Three rivals, one cap table: reading the Kling round

Kuaishou disclosed on Friday that it had taken 19.04 billion yuan (about $2.8bn) into its AI video subsidiary Kling, at a post-money valuation of roughly $18bn. Tencent, Alibaba and Baidu are all on the cap table. That last sentence is the one worth reading twice.

The three Chinese platforms that spend most of their working lives trying to take share from one another have co-invested in the same AI video company, in the same round, ahead of a planned Hong Kong spinoff. This is a slightly strange arrangement, and I think it is going to produce slightly strange outcomes.

What was actually disclosed. Kuaishou's Friday filing describes a "capital injection" into the Kling subsidiary rather than a primary equity raise at the parent. Post-round, Kuaishou's stake sits at approximately 68%, down from 100%. The named investors are CPE, Guofang, BlueFive Capital, Tencent, Alibaba, Baidu and Citic Securities.¹ Kuaishou also disclosed that Kling generated roughly 650 million yuan of revenue in Q1 2026, about four times the year-earlier quarter.² The company's Hong Kong-listed shares closed the day up as much as 6.89%.³

The AI-carrying-parent trade, in one chart you don't need. The market's reaction is the tell. Kuaishou's operating business, short video, live commerce, ads, did not change on Friday. What changed is that a subsidiary was priced. Once Kling has a public number attached to it, the parent gets marked to sum-of-parts, and the residual platform business is implicitly re-valued as whatever is left over. A 6.89% pop on a day with no operating news is the market doing that arithmetic in public.

This is the Baidu/Ernie pattern, and before that the SoftBank/ARM pattern: a legacy parent with a declared AI child, where the child is doing most of the equity-story work. The spinoff makes the trade explicit. Post-listing, Kuaishou shareholders own a majority stake in a separately traded AI video company plus a short-video platform, and the market will price those two things independently. That is usually good for the sum and awkward for the parent's management, who now have two boards, two disclosure regimes, and one set of GPUs to allocate between them.

The strategic détente. The more unusual feature of this round is not the number. It is the guest list. Tencent, Alibaba and Baidu compete across cloud, search, social, payments, and enterprise AI. They do not habitually share cap tables. The last time all three appeared as strategic investors in the same private Chinese technology company at scale, I struggled to think of a clean precedent.

There are two readings, and they are not mutually exclusive. The first is defensive: none of Tencent, Alibaba or Baidu could tolerate the other two acquiring an exclusive commercial relationship with what may become the dominant Chinese AI video model, so each bought optionality. The second is coordinative: Chinese AI video is being treated, implicitly, as a strategic sector where a single national champion is preferable to fragmented competition, and the three platforms are collectively underwriting it. Citic Securities' presence, a state-adjacent investment bank rather than a pure financial sponsor, nudges the reading toward the second. Either way, the round documents a market where the biggest domestic buyers of AI capacity have decided they would rather co-own the supply than compete for it.

The revenue multiple, which is where the frame strains. 650 million yuan of Q1 revenue annualises to roughly $360m, against an $18bn valuation. That is about 50 times forward revenue (revenue multiple: valuation divided by annualised revenue), which is aggressive even in current AI comps. Runway, the closest Western analogue, is reportedly valued at around $4bn on materially smaller disclosed revenue; the multiple is directionally similar, the absolute scale is not.

Four-times year-on-year growth makes the multiple defensible if you believe the growth curve continues. It makes it indefensible if you believe AI video monetisation flattens as competition from Veo, Seedance and Runway intensifies and the initial curiosity-driven adoption saturates. I do not know which of those is right. I do know that $18bn is not a price on 650 million yuan of quarterly revenue; it is a price on the belief that AI video becomes a structural category and Kling becomes the Chinese entry in it. That is a performativity bet, the valuation exists to make the outcome more likely, and it will be judged on whether Q2 and Q3 revenue keep the four-times trajectory alive.

Is this a round or a cornerstone? The investor mix reads more like the cornerstone book of a Hong Kong IPO than a conventional growth round. Cornerstone investors (institutional buyers who commit to a fixed allocation at IPO pricing in exchange for a lockup) are a standard feature of HK listings, and the composition here, one state-adjacent bank, three strategic corporates, two PE houses, is close to what a bookrunner would assemble ahead of a marketed deal. The public disclosure calls it a capital injection and does not specify lockup terms or IPO-linked conditions. If those exist in the actual subscription documents, the round is functionally pre-IPO cornerstone capital wearing venture clothing. Worth watching for in the eventual prospectus.

What this is a case of. Chinese AI companies are choosing Hong Kong over US exchanges, deliberately. MiniMax and Zhipu are pipeline HK listings; Kling now joins them. The regulatory, geopolitical and capital-access reasoning is by now familiar, but the cumulative effect is that Hong Kong is becoming the primary listing venue for Chinese AI pure-plays, backstopped by domestic strategic capital rather than by Western public-market demand. That is a structural market development, and Citic's participation is the clearest signal yet that Chinese capital-markets infrastructure is actively routing AI champions toward it.

What I'd watch. Q2 revenue disclosure, when it comes: the multiple lives or dies on the growth curve. The prospectus, when it lands: whether the July round is retrospectively re-described as cornerstone capital. Any disclosure on GPU allocation between Kuaishou's short-video ranking stack and Kling's inference workload, once the two entities are separately governed. And any sign that the Tencent-Alibaba-Baidu détente extends beyond this cap table — whether, say, Kling's model gets preferential distribution across all three platforms, or whether each strategic investor quietly builds a competing internal capability while owning a stake in the champion.

The round is priced. The story it tells about Chinese AI market structure is the more interesting artefact.

Footnotes