NextFin News - The global artificial intelligence landscape reached a pivotal milestone this month as China’s leading large language model (LLM) startups transitioned from venture-backed unicorns to public entities. In a rapid-fire sequence of listings on the Hong Kong Stock Exchange, Beijing-based Zhipu AI and Shanghai-based MiniMax successfully completed their initial public offerings (IPOs), collectively bringing nearly $16 billion in market capitalization to the floor. Zhipu, founded by Tsinghua University researchers, raised $558 million with a debut valuation of approximately $4.4 billion. Its peer, MiniMax, founded by former SenseTime executives, saw its shares surge as much as 54% on its first day of trading, pushing its valuation to a staggering $11.4 billion.
These listings occur against a backdrop of intense technological rivalry between Washington and Beijing. While U.S. President Trump has maintained a rigorous stance on technology transfers, the successful IPOs of these firms demonstrate a robust appetite for domestic AI champions. According to Recode China AI, Zhipu and MiniMax are being hailed as "China’s OpenAI," yet their financial disclosures reveal a business model far removed from the patient, research-first approach of their Silicon Valley counterparts. The capital raised is intended to fuel the next generation of models, such as Zhipu’s GLM-5, and to secure the massive computing power necessary to compete on the global stage.
The sky-high valuations of these companies are not merely a reflection of their current revenue, which remains modest compared to their market caps, but rather a result of a "scarcity premium" within the Chinese ecosystem. With U.S. tools like ChatGPT and Claude restricted in mainland China, a massive market vacuum has been created. Domestic enterprises and government agencies are under significant pressure to adopt homegrown solutions to ensure data sovereignty and technological self-reliance. This captive demand allows firms like Zhipu to report annual revenue growth exceeding 130%, even as they burn through billions in R&D. For instance, Zhipu’s R&D spending in 2024 represented over 700% of its total revenue, a ratio that would be untenable in almost any other industry.
A closer look at the revenue structures reveals a fundamental divergence in how these companies justify their valuations. Zhipu has leaned heavily into "on-premise" deployments—customized AI solutions installed directly on client servers. This project-based model accounted for roughly 85% of its 2024 revenue. While this ensures high security for government clients, it mirrors the labor-intensive scaling challenges that plagued the first generation of Chinese AI firms like SenseTime. In contrast, MiniMax has pursued a more international and consumer-centric path. Its AI companion app, Talkie, became a viral hit in the U.S. market, helping the company derive nearly 70% of its revenue from overseas. This diversification has made MiniMax more attractive to global investors, explaining why its valuation is nearly triple that of Zhipu despite the latter often leading in technical benchmarks.
However, the sustainability of these valuations faces a looming "DeepSeek Challenge." The emergence of DeepSeek, a research lab backed by the quantitative trading firm High-Flyer, has disrupted the narrative that only massive IPO-funded startups can innovate. By focusing purely on research excellence without the immediate pressure of commercialization, DeepSeek has produced models that rival the best in the world at a fraction of the cost. This has forced public companies like Zhipu and MiniMax into a difficult paradox: they must continue to show commercial traction to satisfy public shareholders while simultaneously pivoting back to foundational research to avoid being leapfrogged by more agile, research-focused labs.
Looking forward, the trajectory of China’s AI IPOs will likely be defined by their ability to transition from "project-heavy" service providers to true "Model-as-a-Service" (MaaS) platforms. The current price wars in China’s cloud API market have driven gross margins for cloud-based AI services to near-zero or even negative levels. For these sky-high valuations to hold, Zhipu and MiniMax must prove they can move beyond the "AI-in-a-Box" era and build scalable, high-margin software ecosystems. As U.S. President Trump continues to shape the global trade environment, these firms will also need to navigate an increasingly bifurcated tech world, where their success depends as much on geopolitical maneuvering as it does on algorithmic breakthroughs.
Explore more exclusive insights at nextfin.ai.
