Is the AI Business Boom Over? The Current State in 2026, Seen Through Published Cases

Several years have passed since generative AI became a tool for individual developers, and the track record of “businesses built with AI” has now accumulated into real numbers. This article surveys the AI-related cases published on this site to sort out what is working and what isn’t. The following analysis is based on the 337 cases published as of August 12, 2026 (168 sold, 165 operating) as the population, from which we extracted those that place generative AI at the core of the business.

The Pattern That Works Is Surprisingly Mundane

Lining up the AI businesses whose revenue has kept going, one common thread stands out. They replace existing “expensive-when-you-hire-a-human” services, within the range where the deliverable is complete as an image or text.

Making AI portraits in place of a photo studio (Photo AI), generating redecoration proposals from photos of a room in place of an interior coordinator, reading PDFs and answering questions (PDF.ai), and supporting academic paper writing. In every case, the comparison point, “how much would it cost to hire a professional for this task”, is clear, and AI delivers it at a fraction of that price, sometimes a tenth or less. The price gap itself becomes the value proposition.

Conversely, no business that sold itself on the novelty of the technology has survived. A pitch of “you can do X with AI” has no basis for its price, because there is no point of comparison.

The Boom’s Initial Speed, and What Comes After

Another feature of AI businesses is that their initial speed is unusually fast, and the decay is just as fast. Among the published cases is one that sold about $150K in a single week from AI avatar generation, then shut itself down right after seeing well-capitalized companies enter the space. The stated reason was “it’s too cheap [to build a moat with]. People will get bored of it.”

In hindsight, that judgment was accurate. Entertainment-leaning AI services get boring within weeks to months, turning into a war of attrition against companies that can pour money into advertising. An individual’s only advantage is “speed,” and that advantage is structurally short-lived — this is the consistent lesson that emerges across the published cases.

On the other hand, AI tools embedded in real work keep growing quietly. They don’t run into the boredom problem, because users open them to get work done, not for entertainment.

“Selling to AI” Beats “Building with AI”

It’s easy to overlook, but the most stably profitable players in the AI boom may be the ones selling tools to the people who use AI. The published cases include several businesses earning revenue around the periphery of AI execution, prompt management, proofreading AI-written articles, media comparing AI tools. The classic pick-and-shovel play from a gold rush is being replayed exactly.

A Warning for Reading the Numbers

Numbers in the AI space are more prone to exaggeration and selective disclosure than in any other field. Presentations like annualizing (ARR) a launch-day burst of momentum, or citing free-user counts as an achievement, are rampant. This site limits itself to cases where the operator disclosed a specific figure, but even then, it’s often impossible to know how many months that number actually held.

An AI business should be evaluated by how long it lasts, not by the size of the number. $40K a month sustained for three years is a stronger business than $150K in a single week.

The Seats Still Open

Looking across the published cases, the gap that stands out is practical, work-oriented tools for the Japanese-language market. In English-speaking markets, many tools have emerged that slot AI into professional fields such as licensed professions, healthcare, and real estate, but equivalent tools are still thin on the ground in Japanese. Because language and business-custom barriers function as an entry barrier, “AI tools that work in English-speaking markets but don’t yet exist in Japan” will likely remain a realistic target for the time being. Even so, this is a pointer to a business opportunity, not a guarantee of success. Nearly half of the published cases either failed to grow, or grew and were shut down anyway.

What’s Happening in the Market for Selling AI Businesses

Records have accumulated on the selling side as well as the building side. The published cases include an SEO SaaS acquired by an AI content company at 2.5x revenue, a case where a repository bought for $20K was grown to $1.5M ARR within months, and growth records for AI writing-assistance tools.

A 2.5x multiple is clearly lower than the usual SaaS market rate (3–6x annual revenue). Both sellers and buyers see it the same way: “this feature will eventually be absorbed into the LLM itself.” The premise that AI-space assets depreciate fast is priced in.

Flip that around, and for an AI business, “sell while it’s still running” tends to be a more rational call than “grow it and figure it out later.” The sharpest example among the published cases is the decision to shut a business down mid-boom and move on to the next product, not a retreat, but a reallocation of resources.

Where Individuals Can Still Win

The gap between well-capitalized companies and individuals, in the AI space especially, comes down to “speed” and “how narrow a niche can be.” Large companies can’t move unless the market size clears a certain threshold, but an individual can make a business work on monthly revenue in the hundreds of thousands of yen.

The winning conditions that can be extracted from the published cases come down to three. (1) The deliverable is complete as an image or text, in a domain where the responsibility for accuracy is light. (2) The price of the existing human service that serves as the comparison point is clear. (3) Users use it for work, not for fun, boredom only fails to kill revenue when this condition is met.

Conversely, cases missing any one of these three show an initial burst of speed but leave no record six months later. Across the whole set of cases, one conclusion emerges: AI made entry faster, but it left the conditions for staying in business unchanged.

Premises for This Tally

Among the cases published on this site, this piece covers those that use generative AI, image generation, or LLMs at the core of the business. All are sourced from information disclosed publicly by the operator or the party involved. Amounts are converted at ¥150 to $1.