Kling's Spin-Off Signals a New AI Playbook. Is Doubao Next?
Kuaishou Technology is moving to spin off its video-generation unit Kling AI at a reported $20 billion pre-IPO valuation — a structural shift that exposes a fundamental tension between hypergrowth AI assets and the consolidated balance sheets of China's internet giants, and raises a pointed question for ByteDance: is Doubao next?
The spinoff plan, which surfaced May 11, 2026, sent Kuaishou's Hong Kong-listed shares surging approximately 11% in a single session — a market signal that investors prefer AI-driven burn rates to be someone else's problem. The company subsequently confirmed the process is underway, though no binding terms have been disclosed. Some market sources suggest the actual valuation could exceed the $20 billion figure already in circulation.
The episode is not an isolated corporate restructuring. It reflects a broader playbook taking shape across China's technology sector: carve out capital-intensive AI units, give them independent financing channels, and let growth-stage valuation multiples — rather than conglomerate P/E frameworks — determine their worth.
Kling's Numbers Justify the Spinoff Math
Kling AI's trajectory makes the strategic case almost self-evident. Launched in June 2024 as one of the first publicly accessible DiT-architecture video generation models positioned against OpenAI's Sora, the product posted full-year 2025 revenue of approximately RMB 1.04 billion (US$144 million), with quarterly figures accelerating from RMB 150 million in Q1 2025 to RMB 340 million in Q4 2025.
The momentum sharpened further in 2026. Kling AI revenue reached RMB 650 million (US$90 million), up more than 300% year-on-year. Annual recurring revenue (ARR) approached US$500 million, compared with roughly US$100 million a year earlier. The platform has crossed 60 million global users and serves more than 30,000 enterprise clients and developers.
The problem is structural: Kling's revenue, while growing at a triple-digit rate, is dwarfed by the capital expenditure it demands. approximately RMB 26 billion (US$3.6 billion), up roughly RMB 11 billion (US$1.53 billion) from 2025 — with the incremental spending directed primarily at compute infrastructure, server procurement, and data center buildout for Kling and other foundation models.
The arithmetic is stark: a business generating low single-digit billions in annual revenue is absorbing capex measured in the tens of billions. Kuaishou's Hong Kong-listed shares fell as much as 14% following its 2025 annual results, suggesting investors were already pricing in the drag before the spinoff option emerged.
At Kuaishou's current market capitalization of HK$230 billion (approximately US$29.5 billion), a standalone Kling valuation of $20 billion would represent nearly two-thirds of the entire parent company — a valuation wedge that makes the case for separation almost without further argument.
Doubao Replicates the Pre-Spinoff Playbook, Step by Step
If Kling represents the template, ByteDance's Doubao is the most plausible next candidate to follow it. The comparison holds across multiple dimensions: scale, burn trajectory, and — critically — the institutional groundwork already being laid.
Doubao is China's largest AI conversational assistant and agent platform, closer in product architecture to ChatGPT than to Kling's specialized video-generation use case. By end-2025, Doubao had crossed 100 million daily active users, becoming the first AI-native application in China to reach that threshold. According to QuestMobile data, Doubao's monthly active users reached 227 million in Q4 2025, ranking first in the domestic AI application category — a position it has maintained into Q1 2026 across multiple measurement indices.
The compute economics are different from Kling's but no less demanding. Doubao's text-based inference carries a lower per-query cost, but the platform processes approximately 120 trillion tokens daily, converting scale into a different form of structural cost pressure. Every percentage point of free-tier growth widens the gap between revenue and infrastructure spend.
ByteDance has begun closing that gap. On May 4, 2026, Doubao quietly introduced a three-tier subscription structure: a standard plan at RMB 68/month (US$9.40), an enhanced plan at RMB 200/month (US$27.80), and a professional plan at RMB 500/month (US$69.40), alongside a retained free tier. For China's largest AI application, this marks the first meaningful attempt at revenue layering — a transition from fully subsidized growth toward a freemium model with monetizable upper tiers. By contrast, Kling already derives approximately 70% of its revenue from professional creators and enterprise clients, indicating a more mature commercial model that Doubao is now beginning to approximate.
More telling than the pricing move is the internal equity architecture ByteDance has constructed around Doubao. In Q4 2025, ByteDance launched a "Doubao Long-Term Incentive Plan," issuing virtual equity tied to Doubao's business performance to employees within the Seed department — the internal unit responsible for large-model development. Initial grant prices were set at US$10 per unit. By late May 2026, the latest grant round was priced at US$13 per unit, implying a roughly 30% internal valuation step-up in under two quarters.
ByteDance has characterized the "Doubao shares" as an internal retention tool rather than a precursor to external financing or a formal spinoff. But the architecture speaks for itself: Doubao now has its own implied valuation, its own price signal, and its own incentive target — the foundational elements of a subsidiary equity structure, assembled without yet requiring external capital markets.
ByteDance's Unlisted Status Accelerates, Not Delays, the Spinoff Logic
A surface reading might suggest that Kuaishou's urgency stems from financial pressure that ByteDance, with its stronger cash position, does not share. That framing misses the more important dynamic.
Kuaishou is listed and can execute a conventional subsidiary IPO. ByteDance remains private, its own IPO timeline unresolved after years of regulatory and structural complexity. Precisely because a ByteDance group listing remains indefinite, spinning out Doubao as a standalone public entity would allow the AI asset to capture its growth-stage valuation premium years ahead of any parent-level liquidity event. A listed Doubao would simultaneously provide ByteDance with a public market reference point — a pricing anchor that could inform and potentially accelerate the eventual valuation of the broader group.
The pattern has precedent in global technology markets. ByteDance's unlisted status is not a constraint on Doubao's independence; it may be the most compelling argument for accelerating it.
A Structural Shift Across the Sector
Kuaishou and ByteDance are not acting in isolation. Baidu has been advancing its chip subsidiary Kunlun Tech toward a dual A-share and H-share listing. Alibaba has reportedly been exploring a spinoff of its chip design unit T-Head Semiconductor. The common logic is identical: AI and semiconductor assets require investment cycles and valuation frameworks incompatible with diversified internet conglomerate reporting structures. Separation unlocks independent financing, removes earnings drag from the parent, and allows growth-stage pricing.
The conditions that triggered Kling's spinoff — sufficient scale, accelerating revenue, heavy capex, rising token costs, intensifying external competition, and acute AI talent competition — describe Doubao with equal or greater precision. The question is not whether Doubao will eventually seek independent capital market access, but how long ByteDance will wait before the internal virtual equity structure becomes an external one.
Related Coverage:
Doubao Introduces Paid Tiers:China AI Industry Shifts to Monetization
Kuaishou's Kling AI Targets $20B Valuation in Planned Spinoff