SHANGHAI, China — On the same day OpenAI retired Sora, the text-to-video model that had defined the category for two years, Kuaishou Technology tucked a figure into its annual results that most Western observers missed. Kling AI, the short-video giant’s video generation platform, had crossed a 300millionannualizedrevenuerunrateasofJanuary2026.Fourth−quarterrevenuecameinatRMB340million,about47 million, and management projected the total would more than double through the year.
The First Real Revenue Story in Chinese AI
The contrast was hard to miss. Sora, by one estimate, had pulled in 2.14millioninrevenuewhileburningroughly1 million a day in compute — a product that collapsed under its own cost structure. Kling, by the same week, was approaching 500millioninannualizedrevenueandhadraisedcloseto3 billion at an $18 billion valuation, with a Hong Kong listing expected within a year.
That round, disclosed on July 2, drew an unusual line-up of backers. Tencent, CPE, CITIC Securities, Alibaba Cloud and Baidu all joined, with Kuaishou retaining about 68 percent. Bloomberg noted the rare sight of rivals Alibaba and Tencent investing in the same asset. For a category long dismissed as a demo reel, it was a statement: generative video had become a paying line of business.
The wider shift is captured in Andreessen Horowitz’s sixth Top 100 generative AI consumer ranking, published in March 2026. ChatGPT still led with 900 million weekly active users. But Chinese-origin products occupied a much larger share than in any prior edition — DeepSeek ranked fourth, and Kling, Manus, Kimi and Alibaba’s Qwen all appeared. Meitu placed five separate apps in the global mobile Top 50, all serving primarily overseas users. Of the 22 Chinese applications in the mobile Top 50, 19 targeted audiences outside China.
Where the Revenue Actually Accumulates
The lesson buried in these numbers is counterintuitive for an industry obsessed with foundation models. Revenue is accumulating in the application layer — the products built on top of cheap, commoditized models, not in the models themselves.
“A year ago, the conversation was about who had the biggest model,” said a recruitment specialist at SunTzu Recruit. “Now everyone can license a foundation model for a fraction of what it cost to train one. The scarce resource is the person who can take a $300 million video product and grow it to a billion across five language markets.”
That person is not an engineer. Kling ships native audio and lip-sync in Chinese, English, Japanese, Korean and Spanish — which means geographic expansion is a product feature, not just a sales motion. Growing it demands leaders who understand how a Tokyo advertising agency buys creative software, how a Seoul production house budgets for AI tools, and how a London brand team evaluates a video model against a human director. The talent pipeline for that profile, in most of these markets, barely exists.
The Talent Bottleneck Has Moved
For years, China’s AI talent crunch was framed as a researcher problem — companies paying seven-figure salaries for PhDs who could train models. That framing is now stale, according to a senior consultant at SunTzu Recruit. The models are good enough and cheap enough that the constraint has migrated to the commercial and organizational layer.
Searches for AI commercial leaders, localization heads and standalone-company executives in Shenzhen and Shanghai now routinely run six months or longer, the consultant said. The typical candidate has spent a career in one market — an advertising executive in Tokyo, a film-studio head in Los Angeles — and lacks the cross-border instincts the role requires. “We are looking for someone who can close a deal in Seoul on Monday and brief a London team on Friday. That person is rare in any market, and rarer still when you need them to understand a Chinese parent company’s governance.”
The supply chain that produces this kind of leader is thin. International firms that sell AI video into five language markets need executives who pair commercial judgment with genuine cross-cultural fluency, and the two rarely co-exist in a single résumé.
The spin-off structure compounds the problem. As Kuaishou carves Kling out into a standalone company headed for public markets, it needs a leadership bench that a unit inside a consumer app never had to build: a CFO comfortable with Hong Kong listing rules, independent directors, investor-relations officers. A partner at SunTzu Recruitment, which advises AI companies on these transitions, said the demand has appeared “almost overnight,” well before the supply of experienced executives.
Not Every App Will Find the Same Path
That said, the video category is still small in absolute terms. A $300 million run rate is a rounding error next to ChatGPT’s traffic, and the path Kling found — enterprise and creative customers paying for a workflow tool — will not be open to every Chinese AI app. Some will stay consumer-only and fight for attention against free Western alternatives. To be fair, the initial read that “China only builds models” was already half wrong; what is now clear is that the application layer advantage depends on commercial leadership, not just product merit. Searches that SunTzu Recruitment has run for these roles confirm the imbalance: candidate pools are a tenth the size of comparable engineering searches, and closing them takes twice as long.
One recent mandate illustrates the friction. A Shenzhen-based AI video company, profitable in China through enterprise clients, wanted to open a Southeast Asia operation. The founders assumed they needed a sales director. After six months they realized the real gap was a country manager who could navigate local content rules, partner with regional studios and recruit a bilingual team — a profile SunTzu Recruit ultimately filled only after widening the search beyond the obvious candidate pools.
The Market Call
The forward-looking read from a SunTzu Recruit recruitment specialist is blunt: the next two years of China’s AI story will be written by commercial and organizational leaders, not researchers. The companies that can staff the application layer — the multilingual go-to-market heads, the creative-market operators, the executives who can run a standalone listed AI business — will convert technical parity into durable revenue. The ones that keep hiring engineers alone will watch that revenue leak to rivals who figured out the commercial half first.
For overseas clients weighing a China hire, the signal is the same. The engineering talent is abundant and inexpensive relative to its output. The commercial and cross-border leadership is not — and it is the part that determines whether a 300millionproductbecomesa1 billion one across international markets.
FAQ
Q: Why has the AI talent bottleneck shifted from researchers to commercial leaders?
A: Foundation models are now cheap to license, so few companies need to train their own. The scarce resource is the executive who can take a video product built for one market and grow it across five language markets, which is a commercial and organizational problem rather than a technical one. Searches for these roles now run six months or longer.
Q: What does a cross-border AI commercial leader actually need to do?
A: The role pairs commercial judgment with genuine cross-cultural fluency, a combination that rarely appears on one résumé. A candidate may have spent a career in a single market as an advertising executive or a studio head. The job requires understanding how a Tokyo agency buys creative software, how a Seoul production house budgets for AI tools and how a London brand team evaluates a video model against a human director. Closing a deal in Seoul on Monday and briefing a London team on Friday is the benchmark.
Q: Why do these searches take so long and how thin are the candidate pools?
A: Candidate pools for commercial leadership run about a tenth the size of comparable engineering searches and take roughly twice as long to close. The supply chain that produces such leaders is thin: executives who pair commercial judgment with cross-cultural fluency rarely coexist with the governance instincts required to work inside a Chinese parent company. Searches in Shenzhen and Shanghai now routinely run six months or longer.
Q: When an AI unit is carved out into a standalone listed company, what hiring does that create?
A: It requires a leadership bench the unit never had to build inside a consumer app: a chief financial officer comfortable with Hong Kong listing rules, independent directors and an investor-relations function. The demand appears almost overnight and arrives well before the supply of experienced executives. One mandate illustrates the friction on the operating side: a profitable enterprise-facing AI video company opening a Southeast Asian operation found its real gap was a country manager, not a sales director.
Sources: Hello China Tech (Poe Zhao, Mar 30, 2026); Kuaishou Q1 2026 earnings; South China Morning Post and TechNode (Kling AI spin-off, Jul 2, 2026); Bloomberg; Sensor Tower via type0.ai; Andreessen Horowitz Top 100 Gen AI Consumer Apps, 6th edition (Mar 2026).
Comments are closed