Li Kaifu’s Zero One (零一万物) pivots to “profitable AI 2.0,” eyes 2027 IPO and heavy employee incentives

Turnaround pitch: results over model worship

Li Kaifu (李开复) is repositioning Zero One (零一万物) from a bleeding-edge large‑model hopeful into a commercial, revenue‑driven AI company. It has been reported that Li disclosed plans in an internal anniversary speech to push Zero One to become “China’s first profitable AI 2.0 company,” with an ambition to file for IPO in 2027 and an immediate allocation of 20 million stock options to staff. The shift is concrete: Zero One has reportedly moved most pre‑training and AI‑infra teams into Alibaba (阿里巴巴), abandoned the race for ever‑larger foundational models, and doubled down on To‑B sales and deployable products that generate recurring revenue.

Industry and geopolitical context

China’s large‑model sector is fragmenting. Some firms continue heavy investment in pre‑training and GPU scale — for example Zhipu AI (智谱) and MiniMax — while others, Zero One among them, have pivoted toward industry applications and commercialization. The change is not purely strategic: export controls and tighter US‑led restrictions on advanced AI chips have made the high‑cost, GPU‑heavy path riskier. At the same time, the launch of inference‑focused models such as DeepSeek‑R1 accelerated market pressure to show business outcomes rather than research milestones. Can business realism outpace technological idealism? Li is betting yes.

Incentives, structure and the new culture of “DRI”

Internally Zero One has instituted what it calls a “second founding”: a DRI (directly responsible individual) structure, quarterly performance reviews tied to entrepreneurial output, and a new partner program. It has been reported that the company set targets of RMB 500 million in 2025 orders and RMB 250 million in audited revenue, and that orders had reportedly climbed to over RMB 1.5 billion by May 2026. The firm is pushing engineers and sales—and Li himself is out courting large clients in Central Asia, the Middle East and Africa—to win long‑term ARR and joint ventures rather than one‑off demos. Li frames the change bluntly: “bringing clients real money matters more than any valuation.”

Zero One’s turn toward industrial AI — a model closer to Palantir’s playbook, Li says — illustrates a broader theme in China’s AI ecosystem: survival now demands monetization, not just model scale. It has been reported that new CEO incentive pools and a roughly 20‑million‑share option grant are meant to align staff to that reality. For Western investors watching China’s AI race, the question is whether such commercialization will produce sustainable profits in a geopolitical environment where access to top‑end hardware and international markets is increasingly politicized.