<img src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/9bb4f313-6104-4e1c-b662-0feab500b315.png?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="" original-src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/9bb4f313-6104-4e1c-b662-0feab500b315.png"/>
On March 31, KNOWLEDGE ATLAS AI released its 2025 performance report. The report shows that the company maintained a high-speed growth trend in its first year post-listing, but losses also widened alongside massive strategic R&amp;D investment.
In 2025, the company achieved total revenue of 724 million yuan, a year-on-year increase of 131.9%. Among this, cloud services and AI agent businesses have become the core growth engines. However, during the same period, the company's net loss was 4.718 billion yuan, an increase of 59.5% from 2.958 billion yuan in the previous year. The main reason for the expanded loss was a substantial increase in R&amp;D expenses, with R&amp;D investment reaching 3.18 billion yuan during the year, a year-on-year increase of 44.9%, reflecting the company's strategy of driving long-term development through technological investment.
The company has made significant progress under the guidance of its core strategy of &#34;Intelligent Upper Bound × Token Consumption Scale&#34;. Its GLM series models maintain a leading position in multiple international evaluations. On the commercialization front, the MaaS platform has exceeded 4 million registered users. After the release of the GLM-5 model, it was quickly adopted by domestic leading internet companies. It is worth noting that the company proactively increased prices for its programming package products in February 2026, and the market acceptance was good, which is considered a mark of recognition for its technological value. Furthermore, the company's business has covered 218 countries and regions globally, actively exploring global cooperation models such as &#34;Sovereign AI&#34;.
<img src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/9ad77b1b-f954-4528-943a-137d20cfed84.png?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="" original-src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/9ad77b1b-f954-4528-943a-137d20cfed84.png"/>
<h2>Cloud Services Revenue Triples, AI Agent Business Becomes New Engine</h2>
Business breakdown reveals diversified growth drivers. During the reporting period, the company optimized its revenue classification disclosure to more clearly reflect its strategic layout. Among them, the open platform and API business performed the most impressively, with revenue soaring from 48.48 million yuan to 190 million yuan, a year-on-year increase of 292.6%, mainly due to the improvement in model capabilities and the surge in call volume. This signifies the acceleration of the scaling process of its MaaS (Model as a Service) platform.
The enterprise-level AI agent business, as an emerging growth pole, saw its revenue increase from 47.49 million yuan to 166 million yuan, an increase of 248.8%. This product line aims to build autonomous intelligent systems for complex enterprise scenarios, and the explosive market demand verifies the trend of AI applications evolving in depth and automation.
Revenue from the traditional advantageous business, enterprise-level general large models (primarily using localized deployment), increased from 215 million yuan to 366 million yuan, a growth of 70.5%, still constituting half of the revenue (accounting for 50.4%).
In terms of deployment methods, cloud services grew faster than localized services. Cloud deployment revenue increased by 292.6% year-on-year, and its proportion of total revenue increased from 15.5% to 26.3%. This reflects the market's increasing acceptance of standardized and agile AI services, and it also aligns with the company's long-term strategy of promoting model-as-a-service.
<h2>Cost Structure: R&amp;D Investment Exceeds 3.1 Billion, Gross Margin Under Pressure</h2>
Behind the rapid revenue growth are equally astonishing investments and a profitability level that is under pressure at this stage. In 2025, the company's gross profit was 297 million yuan, a year-on-year increase of 68.7%. However, the overall gross margin decreased from 56.3% in 2024 to 41.0%.
The financial report explains that this is mainly due to the increased proportion of cloud deployment business and the delivery resources invested in localized deployment business to meet customer needs, leading to a phased decline in gross margin. Among them, the gross margin of enterprise-level general large models decreased from 69.6% to 47.0%, which is a significant change.
Concurrently, the company's cost of sales increased by 213.3% year-on-year to 428 million yuan, consistent with the pace of business expansion. The real &#34;money-burning&#34; part is R&amp;D. In 2025, the company's R&amp;D expenses reached a staggering 3.18 billion yuan, a year-on-year increase of 44.9%, accounting for 439% of total revenue.
The driving factors behind this are clear: increased employee costs (expanding the team and share-based payment expenses), and a surge in third-party computing power service fees paid for iterating foundational models and investing in advanced training facilities. The company is continuously increasing its investment during the sprint phase of a long-term technological race.
<h2>Technical Barriers: The Leap from Vibe Coding to Agentic Engineering</h2>
In its business review, management outlines a clear path of &#34;intelligent evolution&#34;: from Vibe Coding, to Agentic Engineering, and then to &#34;digital engineers&#34; capable of executing long-term tasks in the future. In 2025, KNOWLEDGE ATLAS claimed to be the first to complete the leap from Vibe Coding to Agentic Engineering.
To achieve this leap, the company has overcome multiple technical challenges at the base layer. Through Muon Split optimization strategies and MLA-256 improvements, it reduced the GPU memory usage for model training while maintaining performance; the dynamic sparse attention mechanism has tackled the computational challenges of long sequence inference, claiming to reduce deployment costs by 50%.
Particularly crucial is its launch of the Slime asynchronous reinforcement learning framework, designed to solve the &#34;idle running&#34; pain point when AI agents execute long-term tasks, decoupling generation from training. Coupled with its original direct double-sided importance sampling algorithm, it allows models to learn efficiently from over 10,000 real software engineering environments.

02513

KNOWLEDGE ATLAS's 2025 revenue reached 724 million yuan, a year-on-year increase of 131.9%. Revenue from the Open Platform and API (Cloud Services) increased 292.6% year-on-year to 190 million yuan; revenue from the enterprise-level AI agent business grew 248.8% year-on-year to 166 million yuan. The company's R&D investment reached 3.18 billion yuan, up 44.9% year-on-year, leading to a 59.5% year-on-year expansion in net loss to 4.718 billion yuan

- On March 31, 2025, Zhipu AI released its performance report showing significant revenue growth of 131.9% to 724 million yuan, despite a net loss of 4.718 billion yuan.  
- The increased losses, up 59.5% from the previous year, were primarily due to a substantial rise in R&D expenditures of 3.18 billion yuan, reflecting a long-term strategic focus on technological development.  
- The company's core offerings, including cloud services and intelligent systems, are showing strong growth, notably with the MaaS platform exceeding 4 million users and a significant increase in international engagement.

Wallstreetcn

KNOWLEDGE ATLAS

KNOWLEDGE ATLAS 2025 Revenue Surges 132%, Cloud Services and AI Agent Businesses Both Explode, Massive R&D Investment Leads to Widened Loss | Earnings Insights