Is Zhipu really reliable?

This image is AI-generated

Earnings Report Summary:

Zhipu (02513.HK) released its first annual financial report since listing. In 2025, it generated total revenue of RMB 724 million, up 131.9%. Its net profit, however, suffered a large loss of RMB 4.72B, but after adjustments the net loss for the year was RMB 3.18 billion, widening 29.1% year over year. During the reporting period, Zhipu continued to expand its R&D investment. Its R&D expenses were RMB 3.18 billion, which is the main reason for the loss.

In terms of its principal business, last year Zhipu’s consolidated gross margin was 41%. By business segment, its enterprise-level general-purpose large model business revenue was RMB 366 million, accounting for 50.4% of total revenue, and remains the core pillar. Revenue from its enterprise-level intelligent agent business was RMB 166 million, up 248.8% year over year; its revenue share increased to 22.9%, becoming a new growth engine. The open platform and API (Token sales) business was the most impressive: revenue was RMB 190 million, with a surge of 292.6%, and its revenue share reached 26.3%.

Main viewpoints:

1. Revenue and Business: Early revenue figures don’t have to be taken too seriously—mainly look at expectations for scenario-driven breakout

At this stage, the new species’ significance is not very meaningful if you only look at its revenue scale and gross profit margin—after all, it is still very early, and commercialization is just getting started.

However, it is important to pay close attention to this figure: Zhipu’s full-year revenue last year was only RMB 724 million, but its accounts receivable were as high as RMB 699 million. What does that indicate? The quality of revenue likely is not very high—there may even be a lot of “water.” This is also how Sunrise did it early on: first, make the revenue numbers look good (using accounts receivable to expand revenue). As for bad debts, deal with them later (more than the last third of accounts receivable became bad debts). In that situation, is that business revenue really “revenue created based on real business, normal commercial logic, and meaningful external contracts”? Judging from Sunrise’s early data, whether it is the revenue recognition criteria or the seriousness of cooperation contracts, there should be clear issues. Would Zhipu also be like Sunrise? Most likely, it will—though it may just be a matter of degree!

But if that is the case, is Zhipu unreliable? Not necessarily. For this disruptive AI-native new “species,” in the early stage—while cultivating customers and educating the market—it can only do things this way. The good news is that now, with the breakout of enterprise-level and even individual AI applications such as “lobsters,” opportunities for foundational large-scale vendors like Zhipu will come very quickly as well—which is also why it is worth 300 billion yuan!

However, here we need to make up for it based on objective circumstances: Zhipu should be much more reliable than Sunrise early on. Sunrise had a large amount of hardware integration business in its early stage (so revenue could be booked quickly), whereas Zhipu mainly focuses on technology and API-calling services. Therefore, even if there is “water” in the RMB 700 million revenue, in theory it should be far smaller than Sunrise’s.

2. Financial Soundness: On-hand funds may be enough to burn for another two years; a 300 billion market cap creates a huge financing leverage

For a new “species” like Zhipu, besides revenue data and expectations for scenario breakout, there is an even more important issue: financial soundness—given the current burn rate and combined with business growth expectations, can the company’s capital chain hold up?

First, take a look at its losses: last year it recorded a book loss of RMB 4.72B, but the loss under the adjusted accounting basis was RMB 3.18B. Currently, Zhipu has only released its performance report and has not published a complete audited annual report. The outside world therefore cannot see its complete statement of cash flows, but based on the adjusted basis in the table above, the adjusted loss of about RMB 3.18B also allows us to judge the magnitude of cash burn on the operating side—because the adjusted figures include asset items that need to be amortized and depreciated, and the balances of assets (including fixed assets) are not large. Taking into account its revenue scale and accounts receivable items, its operating cash flow deficit estimate may also be around RMB 3 billion.

Next, look at its cash reserves on the books: as of the end of last year, Zhipu’s readily convertible quasi-cash reserves were about RMB 2.7 billion (bank deposits and cash of RMB 2.26B + time deposits of RMB 108 million + short-term investments measured at fair value of RMB 377 million). However, this is only a snapshot as of December 31, 2025. In fact, Zhipu completed its IPO issuance in January 2026, and the final net proceeds raised were about HKD 4.7 billion (including the green shoe option), which is roughly RMB 4.3 billion. That means, in theory, after Zhipu’s IPO issuance, its on-hand quasi-cash reserves could be around RMB 7.0 billion.

Based on the two estimation figures above (operating net cash burn of about RMB 3.0 billion in 2025 vs. quasi-cash reserves on hand of about RMB 7.0 billion), the funds on Zhipu’s books today may be enough to burn for roughly two years. Considering the dynamic expansion of R&D investment, it may not even be enough for two years. Still, the good news is that Zhipu has a market cap of 300 billion yuan. In theory, there is a very high equity financing leverage, and the difficulty of raising additional capital through the capital markets is not big. Therefore, in theory: as long as Zhipu’s story remains clearly visible, the pressure on its capital chain also should not be too hard to ease.

After we clarify these two key technical financial issues above, what remains is the matter of Zhipu’s business and scenario implementation. We can discuss that problem later if we have the opportunity. Everyone is also welcome to discuss it! END

Lentil Company Research****Classic Review-