SiliconFlow Files for Hong Kong IPO: Revenue Surges 653% but Gross Margin Turns Negative, Token Factory Model Faces Profitability Test

On June 30, 2025, AI infrastructure company SiliconFlow submitted a listing application to the Hong Kong Stock Exchange, aiming to become the 'first AI Token factory stock'. According to the prospectus, the company's 2025 revenue reached RMB 55.33 million, up 653.2% year-on-year, but gross margin turned negative to -24.0%, with public cloud service gross margin at -119.0%. The company reported a net loss of RMB 345 million for the year, with an adjusted net loss of RMB 187 million. SiliconFlow positions itself as a Token factory for the AI inference era, optimizing computing power and model matching through its proprietary software platform to sell Tokens to customers. As of the end of April 2026, it had 10.28 million registered users, a daily average Token throughput of 578.5 billion, and served over 13,000 enterprise customers. According to Frost & Sullivan, by annual Token throughput in 2025, SiliconFlow ranked fourth among Token supply platforms in China and first among independent ecosystem platforms. The company has completed 7 rounds of financing, with a valuation of RMB 7.74 billion, and investors include Alibaba, Meituan, Huawei Hubble, Zhipu AI, SenseTime, among others.

Business Model and Market Positioning

SiliconFlow's core business is the 'Token factory': it leases or purchases chips, deploys open-source large models, optimizes them through its proprietary software platform, and sells Tokens to users. Revenue sources are divided into public cloud Token sales (API calls) and on-premises deployment solutions (SaaS subscriptions and dedicated services). In 2025, public cloud service revenue was RMB 29.26 million (52.9% of total), and on-premises deployment revenue was RMB 26.07 million (47.1%). As of the end of April 2026, public cloud business grew faster, while on-premises deployment saw higher average order value but fewer projects. The company emphasizes that it is not a model company but an infrastructure layer for the AI inference era, focusing on packaging and delivering heterogeneous computing power, multiple models, and enterprise needs.

Financial Data and Reasons for Loss

• Revenue and Loss: 2025 revenue was RMB 55.33 million, up 653.2% from RMB 7.346 million in 2024; net loss for the year was RMB 345 million (compared to RMB 81.915 million in 2024), adjusted net loss was RMB 187 million, operating cash outflow was RMB 172 million, and monthly cash burn rate was RMB 14.8 million.
• Gross Margin Plunge: Gross margin was 39.4% in 2024 but dropped to -24.0% in 2025. The prospectus explains that this was mainly due to the expansion of the public cloud business, where the company issued a large number of free Token vouchers to attract users, while bearing the cost of computing power leases, causing sales costs to surge from RMB 4.452 million to RMB 68.632 million.
• R&D Investment: R&D expenditure was RMB 64.5 million in 2024, increasing to RMB 209.1 million in 2025, up 224.4%. The company has about 100 employees and is technology-driven.
• Cash Flow: As of the end of 2025, the company held RMB 100 million in time deposits and RMB 172 million in cash and cash equivalents.

User and Market Data

• User Scale: Registered users grew from 127,000 at the end of 2024 to 10.2824 million at the end of April 2026.
• Platform Capability: As of April 30, 2026, daily average Token throughput was 578.5 billion, supporting over 170 models cumulatively, and serving over 13,000 enterprise customers.
• Industry Ranking: According to Frost & Sullivan, by annual Token throughput in 2025, SiliconFlow ranked fourth among Token supply platforms in China (1.5% market share) and first among independent ecosystem platforms.

Investors and Industry Chain Layout

SiliconFlow has completed 7 rounds of financing, with a valuation of RMB 7.74 billion. Investors cover multiple industry types:

• Platform and application: Alibaba, Meituan, Ctrip, Kingdee, iSoftStone, Wondershare, etc.;
• Computing power and infrastructure: Huawei Hubble, Biren Technology, Intellifusion, China Unicom, 360, etc.;
• AI models and companies: Zhipu AI, SenseTime;
• Professional investment institutions: Sinovation Ventures,耀途资本, Primavera Capital, GGV Capital, Huakong Capital, etc.;
• State-owned and industrial funds: Government-backed funds from multiple regions.

Risks and Challenges

• Limited Pricing Power: Intense price competition in the Token market restricts the company's bargaining power.
• Cost Pressure: Computing power lease costs may continue to rise, especially against the backdrop of computing power shortages.
• Technology Barriers: The long-term competitiveness of the proprietary software platform needs continuous validation, facing competition from similar vendors.
• Profit Model: Currently relies on subsidies to acquire customers; paid conversion rates need improvement, and profitability has yet to be validated by the market.