One of China's leading artificial intelligence startups is closing in on a milestone that the rest of the industry has struggled to achieve: $1 billion in annual sales. Z.ai, formerly known as Zhipu and the creator of the GLM family of large language models, is on track to become the first independent Chinese AI firm to reach that revenue figure, according to a report by Bloomberg. The projection is based on annualized recurring revenue and rapid growth in enterprise and cloud markets, but it underscores a fundamentally different approach to monetizing AI than the one adopted by most Western labs.
The Numbers Behind the Milestone
Z.ai's reported revenue for 2025 stood at approximately 724 million yuan, or about $100 million, representing a 132% year-over-year increase. While the base is modest, the growth trajectory is steep. JPMorgan Chase analysts project that Z.ai's revenue will reach around 4.6 billion yuan in 2026 and surge to 30.9 billion yuan by 2028, the year they expect the company to finally turn a profit. The $1 billion milestone is, however, a forward-looking estimate that leans partly on annualized recurring revenue (ARR) — a run-rate snapshot that may not fully reflect booked sales. Even JPMorgan's 2026 forecast, at roughly $630 million, sits below the billion-dollar mark in dollar terms.
Yet the trajectory behind those numbers is real. Z.ai has managed to scale revenue while simultaneously giving away its most capable models as open-source software — a paradox that has confounded many Western AI companies. Most startups in the United States, such as OpenAI and Anthropic, rely on proprietary models and massive venture capital funding to cover losses that often exceed $1 billion per year. Z.ai, by contrast, is building a revenue engine that, while still unprofitable, is approaching a scale that suggests commercial viability.
How Z.ai Makes Money
The company's revenue mix is heavily tilted toward the enterprise sector. A significant share comes from on-premises deployments for state-owned enterprises and financial institutions, which require secure, locally hosted AI solutions. These contracts often involve customizing the GLM models for specific use cases, from customer service to data analysis, and include ongoing support and maintenance fees. Additionally, Z.ai operates a fast-growing cloud business that allows organizations to access its models via APIs, with pricing that undercuts many Western competitors.
The API side has become the most dramatic growth driver. Z.ai's open platform, where developers can access the GLM models through a paid API, has seen its annualized recurring revenue hit 1.7 billion yuan, up a staggering sixtyfold in a single year. This growth is fueled by the company's aggressive pricing strategy and the high quality of its models, which have consistently ranked among the top performers in Chinese benchmarks for natural language understanding and generation.
The Open-Source Paradox
Here is the part that defies the conventional playbook. Z.ai releases its most capable models, including the latest GLM-5.2, as open-source software that anyone can download and run for free. The conventional wisdom in AI holds that giving away the core product destroys the ability to charge for it. Z.ai's founder, Tang Jie, has publicly defended the opposite approach, arguing that frontier AI should remain open to everyone to foster innovation and trust. The revenue figures suggest that open-sourcing the best models can actually drive adoption, which in turn fuels demand for paid services like cloud hosting, technical support, and on-premises deployments.
This strategy mirrors what companies like Red Hat and MongoDB have done in the software world: offer a free, open-source core and monetize through enterprise-grade services. But in the AI industry, where training costs can run into the hundreds of millions, many labs have opted for proprietary models to protect their investments. Z.ai is betting that commoditizing the model itself will create a larger ecosystem of users and developers, ultimately leading to more revenue from auxiliary services.
Why It Matters for the Global AI Race
Z.ai's success is a case study in China's approach to AI commercialization. While American companies have focused on building ever-larger models and securing exclusive access to data and compute, Chinese firms have emphasized speed, scale, and cost efficiency. The Chinese government has also played a role by subsidizing AI research and encouraging state-owned enterprises to adopt domestic AI solutions. This has created a ready market for Z.ai's offerings, even as the company faces intense competition from other Chinese players like Baidu’s Ernie Bot, Alibaba’s Tongyi Qianwen, and Tencent’s Hunyuan.
Bloomberg notes that Z.ai’s revenue projection relies partly on state-owned buyers, which blurs the line between genuine commercial demand and government support. Critics argue that such revenue is not purely market-driven, but supporters point out that the government is a legitimate customer, just like any large enterprise in the West. The company has also raised billions of dollars in follow-on share sales, with a valuation that has soared to roughly $112 billion — a rally of over 1,000% since its January 2025 listing on a Chinese stock exchange. That valuation assumes the ambitious projections come true rather than reflecting current fundamentals.
Challenges Ahead
Despite the impressive growth, Z.ai faces significant headwinds. The Chinese AI market is brutally competitive, with many startups and tech giants undercutting each other on price. The open-source model may drive adoption, but it also means that competitors can freely reuse Z.ai’s technology, potentially cannibalizing its own revenue. Profit margins remain thin, and the company is still lossmaking: its losses have continued to climb even as revenue soars, a pattern common across the AI industry.
Moreover, the billion-dollar projection is just that — a projection. If growth slows or if a major competitor releases a significantly better model, Z.ai could struggle to meet expectations. The company’s heavy reliance on state-owned clients also exposes it to political risks, such as changes in government procurement policies or export controls that could limit access to essential hardware like advanced GPUs from Nvidia.
Nevertheless, Z.ai’s trajectory offers a glimpse of a future where AI companies can achieve significant revenue without locking their best technology behind paywalls. The open-source paradox may prove to be a sustainable model, especially in markets where cost sensitivity is high and trust in proprietary systems is low. If Z.ai can maintain its growth rate and eventually turn a profit, it will have demonstrated an alternative path to commercialization that the West has largely ignored.