AI Infrastructure Firm SiliconFlow Seeks Hong Kong IPO Amidst Heavy Losses
SiliconFlow, an AI inference infrastructure company founded less than three years ago by AI veteran Yuan Jinhui, has filed for an IPO on the Hong Kong Stock Exchange under the 18C category for pre-commercialization companies. The company aims to become a leading open and independent token supply platform in China, aggregating computing power from various providers like Nvidia and AMD, and translating it into standardized token supply. Despite rapid user growth, reaching over 10 million registered users and a significant increase in daily token throughput, SiliconFlow faces substantial financial challenges. Its prospectus reveals a stark reality behind its impressive growth metrics: the company is selling tokens at a loss, with a projected gross margin of -24% for 2025. The cost of sales is expected to exceed revenue, largely due to the high cost of leased computing resources, which constituted 86.9% of sales costs in 2025. Research and development expenses are also significant, projected to be 378.1% of revenue in 2025, alongside sales and marketing expenses at 151.4% of revenue. Consequently, SiliconFlow anticipates substantial net losses, with a projected net loss rate of -624.4% in 2025, amounting to a net loss of 345 million yuan. The company has raised seven rounds of funding, significantly increasing its valuation from 280 million yuan at the angel round to 7.74 billion yuan at the B+ round just before its IPO filing. Key investors include Alibaba, Huawei, and Source Code Capital, among others. Yuan Jinhui, who previously founded OneFlow, leads SiliconFlow, holding significant voting power through various entities. The company's reliance on the availability and licensing terms of open-source AI models presents a notable risk.
SiliconFlow's IPO filing highlights a critical tension in the current AI infrastructure landscape: the pursuit of rapid market share and user acquisition versus sustainable profitability. The company's strategy of offering tokens below cost, driven by intense price competition and the high cost of leased computing power, presents a significant financial challenge. While the substantial user growth and valuation increases reflect strong investor confidence in the AI sector's potential, the projected negative gross margins and escalating net losses raise questions about the long-term viability of this business model. The company's heavy reliance on leased resources and open-source models also introduces supply chain and intellectual property risks. Future success will likely depend on its ability to secure more cost-effective computing resources, optimize its operational efficiency, and potentially shift towards higher-margin services or a more sustainable pricing strategy, all while navigating the evolving regulatory and competitive AI ecosystem.
AI-generated to prompt reflection — not editorial opinion, not advice, not a statement of fact. How this works.