Consumer AI Startups Ride Power Users as Spend Hits $40B

A market that grew deeper, not wider

The main consumer technology story of autumn 2026 is not a product launch. It is a change in how people relate to the software they already use. Menlo Ventures estimates that global consumer spending on AI reached $40 billion this year, more than three times its 2025 figure. Over the same period, the share of U.S. adults using AI rose only three points, from 61% to 64%.1 Put simply, the market grew because the same people used it more, not because more people joined. That has direct consequences for which emerging startups get funded, how they price, and what they build.

Menlo's figures come from a July 2026 survey of 5,067 U.S. adults run with Morning Consult. They show daily AI use rising from 19% to 25% of Americans, and 52% of AI users say they use it more than a year ago.2 General-purpose assistants alone are now used daily by 22% of consumers. That is more than ride-sharing and food delivery combined.1 The firm also says consumer AI hit the $40 billion mark about three and a half years after ChatGPT launched. App-store spending took roughly eight years to reach that level, and U.S. streaming subscriptions about sixteen.4

The payer question: two very different numbers

The coverage splits most sharply on how many people actually pay. Menlo reports that 55% of U.S. AI users pay for at least one AI product, roughly 90 million Americans. A year earlier, it estimated that about 3% of AI users worldwide were paying.2 Andreessen Horowitz's latest consumer AI ranking, based on YipitData's U.S. e-receipt panel, gives a much smaller number. Only 4.5% of eligible consumers had an active paid personal subscription to ChatGPT, Gemini, or Claude in August 2026, up from 2.1% a year earlier.9

The two numbers are less contradictory than they look, but the gap still matters. Menlo counts any paid AI product and divides by AI users. It also counts people whose subscriptions are covered by family members, employers, or schools.1 The a16z figure tracks card receipts for three flagship assistants across a broader consumer base.9 My reading is that the receipt data is the better guide to how many people pay for themselves, and the survey is the better guide to how far AI spending has reached into households. Founders who quote the 55% figure to investors as proof of mass-market willingness to pay are overstating it.

On one point both datasets agree: spending is highly concentrated. Menlo finds the 14% of payers who spend $100 or more a month account for 60% of consumer AI spending, while the typical payer spends $20 to $49.1 The a16z data is starker still. The top 1% of AI payers averaged $903 a month and made up 19.5% of observed spending, more than the bottom half of spenders combined. The top 1% also raised their spending by 80% over 18 months, while median spending barely changed.9

The heaviest spenders are buying tools to build things. They lean toward automation and product-building services such as n8n, fal, Manus, and Nous Research's Hermes Agent, and creative tools like Higgsfield, Figma, and HeyGen.9 Menlo's profile of a typical $100-plus spender is a Millennial parent with a postgraduate degree who works in technology or finance. This person is nearly three times as likely as the average American to earn $100,000 or more a year.2

Why the venture market looks the way it does

That consumer pattern helps explain where startup capital is going. Crunchbase counted $159 billion in global venture funding in the third quarter across close to 6,000 startups. AI-driven companies took $102 billion of it, or 64%.11 A record 27 companies raised rounds of $1 billion or more, and together they took around a third of the quarter's capital.11 Half of all startup capital invested in the quarter went to companies founded since 2022.11

The headline number fell from the previous quarter, and the coverage mostly agrees on why. In North America, funding dropped 35% from the second quarter to $92 billion but was still up 50% from a year earlier. Crunchbase attributes the drop mainly to the absence of new megarounds for OpenAI and Anthropic, not to a broad cooling.12 Some analysts read the more important signal as concentration. One argued that with AI taking about two-thirds of North American venture dollars, non-AI startups now compete for the same limited-partner money as AI infrastructure deals.18 Carta figures cited in founder-focused coverage put the median Series A valuation for an AI foundation-model startup near $300 million, against $55 million for a non-AI company at the same stage.16

The power-user economics also show up in individual deals. Manus, whose agent product appears among top spenders' purchases, raised more than $500 million to rebuild as an independent company after Chinese regulators forced Meta to unwind its acquisition.22 Nous Research, maker of Hermes, closed a $90 million Series B at a $1.5 billion valuation, with Nvidia, Microsoft's M12, and Samsung participating.26 Smaller bets point the same way. Eden AI Labs raised $1.5 million for a personal AI agent.25 Sabi raised a $50 million seed round led by Khosla Ventures for a brain-AI interface meant for everyday wear.25

Agents are where trust is tested

The newest test of consumer behaviour is delegation. Menlo found that 41% of AI users have tried an agent, 24% use one regularly, and 32% have let AI act for them without final approval.1 Agent users are also the most valuable customers: 92% pay for AI and 63% use it daily.1 Consumers have given agents access to their email (36%), browsers (33%), and messaging apps (31%).2

Other surveys suggest people are more cautious about money. In a Spreedly survey, 64% of Americans said they would not let an AI shopping assistant spend without their approval, and only 19% would hand over payment details in exchange for a guaranteed lowest price.6 The same survey found brand loyalty surprisingly easy to shift. Sixty-three percent would consider letting AI choose their brands if it reliably saved them 20%, and 31% of Gen Z respondents named ChatGPT as their most trusted source for product research.6 Ping Identity surveyed 11,000 consumers in 12 countries. Respondents set an average spending limit of $200 for routine agent purchases, but 42% capped it at $100 or less and 22% would not trust AI with any financial transaction.7

Global Payments' research shows the trend moving quickly. Consumers now expect AI agents to make 15% of their purchases within five years, up from 9% a year ago.8 The share comfortable with AI spending small amounts on things like cinema tickets rose from 32% to 82% in a year, and for meal delivery from 30% to 78%.10 Half of Americans still worry about payment security, however.8

Taken together, these surveys describe one consistent pattern rather than conflicting ones. Consumers will let AI influence their choices and handle small, low-risk purchases. They are much slower to give it unrestricted access to their money. This is why investors are funding the security and transaction layers around agents. Monid, which builds a marketplace where agents buy tools, raised $7.7 million.26 Agent-security startup Rein Security raised $25 million.22 Armadin, the AI security company led by Mandiant founder Kevin Mandia, raised $255.5 million at a valuation above $2.5 billion.23

The backlash consumer startups can't ignore

Menlo also reports a counter-trend. Accuracy (45%) and trustworthiness (40%) now top consumers' criteria when choosing an AI product.3 Thirty-six percent say they are less likely to engage with content they know is AI-generated.2 Ping's survey found 56% value knowing when AI is involved more than getting the fastest answer. A fifth say they would leave a website immediately if they suspected undisclosed AI.7 Coverage of business-to-consumer startups this month says the companies winning are the ones that show retention, trust, and repeat use, not broad mass-market ambitions.24

The takeaway

In my reading, the 2026 consumer tech market rewards depth over reach. The money is in a relatively small group of affluent, heavy users who treat AI as working equipment. Startups that serve them with agents, automation, and creative tools are getting priced as such. The risk is that the industry mistakes a narrow base of intense spenders for broad demand. The receipt data showing single-digit paid adoption of the main assistants is a reminder that most people have not opened their wallets.9 For the next group of consumer startups, the harder job is making delegation safe and AI use transparent enough that ordinary users will pay, not winning the power users who already do.