Kuaishou is injecting 1.4 billion yuan for roughly 1% of Kling AI, pushing its valuation past 100 billion, but the buyback clause is effectively an IPO bet, leaving compute costs and monetization challenges unresolved.
When Kuaishou disclosed in an announcement on the last day of August that Beijing Kling had brought in two new investors, and that one of them—the National Artificial Intelligence Industry Investment Fund—had injected 1.4 billion yuan in cash in exchange for only about 1.14% of the expanded registered capital, the shock felt across the AI large-model and internet secondary markets lay not only in the fact that the implied standalone valuation had immediately breached the 100 billion yuan mark, but also in the fact that a national-level sovereign fund had taken such a precise position—at such a rare, ultra-high valuation and ultra-small stake—in a generative video model entity incubated by an internet giant.
However, if one looks past the dazzling halo of a post-money valuation in the tens of billions and national-team backing to examine the full drawdown of the 20.4471 billion yuan subscription limit and the "repurchase rights" held by both new investors, it becomes clear: this seemingly blazing capital carnival is by no means a routine financial capital increase, but rather a life-or-death capital gambit custom-built by Kuaishou for Kling AI—spinning off a subsidiary for independent financing, binding national strategic capital to share the computing-power losses, and simultaneously signing a stringent IPO bet agreement—as it confronts the extremely terrifying computing-power consumption and lengthy monetization cycle of generative video models.
Over the past year or more, text-to-video technologies represented by Sora and Kling have pushed generative AI into a new era of visual explosion. With its formidable performance in simulating the laws of physical motion, high definition, and coherent long-video generation, Kling AI once caused a viral sensation on social platforms at home and abroad, becoming one of the very few super technical assets Kuaishou has possessed in recent years with global competitiveness.
Yet the text-to-video model is, in commercial essence, an almost bottomless incinerator of computing power. The cluster GPU computing cost behind generating every second of high-definition footage is several orders of magnitude higher than that of plain-text tokens. For Kuaishou, whose main platform still relies on refined cost control to sustain tens of billions in profit, if it were to keep bearing Kling's billions in R&D depreciation, computing-power data-center leases, and high-end graphics card procurement entirely within the listed company's standalone income statement over the long term, it would inevitably severely drag down the main platform's earnings quality and shareholder returns.
Following the traces left in the business registration underpinnings to penetrate this AI unicorn's trajectory of transformation, its operational thread—from a marginalized internal project within a big tech company to an independent platform directly carrying national-level capital—is laid completely bare in the Tianyancha records. Tianyancha business registration data shows that Beijing Kling Technology Co., Ltd. was established in April 2026, with registered capital of only 10 million yuan, legal representative Yang Fan, registration status active, and enterprise type wholly owned by a foreign-invested enterprise legal person.
In the historical information accumulated by Tianyancha, this entity was formerly named "Beijing Xingwan Technology Co., Ltd.," with its business scope precisely covering key areas such as artificial intelligence basic software development, application software development, and second-category value-added telecommunications services.
From a 10-million-yuan foreign-funded shell company established just four months ago, whose former name was steeped in early incubation overtones, to a cash-absorbing behemoth now directly carrying more than 20 billion yuan in capital capacity, this lightning-fast legal-person change and asset injection reflects the extremely seasoned capital-isolation financial engineering of a big tech company spinning off a high-valuation business.
After completing this round of capital injection, Kuaishou still indirectly holds absolute control of about 68.33% of Beijing Kling and continues to consolidate its financial statements into the listed company's overall accounts. This means that while firmly locking down core control, Kuaishou has successfully drawn down the full 20.4471 billion yuan subscription limit previously set by releasing about 30% of equity to outside parties. This operation—using the real money of external institutions to bear the高昂 computing-power costs of model pre-training and inference—not only preserves absolute dominion over the underlying technological achievements and commercial ecosystem, but also builds an extremely thick loss buffer for the main platform at the financial level.
The more subtle and perilous game lies deep within the "repurchase right" clauses obtained by the National Artificial Intelligence Fund and Chia Tai Robotics.
The entry of a national-level industrial investment fund first secures for Kling the highest level of credit endorsement and protective amulet in algorithm filing, cross-border data flow compliance, and the scheduling of key domestic government and enterprise computing-power resources. But under the rigid discipline of preserving and increasing the value of state-owned capital, sovereign capital will never act as a pure venture philanthropist. The harsh consideration of 1.4 billion yuan for only 1.14%, paired with an explicitly written repurchase right, is in the hidden lexicon of investment banking operations almost directly equivalent to a stringent IPO bet agreement.
The existence of the repurchase right means that if Beijing Kling cannot achieve an independent spin-off listing within the agreed time window (typically three to five years), or if its commercialization and cash-generation capability fails to meet preset assessment targets, Kuaishou's parent body or Kling must rigidly repurchase this 1.4 billion yuan plus interest, along with tens of millions of dollars in premium. This clause not only locks in the exit safety line for the national fund, but also hangs like a Sword of Damocles over Kling's management, forcibly pushing this startup entity, established only months ago, onto an IPO countdown channel where only success is permitted and failure is not.
Yet the commercialization and monetization of the text-to-video track still confronts a brutal chasm of reality.
At present, global text-to-video tools generally rely on monthly subscription fees from individual creators (SaaS) and small-scale procurement by film and television special-effects teams, and their average order value and paid penetration rate are far from catching up with the suffocating inference computing costs behind them. Even if Kling has launched API interfaces to charge advertising and gaming companies, under the relentless pressure of industry price wars and open-source models, the high unit computing cost is still ruthlessly devouring gross margin.
If Kling cannot, within the coming strategic window, truly embed its dazzling generation capabilities into industrial-grade film and television pipelines or the core livestream e-commerce selling chain to achieve positive cash flow, then the tens-of-billions valuation bubble will ultimately collide with the extremely stark hard landing of commercial monetization.
This national-level heavyweight capital injection in late autumn sends the coldest signal to the entire generative AI track: the era of going it alone in the large-model arms race has completely and definitively ended. When R&D investment rises to the scale of tens of billions, only top players capable of packaging the supercomputing story into an independent entity, cleverly leveraging national sovereign capital and industrial giants to share risks, and daring to sign repurchase agreements to bet on the future can obtain a ticket to the final decisive battle.
Under the double squeeze of a hundred-billion valuation and the shackles of repurchase, the time left for Kling to convert cutting-edge technology into real commercial profit is already counted precisely in months.