Behind Zhipu AI's slight capital increase, Meituan and Tencent are closing in with their bets, a large-model oligopoly game played amid the costly burning of compute and the struggle to generate revenue.
When Zhipu AI, which stands in the first tier of domestic large models, made a seemingly insignificant business registration change, anyone who focused only on the increase of just over 2 million yuan in registered capital would completely misjudge the brutal scale of stakes in the current artificial general intelligence race. Under the leverage rules of the venture capital market, this very small portion of newly added registered share capital, increasing from 46.56 million yuan to 48.76 million yuan, often corresponds to a real injection of hundreds of millions or even billions of yuan in actual capital from the capital reserve account.
This is by no means a simple financial transfusion for a startup, but a calculated defensive marriage between a leading large-model player and internet giants during an anxious cycle in which computing power is running out and commercialization has yet to generate self-sustaining revenue.
Shareholder Structure: A Miniature Live-Fire Exercise of China's Internet Map
Following the capital trajectory at the bottom layer of commercial registration to penetrate this AI unicorn's hand reveals an extremely massive intertwining of internet power fully exposed in Tianyancha records. Tianyancha business registration data shows that Beijing Zhipu Huazhang Technology Co., Ltd. was established in 2019. Recently, it not only completed an expansion of registered capital, but its complex shareholder structure is also a miniature live-fire exercise of China's internet map.
In addition to core founding team members such as Tang Jie, who represents the Tsinghua-affiliated technology source, giant investment entities such as Tianjin Sankuai Technology Co., Ltd. and Guangxi Tencent Venture Capital Co., Ltd. are prominently listed. The appearance of Meituan and Tencent in the same frame precisely pierces through the underlying panic of big tech when facing technological disruption.
Why Are Big Tech Companies Repeatedly Betting on External Independent Large-Model Companies?
In public perception, Tencent has Hunyuan, and Meituan is also frantically recruiting AI talent to develop its own models. Since the major tech companies are concentrating computing power on building their own ecosystems, why are they still spending huge sums to invest in an external independent large-model company?
This investment layout, which appears to duplicate resources, is essentially an astronomical insurance policy bought by internet giants to guard against missing the technological boat. Although big tech companies hold massive consumer data and excellent application scenarios such as ride-hailing and food delivery, they are constrained by internal profit assessments and computing-power allocation barriers, making it often difficult for them to pursue unadulterated brute-force breakthroughs at the foundation layer in the way pure academic geek teams can.
Zhipu AI's pure pedigree, born out of Tsinghua University's Department of Computer Science and Technology, has supplied it with the nation's top algorithm engineer talent, which is precisely the foundational R&D infrastructure that big tech companies, relying on high-salary poaching, lack most. The capital injection from Tencent and Meituan is not simply for financial returns, but to use money to buy a front-row seat to watch the evolution of large-model technology. Once their internal proprietary models underperform expectations, big tech can immediately connect Zhipu's foundational capabilities directly into their own massive business interfaces through capital ties, ensuring they are not kicked out of the race for the entry point in the AGI era.
Heavy-Asset Cash Burn: The Computing-Power Money Furnace and the Revenue-Generation Dilemma
However, accepting the enormous funds from the giants' encirclement is also a heavy-metal feast that Zhipu AI has no choice but to swallow.
An examination of its approved business scope shows that artificial intelligence foundational software development and data processing support services constitute its commercial skeleton. But behind this seemingly extremely asset-light code logic lies an extremely asset-heavy cash-burning black hole. The training and iteration of each generation of hundred-billion-parameter models requires tens of thousands of high-performance GPU clusters roaring day and night, followed by suffocating hardware depreciation and astonishing electricity bills.
Against the reality that a consumer-facing super app has not yet exploded and business-to-business enterprise customers remain extremely stingy in their willingness to pay for private deployment, large-model companies' own revenue-generating capacity simply cannot support the pre-training expenses of the next generation of models. This newly added capital injection will, the moment it arrives, turn into computing power burning in the data center.
Shuffle Signal: The Oligopoly Game and the Neutral Tightrope
This capital expansion in early autumn sends the coldest reshuffling signal to the entire Chinese artificial intelligence sector: the unruly carnival of the hundred-model war is over, and the large-model table has completely become an oligopoly game that only trillion-yuan-market-cap giants and top-tier capital can sit at. How to take Meituan's and Tencent's money while still maintaining a neutral posture in the commercialization of cloud services and API calls, and not becoming a computing-power vassal of any one big tech company, is the tightrope this Tsinghua-affiliated unicorn must successfully walk before it burns through this round of financing.