A significant announcement on July 20, 2026, captured market attention: 01.AI, the AI large model company founded by computer scientist Kai-Fu Lee, is advancing its Pre-IPO funding round with the goal of listing on the Hong Kong Stock Exchange in 2027. Lee confirmed in an interview with Bloomberg during the Shanghai World Artificial Intelligence Conference that the company is in the process of unwinding its offshore holding structure, a prerequisite for an overseas listing.
From its founding in May 2023, when it was hailed as a potential "Chinese OpenAI," to the formal emergence of its 2027 listing plan, 01.AI has charted a unique transformative path over four years. This is more than just the IPO story of an AI unicorn; it is a microcosm of the Chinese large model industry's shift from technological idealism to commercial application.
Strategic Shifts: From AGI Aspirations to Industrial AI
01.AI was established in May 2023 by Kai-Fu Lee himself. Leveraging Lee's personal reputation as an "AI evangelist" and the breakout performance of its Yi series of open-source models, the company achieved a valuation exceeding $1 billion within just 8 months, joining the ranks of prominent "AI tiger cubs." Investors included Alibaba Group's Cloud Intelligence Group, among others.
However, over the past two years, 01.AI has undergone two radical strategic pivots.
The first pivot occurred in early 2025. The company made a decision that shook the industry: it transferred the majority of its large model pre-training team and AI infrastructure team to Alibaba Cloud, actively abandoning the development of a trillion-parameter super-large foundational model. Lee later admitted that period was one of the "top three darkest moments" of his life, with the team facing a crisis of confidence and office desks left empty by a third.
The reason for the shift was straightforward. "If you want to be number one, how much money do you need to raise? What is the probability of success? This is a business question," Lee stated. The release of DeepSeek's open-source models fundamentally upended the economics of large model training—the business model of building cutting-edge models from scratch had become unsustainable for the vast majority of companies. Only a few with "bottomless" financial resources could still justify the cost of building models from the ground up.
The second pivot followed closely. 01.AI repositioned itself not as a manufacturer of frontier large language models, but as a provider of enterprise AI infrastructure. The company ceased building its own foundational models, instead focusing on fine-tuning and customizing existing domestic open-source models like DeepSeek, Alibaba's Tongyi Qianwen, and Zhipu AI's GLM. It packages these models within software that organizes customer data, layered with AI agents. Products are typically deployed on clients' own servers rather than the cloud, as enterprise buyers increasingly insist on keeping sensitive data in-house.
Lee termed this new product "Boss AI" and described 01.AI as "China's Palantir." Palantir Technologies Inc. is a U.S. big data analytics company with a market capitalization exceeding $320 billion and a net profit margin of 53% in the first quarter of 2026. In an internal letter in May 2026, Lee explicitly set the goal of becoming "the first Chinese AI 2.0 company to achieve profitability breakeven."
Following the strategic shift, 01.AI's product logic fundamentally changed. The company no longer pursues flashy tech demos, but focuses on quantifiable business outcomes. On July 7, 2026, it officially launched its "Top Role AI" product suite, including Boss AI, Sales Champion AI, and Investment Officer AI, targeting CEOs, top sales leaders, and investment heads, respectively. The company's product head revealed that after implementing these products internally, order value increased fivefold and lead conversion rates doubled.
For commercialization, 01.AI adopted a dual-track strategy. Domestically, it jointly announced the establishment of a joint venture, "Wanfeng Intelligence," with the Charoen Pokphand Group, using egg-laying farming as the first pilot to build a vertical AI agent platform for agriculture. This marks a complete departure from traditional IT outsourcing, moving towards deep co-creation where profits and risks are shared with clients. Internationally, the company explicitly chose to avoid the U.S. market—where Lee noted buyers remain cautious about Chinese software—focusing instead on Europe, the Middle East, Asia-Pacific, and Africa, targeting the sovereign AI sector. Lee has been invited to join Kazakhstan's National AI Council, and the company is deploying sovereign AI solutions in Kazakhstan and other countries.
Organizationally, 01.AI maintains a lean team of approximately 240 employees, quite streamlined by industry standards. The company once most resembling "China's OpenAI" is now redefining itself with the agility of a "leopard."
Hong Kong's AI Valuation Frenzy and the Profitability Test
01.AI's IPO plan is not an isolated event but part of a collective wave of Chinese large model companies seeking listings.
In early 2026, Zhipu AI and MiniMax listed on the Hong Kong exchange, staging a capital feast of "soaring on debut." As of July 16, 2026, Zhipu AI's stock price reached HK$1,540 per share with a total market cap of HK$717.06 billion, while MiniMax traded at HK$260.6 per share with a market cap of HK$91.011 billion.
Following them, Moonshot AI initiated its Hong Kong listing process in July, expected to complete within six months. Wenge AI landed on the Hong Kong exchange on June 26, becoming the "first decision-making large model stock," with its share price surging over 105% on its first day. According to disclosures from HKEX and the China Securities Regulatory Commission, other "AI tiger cubs" like Baichuan AI and Stepfun are also advancing capital operations to varying degrees.
Behind this listing wave lies a regulatory incentive from HKEX. Chapter 18C of the Listing Rules, effective March 31, 2023, is specifically designed for Specialist Technology Companies, significantly lowering the barrier for AI and semiconductor firms to raise funds in Hong Kong. By the end of March 2026, 14 companies had listed in Hong Kong under Chapter 18C.
However, the flip side of the capital feast is the underlying concern of a valuation bubble. Both Zhipu AI and MiniMax are currently loss-making, with no positive price-to-earnings ratio to calculate. For instance, at a market cap around RMB 400 billion, Zhipu AI's price-to-sales ratio exceeded 550 times, surpassing 1,000 times at its trillion-yuan peak. A CEO of a listed company remarked, "Zhipu AI has annual revenue of RMB 7 billion but can reach a market cap of RMB 1 trillion. I find this world a bit hard to understand."
Analysts note that while traditional financial metrics may not fully apply to early-stage tech firms, excessively high valuations still signal risk. Some argue that the超高 P/S ratios essentially reflect the market pricing the long-term value of "Chinese AI entering global frontier competition," rather than static financial returns.
Whether 01.AI can attract capital favor in this valuation frenzy depends on two core variables: the verifiability of its revenue growth and the pathway to achieving its profitability targets.
Financially, 01.AI's growth momentum appears strong. For the full year 2025, its audited revenue reached RMB 250 million, with contracted orders totaling RMB 500 million. By May 2026, its order book had more than doubled to exceed RMB 1.5 billion, heading towards a target of RMB 2 billion. Lee revealed that nearly half of these orders are recurring subscriptions, which could enhance its valuation. Annual operating costs are approximately RMB 200 million.
However, there is a lag between orders and recognized revenue. Lee's target is to secure RMB 1.5 to 2 billion in contracts and achieve profitability breakeven in 2026. The company also aims to achieve quarterly profitability in 2027. If these goals are met, 01.AI would be among the first Chinese large model companies to turn a profit, commanding a significant premium in its IPO valuation.
Nevertheless, challenges remain significant.
First, the competitive landscape of the large model industry is still evolving rapidly. DeepSeek's open-source strategy continues to lower industry barriers, while internet giants like Alibaba, Tencent, and Baidu are increasing their investments in large models. By exiting the foundational model race and pivoting to the enterprise AI decision-making track, 01.AI has avoided the most capital-intensive arms race but has also forgone the narrative premium of holding the technological high ground.
Second, it remains uncertain whether the Hong Kong market will fully comprehend the strategy of benchmarking against Palantir. Palantir Technologies Inc. took 17 years to achieve profitability, with its core business moat lying in deeply embedding itself within the data infrastructure of government and large corporate clients. Whether 01.AI can replicate this path in the Chinese market carries considerable uncertainty. Lee himself acknowledged that only companies with "essentially unlimited balance sheets" can still justify the cost of building models from scratch—implying the necessity of 01.AI's asset-light approach.
Third, the Hong Kong market's pricing logic for AI assets is experiencing intense volatility. Since their listings over half a year ago, Zhipu AI and MiniMax have seen their stock prices undergo dramatic swings with divergent trends. As market sentiment shifts from euphoria to rationality, investors are likely to focus more on genuine profitability and cash flow rather than mere price-to-sales narratives.
From 01.AI's perspective, the pace and pricing of its Pre-IPO funding will be a key signal. Lee expressed hope to complete this funding round around the time of its first annual results announcement—suggesting a window likely in early 2027, coinciding with the release of its full-year 2026 financials. The company's move to unwind its offshore structure indicates its IPO preparations have entered a substantive phase.
For China's large model industry, 01.AI's listing will serve as an important case study. If Zhipu AI and MiniMax represent the capital market path for "technology-driven" large model companies, then 01.AI represents a distinctly different route—from "selling models" to "selling solutions," from the "parameter race" to "improving financial statements," from "AGI ideals" to "industrial AI implementation."
Whether this path gains recognition from the capital market concerns not only the fate of 01.AI but will also provide a crucial reference point for the commercial trajectory of China's entire AI industry. As Lee stated, "Three years ago, we set out with AGI ideals; three years later, we are steadfastly moving towards industrial AI." The journey from "China's OpenAI" to "China's Palantir" is, in essence, a footnote of an era where China's AI industry transitions from technological faith to commercial validation.
Comments