Author｜Lin Yi Editor｜Key Point Jun

"Who Represents China's Anthropic: Zhipu or MiniMax?"

On January 8th of this year, Zhipu successfully went public on the Hong Kong Stock Exchange. The following day, January 9th, MiniMax followed suit and completed its listing. Both companies are currently unprofitable, with R&D investments significantly outpacing their revenues. Nevertheless, their post-listing performance in the capital market has been outstanding, with their combined market capitalization now reaching the hundred-billion-dollar mark.

This valuation far surpasses that of their overseas industry counterparts. Earlier this year, Anthropic achieved a valuation of $38 billion after its Series G funding, with annualized revenue of $1.4 billion, corresponding to a price-to-sales (P/S) ratio of approximately 27x. In contrast, Zhipu reported revenue of RMB 724 million in 2025, with a peak market capitalization of around $5.1 billion, resulting in a P/S ratio exceeding 500x. MiniMax, on the other hand, reported 2025 revenue of $79 million, with a peak market capitalization of approximately $4.9 billion, translating to a P/S ratio over 600x.

Several factors contribute to this significant premium, including the direct benefits from the OpenClaw and Lobster trends; a scarcity of free float, with 60% of shares locked in strategic placements and minimal liquidity in secondary markets; and market enthusiasm for the narrative of the "Chinese Anthropic."

Both MiniMax and Zhipu have released their first annual reports post-listing. By analyzing their current financial structures, business models, and growth quality, we can explore which company is closer to Anthropic in its commercial approach.

To B Delivery vs. To C Subscription

Both Zhipu and MiniMax are vigorously developing large models, but their revenue structures are nearly polar opposites.

Zhipu primarily focuses on the domestic government and enterprise market, deriving the majority of its revenue from Chinese government and corporate clients. In 2025, 73.7% of its revenue (approximately RMB 534 million) came from privatized deployments, often highly customized to meet industry and enterprise-specific needs. In terms of market penetration, nine of the top ten domestic internet companies have deeply integrated Zhipu's GLM series models. Total annual revenue reached RMB 724 million, up 131.9% year-on-year.

MiniMax, however, targets overseas consumer markets and creator ecosystems, with core revenue derived from international users. In 2025, total revenue reached $79 million, up 158.9% year-on-year. Of this, 73.0% ($57.66 million) came from international markets outside mainland China. Its product portfolio heavily relies on AI-native applications, accounting for 67.2%, primarily driven by products like the AI role-playing app Talkie and the video generation tool Hailuo AI.

While both companies are similar in scale, their structures differ entirely: one operates domestically with a ToB privatized delivery model, while the other serves overseas consumers through platform subscriptions. This divergence directly shapes their distinct financial performances.

When assessing the commercial health of large model companies, gross margin trends are more meaningful than pure revenue growth.

Zhipu's gross margins are under downward pressure. Its overall gross margin fell from 56.3% in 2024 to 41.0% in 2025. The gross margin for its core privatized deployment business dropped even more sharply, from 66.0% to 48.8%, primarily due to increased delivery resources invested to meet customer demands. This delivery-heavy model entails extremely high labor costs. In 2025, Zhipu's sales and marketing expenses reached RMB 391 million, nearly double its annual revenue from cloud deployment services.

It should be emphasized that in China's enterprise-grade AI market, pricing power for computing resources lies with major cloud players like Alibaba Cloud, Baidu Cloud, and Volcano Engine, which boast vast cloud ecosystems. These giants not only possess self-developed chips, cloud platforms, and models but also cultivate larger customer and developer ecosystems. Even if independent large model companies optimize inference costs for a specific model, major players quickly follow suit. In the long run, independent large model companies are unlikely to compete with these giants on token costs.

MiniMax's gross margin trend tells a different story. Its overall gross margin improved from 12.2% in 2024 to 25.4% in 2025. More critically, while revenue surged, its sales and distribution expenses plummeted 40.3% year-on-year to $51.9 million, thanks to organic growth driven by word-of-mouth for its AI-native products, reducing the need for promotional spending. Its annualized recurring revenue (ARR) soared from $100 million to $150 million in just two months.

Gross margin trends matter more than absolute values. While Zhipu's 41.0% gross margin currently tops MiniMax's 25.4%, the former faces pressures from diseconomies of scale, while the latter demonstrates the network effects characteristic of software subscription models.

Domestic Market vs. Overseas Expansion to Avoid Competition

Zhipu has chosen to focus on China's government and enterprise AI procurement market. By the end of 2025, Zhipu's total accounts receivable stood at approximately RMB 339 million, accounting for nearly 47% of its annual RMB 724 million revenue. Additionally, in 2024, a single client contributed nearly RMB 59.47 million in revenue, representing about 19% of total revenue that year, highlighting significant customer concentration and payment term issues. The privatized deployment business requires extensive customization and local delivery, with revenue and labor costs growing linearly, making it difficult to benefit from the marginal cost reductions typical of AI technologies.

Zhipu's management is clearly aware of these challenges. In 2025, the company proactively optimized its business structure, with revenue from open platforms and API services surging by 292.6%. Cloud deployment revenue's share of total revenue rose from 15.5% to 26.3%. However, the gross margin for cloud services currently stands at just 18.9%. Transitioning to the cloud means enduring short-term gross margin pain while fending off price wars from tech giants like ByteDance and Alibaba.

MiniMax has positioned its main battlefield overseas, avoiding the intense competition in China's ToB market. Its 73.0% overseas revenue ratio not only sidesteps price wars but also grants access to user groups with stronger willingness to pay. Talkie and Hailuo AI have validated the monetization logic for AI companions and video generation scenarios. The absolute decline in sales expenses alongside exponential revenue growth are hallmarks of early-stage product-driven SaaS, aligning more closely with Anthropic's dual-wheel model of C-end and developer-driven growth.

However, overseas expansion carries risks. When facing traditional content giants like Hollywood, copyright issues loom as a Sword of Damocles. Disney, Universal, and Warner Bros. have sued MiniMax, alleging that Hailuo AI lacks basic safeguards—for example, when users input "Darth Vader," the system directly generates infringing videos. The plaintiffs seek damages based on 500 works at up to $150,000 each, potentially totaling $75 million.

MiniMax disclosed in its financial report that some subsidiaries face legal disputes over intellectual property infringement claims. The cases are in their early stages, with outcomes impossible to predict accurately, representing a significant external risk variable.

Yuezhi Anmian is also planning to list. According to media reports, its overseas revenue grew strongly in the first quarter of this year, suggesting its overall revenue mix may resemble MiniMax's.

MiniMax Resembles Anthropic More Than Zhipu

Both Zhipu and MiniMax invest heavily in R&D to achieve technological advantages.

Zhipu's 2025 R&D expenses reached RMB 3.18 billion, up 44.9% year-on-year, resulting in an adjusted net loss of RMB 3.182 billion. For every RMB 1 in revenue, R&D investment amounts to approximately RMB 4.4. Zhipu raised about HK$5 billion in its January IPO, replenishing its resources, but at the current burn rate, capital chain pressures persist. With cash reserves of approximately RMB 2.259 billion, assuming no revenue growth, its cash flow would last less than 12 months, likely necessitating additional financing shortly. In terms of personnel efficiency, Zhipu employs 1,094 full-time staff.

MiniMax enjoys a wider financial safety margin. Its 2025 R&D expenses were $253 million, with an adjusted net loss of $251 million. Its cash, time deposits, and financial assets total $1.05 billion, providing over four years of cash runway. Its workforce is also more streamlined, with just 428 full-time employees, demonstrating higher personnel efficiency while generating revenue on par with Zhipu.

Anthropic's core business model revolves around its cutting-edge Claude series models, featuring a diversified "API + subscription" revenue structure that maintains strong enterprise and developer loyalty.

In terms of model capabilities, both Zhipu and MiniMax demonstrate strong competitiveness. Zhipu's confidence in raising API call prices by 83% in early 2026 without dampening demand suggests it has begun to wield pricing power.

However, from a business model evolution perspective, MiniMax aligns more closely with Anthropic's prototype. MiniMax's steadily improving gross margins, dominance in overseas C-end subscription revenue, transition toward product-driven sales, leaner organizational structure, and stronger cash reserves all point to a healthy, scalable business.

Zhipu's moat lies in its deep integration with China's government and enterprise markets and its strong industry-academia-research background. However, its privatized deployment-driven business resembles traditional IT integrators rather than pure SaaS or MaaS enterprises.

Whether it's Zhipu, MiniMax, or Yuezhi Anmian, surviving the AGI race amid competition from tech giants requires continuous investment in this endless R&D arms race.

Note: The views expressed herein are solely those of the author and do not constitute investment advice. This article has not been endorsed by any related companies.