Unique Research · CHINA AI 100 Survey Notes · October 20, 2026
Aligning two years of slices: five structural shifts in China’s AI startup ecosystem
“At one end, the first cohort of AI companies is turning in revenue and listing answers; at the other, new players are still accelerating into the field.”
Over the past two years, Unique Research (非凡产研) has continuously surveyed and collected data from AI companies that applied to the CHINA AI 100. The sample spans models, infrastructure, applications, intelligent hardware, and industry scenarios.
The vast majority of companies appear in only one of the two years — the sample pool itself is turning over at high speed.
Survey data for a ranking list is, in essence, a slice of China’s AI startup ecosystem.
Align the two years of slices, and some structural shifts come into view.
As applications open for the 2026 CHINA AI 100 (China AI Top 100), we decided to share these observations first — a wrap-up of the past two years, and a draft of next year’s trends.
Among 2024 applicants, the densest labels were “productivity tools,” “industry applications,” and “large language models.”
By 2025, the most conspicuous variables became three words:
The share of application-layer companies among all applicants rose from a bit over 60% to about two-thirds.
At the same time, categories that did not exist the year before — “embodied intelligence” and “MaaS” — appeared on the list for the first time.
That is not to say models no longer matter — the absolute number of model-layer applicants is still growing.
The industry’s center of value creation is shifting from “building the model” to “putting the model into concrete scenes.”
The surge in digital employees deserves special attention.
When a company delivers not “customer-service SaaS” but an “AI customer-service employee,” what changes is not only the pitch — it is the entire pricing logic: from charging per seat to charging per output.
Software is starting to eat services budgets. That is a more accurate description than “AI replacing people.”
In 2024, about a quarter of applicant companies had annual revenue above RMB 100 million.
In 2025, that share rose to more than 40%, and for the first time a cohort appeared with annual revenue above RMB 500 million.
Company maturity is clearly moving up as well: the share of already-listed companies rose from under 1% to about 7%, and the number of companies at Series C or later more than doubled year over year.
But the other end of the data is equally interesting: among 2025 applicants, companies founded after 2023 accounted for nearly 40%, and early-stage founders from seed through Pre-A still made up more than a third of all applicants — the pace of new entrants has not slowed at all.
At one end, the first cohort of AI companies is turning in revenue and listing answers; at the other, new players are still accelerating into the field.
This “both ends thickening at once” dumbbell structure shows the industry has neither entered a zero-sum game over a fixed pie, nor stayed stuck in early trial-and-error — it is completing a generational overlay.
Among 2025 applicants, median headcount was 60; companies with fewer than 50 employees accounted for a full 40%.
More interesting still: a sizable share of these small companies reported annual revenue in the “tens of millions of RMB” or even “over 100 million RMB” bands.
Headcount efficiency is being rewritten by AI itself first.
That matches what we see on the ground: a team of a dozen people, using Agent orchestration, covers content pipelines, customer-service systems, and data analysis that used to take a hundred-plus people.
AI companies’ own organizational form is the first exhibit of AI productivity.
Growing from small to large is no longer the only success path — “small and highly efficient” is becoming a company shape you can choose on purpose.
That is also why this year we created a separate list for “super individuals” (AI OPC 100) — when an individual can assemble a 7×24 Agent team, the boundary of the “company” itself is being redefined.
The 2025 survey collected founder information in full for the first time. The headline figures carried in the section title are worth recording on their own: median founder age 41, and newcomers born in 2002.
Another detail few notice is corporate structure: among 2025 applicants, about a quarter used VIE, red-chip, Sino-foreign joint venture, or pure overseas offshore structures.
AI founding, from day one, is preparing for dollar capital and global markets — something the previous generation of Chinese internet companies usually considered only after Series C.
Among 2025 applicants, nearly 60% flagged overseas business; more than a quarter already have overseas employees; and companies with 100% of revenue from overseas are not just one.
The detail on geographic coverage says even more about direction: Southeast Asia was the most-checked overseas destination, followed by the US, Japan/Korea, and Europe; the Middle East, Latin America, and Africa also began appearing on the same application form.
As a contrast: in the 2024 data, the word “overseas” barely existed in any survey field — not because companies had no overseas business, but because it was not yet worth filling in on its own.
A year later, it became the most carefully completed part of the required fields.
Globalization for Chinese AI companies is no longer “an extension desk of the domestic business,” but “priced for the global market from day one.”
Model capability catching up, lower engineering cost, and the supply chain right behind them — the hand this generation of AI founders holds is completely different from the last generation of SaaS companies.
The 2026 CHINA AI 100 is aimed at AI-native companies: firms whose core technology, products, and business model are built on AI from the ground up — AI as the critical capability for creating value, not a replaceable add-on.
Review uses the TPES four-dimension model — Technology leadership, Product competitiveness, Economic growth, and Social value — with no preset fixed weights, cross-checked through user research, company research, verifiable operating data, and expert review.
Application deadline: November 6, 2026
Award notice: November 9, 2026
The list will be released on November 19, 2026, live at the “2026 Extraordinary Awards · Beijing · AI Creators Conference” (2026 非凡大赏 · 北京 · AI 创造者大会).
The conference theme is “Always-on intelligence, individual evolution” — Day 1 on always-on agents and one-person company operations; Day 2 on how Chinese AI builds a global business. The same event will also release the 2026 AI OPC 100 super-individual ranking.
The application channel will open shortly; follow subsequent posts from this account for the entry point and materials requirements.
If your company is already described by the trends above, you are welcome to apply.
If your company is creating the next trend, you should apply even more.
Tell the AI founders around you — the 2026 list should have their names on it.