8 Technology Shifts and the Race to Get Paid.” Axis Intelligence, November 2, 2026. https://axis-intelligence.com/ai-trends/.
AI in 2027: Questions People Ask
What are the biggest AI trends for 2027?
Eight shifts appear across the 2027 forecasts of Gartner, Forrester and IDC: world models and quantitative models, speech reasoning models, governed AI agents, cheaper and more open inference, sovereign AI, tighter AI budgets, content provenance, and diverging regulation. On adoption, Axis Intelligence Research projects about 4.0 billion monthly generative AI users by October 2027.
What is a world model in AI?
A world model is an AI system that learns how space, time and cause and effect work, so an agent or robot can simulate the result of an action before taking it. Gartner analysts predicted at its October 2026 Symposium that 40% of robotics applications will use world models by 2030, up from less than 1% in 2025. AMI Labs and World Labs each raised about $1 billion for world models in early 2026.
Will AI growth slow down in 2027?
Yes, in percentage terms. Axis Intelligence Research’s base case has monthly generative AI users growing about 23% to roughly 4.0 billion by October 2027, after 57.2% growth in the year to October 2026. The slowdown comes from saturation in the US and Europe, not falling interest.
Is ChatGPT losing users to Gemini?
Not in absolute terms: ChatGPT went from 900 million to 1.2 billion weekly users between February and September 2026, per OpenAI. Its share is what is falling. Sensor Tower estimates ChatGPT’s share of unique AI assistant users dropped below 50% in March 2026 as Gemini and Claude grew faster.
How many teenagers use AI chatbots?
64% of US teens aged 13 to 17, according to Pew Research Center’s fall 2025 survey, with about three in ten using one every day. In Australia, eSafety found 79% of children aged 10 to 17 had used an AI assistant or companion.
Can a 13-year-old legally use ChatGPT or Gemini?
It depends on the country. In Italy, children under 14 need parental consent to use AI under Law 132/2025. In Australia, AI chatbots must restrict under-18s from age-inappropriate content such as sexual or self-harm material under codes in force since March 9, 2026. In New York, from January 1, 2027, AI companions built into social platforms must be switched off by default for minors identified through age assurance, unless a parent turns them on.
Do most AI users pay for it?
In the US, a slight majority do: 55% of AI users pay for at least one AI product, per Menlo Ventures’ 2026 survey. Globally, paid accounts are a small slice; OpenAI reported more than 50 million consumer subscribers against more than 900 million weekly ChatGPT users at the time.
Why does Microsoft say 18.8% of people use AI when other reports say half the world?
Microsoft counts working-age people seen using a generative AI product in its telemetry, which is about 1.0 billion people. DataReportal’s broader estimate counts anyone 16 and over using any AI, including AI answers in Google Search, which exceeds 4.1 billion. Same question, different definitions.
Related Axis Research
- Generative AI statistics 2026 — US adoption by quarter, consumer and enterprise spending
- AI agents trends 2027 — the enterprise agent spending forecast
- Data center trends 2027 — the capex behind the compute
- AI energy consumption forecast
- Future of work trends 2027
- AI cybersecurity trends 2027 and AI scam trends 2027
- ChatGPT statistics and OpenAI statistics
- AI adoption by country
- AI regulation tracker and AI IPO tracker
- Social media trends 2027