Kai-Fu Lee once chased frontier AI models with the same intensity that defined his career at Google and as a venture investor. Now his Beijing-based startup 01.AI has abandoned that race. The company instead builds enterprise software on top of open models from others. And it aims to list in Hong Kong in 2027.
Lee founded 01.AI in 2023. Within eight months the firm reached a valuation above $1 billion. Alibaba’s cloud unit backed the effort. Early ambitions centered on training powerful large language models from scratch. Those plans collapsed under economic reality.
“Only a handful of companies with essentially bottomless balance sheets could still justify the cost of building models from scratch,” Lee told Bloomberg at the World Artificial Intelligence Conference in Shanghai. “Everyone else needed a new business model.”
The shift happened after DeepSeek released strong open-weight models. Training costs no longer made sense for most players. 01.AI now fine-tunes models such as DeepSeek, Alibaba’s Qwen, and Zhipu AI’s GLM. It wraps them into tools that help companies organize their data.
Its main product, called Boss AI, creates isolated data pools for clients. These pools support instant querying and visualization. Buyers in regulated industries want the software deployed on their own premises. Data security drives the preference.
Lee has described the approach as building the Palantir of China. The comparison highlights a focus on top-down sales of customized AI systems to governments and large enterprises. Direct engagement replaces consumer apps or model benchmarks.
Financial results show momentum. The company reported audited revenue of 250 million yuan, about $34 million, for 2025. Contract orders exceeded 1.5 billion yuan, or roughly $207 million, as of May 2026. Management set a 2 billion yuan target for the full year. Recurring subscription revenue already accounts for nearly half the 2026 contract volume.
Operating costs run about 200 million yuan annually. Personnel expenses dominate. The firm employs 240 people. It has integrated its core research team into Alibaba Cloud. That move freed resources for the infrastructure bet.
Orders jumped from about $74 million in 2025 to $220 million in 2026, according to analyst notes shared on X by Poe Zhao. The growth reflects demand for sovereign AI deployments. Clients seek systems they control completely.
Geography matters. Roughly half the business comes from outside China. Markets stretch across Asia, Europe, South America, Central Asia, the Middle East, Southeast Asia, and Africa. 01.AI operates in Kazakhstan and formed a joint venture with Thailand’s CP Group. It avoids the U.S. market. Buyers there remain wary of Chinese software.
The pivot aligns with a broader pattern among Chinese AI startups. The group once known as the AI Tigers includes Zhipu AI, MiniMax, Moonshot AI, DeepSeek, and StepFun. Zhipu and MiniMax have listed in Hong Kong. Moonshot targets a listing within six months at a potential $30 billion valuation. DeepSeek also eyes 2027. All six either prepare for or have completed Hong Kong debuts.
01.AI will unwind its offshore holding structure. Moonshot completed a similar step in May 2026. The process clears the way for a local listing. Lee plans to close a pre-IPO funding round around the release of first annual results. The fiscal year ends in December.
Recent coverage highlights the strategy’s uniqueness. A July 21, 2026 post on X by Jeffrey Towson noted the top-down sales focus for regulated industries and isolated data pools. Yet the marketing struck him as odd. Booth displays at the Shanghai conference featured Lee’s books and cutouts for photos. “Money leopards” in the jungle served as wordplay for rapid monetization.
Another July 20, 2026 X thread from tokenandoAI summarized the pivot away from pre-training and the emphasis on enterprise and sovereign deployments. The account cited the same revenue and order figures while noting the pattern across the six leading startups.
Lee’s career adds weight to the story. He led Google’s China operations before turning to entrepreneurship and investing. His public profile remains high. Some observers question whether celebrity helps or distracts from execution. The numbers, however, point to real commercial traction.
Enterprise buyers prize control. They hesitate to send sensitive data to public clouds. On-premise infrastructure addresses that concern directly. 01.AI positions its software as the layer that makes open models useful inside corporate firewalls.
The model-building frenzy left many firms with high costs and uncertain returns. Only those with vast capital continued. Most others, including 01.AI, adapted. The adaptation looks like a bet on data infrastructure as the durable moat.
Hong Kong listings offer Chinese tech firms access to capital without U.S. regulatory hurdles. Several peers already succeeded. 01.AI’s planned 2027 debut would place it among the later movers. First annual results will provide investors a clearer financial picture.
Lee’s comments at the Shanghai conference signaled confidence. The pre-IPO round seeks to strengthen the balance sheet before listing. Details on size and participants remain undisclosed. Yet the timeline ties funding directly to audited performance.
Analysts watch whether the Palantir analogy holds. Palantir built its reputation on complex data analytics for government and enterprise clients. 01.AI follows a parallel path but in a market shaped by China’s regulatory environment and data localization rules.
The company no longer competes on model leaderboards. Benchmarks matter less than deployment speed and security. Clients want solutions that work with their existing data pools. Boss AI delivers exactly that.
Still, challenges persist. Competition among the remaining AI Tigers stays fierce. Moonshot’s higher valuation target reflects different ambitions. 01.AI pursues profitability and scale through services rather than raw model size.
Recent Bloomberg reporting on July 20, 2026 confirmed the IPO timeline and pre-IPO plans without revealing specific fundraising targets. The story reinforced Lee’s shift from model creator to infrastructure provider.
Conversations on X reflect mixed reactions. Some praise the pragmatic strategy. Others note the marketing emphasis on Lee’s personal brand. Both elements may influence investor perceptions ahead of the listing.
Ultimately 01.AI’s story captures a moment of realism in AI development. Not every company can sustain frontier training costs. Those that pivot early gain time to build defensible businesses. Lee appears to have made that turn at the right moment.
WebProNews is an iEntry Publication