Summary of Key Points
Siliky Flow, a company established less than three years ago, has rapidly expanded through its AI intermediary services (adapting computing power and models). It has raised funds in seven rounds, with its valuation soaring to 7.74 billion yuan. However, its financial performance is dismal: revenue in 2025 increased by seven times to 55.33 million yuan, but the gross margin was -24% (losing 24 cents for every 1 yuan earned), resulting in a cumulative loss of 440 million yuan over three years and a monthly cash burn of 14.8 million yuan. Now, it is applying to list on the Hong Kong Stock Exchange in an attempt to survive, but it faces multiple crises, including competition from large companies, low user loyalty, and a difficult business model to make profits. Will listing be a lifeline or merely accelerate its demise?
Business Model: The “Intermediary Translator” in the AI Ecosystem
Siliky Flow does not focus on developing large models (such as ChatGPT) nor C-side apps; instead, it acts as a bridge between computing power and models. In simple terms:
- Its upstream partners include various chips (NVIDIA, Huawei Ascend, etc.) and AI models with different architectures;
- It uses its proprietary tools to adapt these incompatible components into standardized “Tokens” (the unit of measurement for AI services, similar to data packages);
- It sells these tokens to enterprises and developers, helping them with AI deployment, adaptation, and cost reduction.
For example, when the DeepSeek model became popular in early 2025 and official servers were overwhelmed, Siliky Flow used its domestic chip adaptation technology and collaborated with Huawei Cloud to handle the excess traffic. It also gained fame by offering a free token promotion.
Financial Reality: The Faster Growth, the More Severe Losses
On the surface, the numbers look impressive: revenue increased from 7.34 million yuan in 2024 to 55.33 million yuan in 2025 (a sevenfold increase), and the number of registered users grew from 127,000 to 10.28 million, with its token throughput ranking fourth in China. However, this growth came at a heavy cost:
- Gross Margin Plunge: The gross margin plummeted from 39.4% in 2024 to -24% in 2025 (losing 1.24 yuan for every 1 yuan of revenue from token sales); the public cloud business was even worse, with a gross margin of -119%;
- Exponential Cash Burn: The company suffered a net loss of 440 million yuan over three years and had a monthly cash burn of 14.8 million yuan in 2025, with operating cash flows flowing out by 172 million yuan;
- Growth at the Cost of Subsidies: In 2025, it spent 54.21 million yuan on distributing free tokens (almost equivalent to its annual revenue), but users quickly left—the number of public cloud users dropped from 5.45 million to 1.45 million in the first four months of 2026.
Competition from Large Companies: The Space for Independent Third Parties is Narrowing
Siliky Flow’s profit comes from the difference between computing power and model prices, but this niche is being encroached upon by giants:
- Large Companies Building Their Own Intermediary Services: Alibaba Cloud has established a Token Foundry division (managed by its CEO), and companies like ByteDance, Baidu, and Tencent are increasing their investments to take over these services;
- Price Wars: Leading manufacturers have reduced prices multiple times, with some model tokens decreasing by 90%, and DeepSeek V4-Pro by 75%. Alibaba Cloud even further lowered prices by 97.5%. While large companies have other sources of revenue (e.g., from e-commerce), Siliky Flow can only suffer losses or lose customers;
- Mobile Operators Entering the Market: Mobile and Unicom have launched token packages, making AI services more accessible through data plans, further squeezing independent third parties.
Currently, Siliky Flow holds only 1.5% of the market share, while the top three companies control 60%. Breaking through this barrier is extremely difficult.
Model Challenges: Users Driven by Subsidies Lack Loyalty
Siliky Flow’s strategy was to attract users with free tokens and then convert them into paying enterprise customers, but reality is harsh:
- User Behavior: Subsidies brought in millions of users, most of whom are price-sensitive and leave as prices rise. In the first four months of 2026, the number of high-value customers (purchasing dedicated computing instances) dropped from 49 to 20, and local deployment customers decreased from 28 to 20;
- Low Conversion Rate of Paid Users: Although the penetration rate of paid public cloud services increased from 13% to 44%, most transactions were small, and the pool of high-value customers is shrinking. Although the average transaction value has increased, the total volume is too low to sustain profitability;
- Insufficient Scale Effect: Public cloud APIs are becoming a “bulk data sales” model, and only extremely large companies can make a marginal profit. Siliky Flow’s scale is far from being significant enough.
Listing: A Lifeline or a Death Knell?
Siliky Flow’s listing coincides with the Hong Kong Stock Exchange’s 18C regulations for specialized technology companies (with a valuation of 7.74 billion yuan, approximately 800 million Hong Kong dollars). The funds raised could temporarily boost its cash flow, but there are significant risks:
- Limited Investor Patience: Interest in the AI infrastructure sector has cooled down (e.g., SenseTime’s stock value dropped by 80% after listing). Will investors continue to support losses in exchange for growth?
- Risk of High Customer Concentration: The top five customers account for 60% of revenue, and a single major customer could cause a collapse if they stop using the service;
- Stock Price Pressure: If losses persist after listing, a drop in stock price could demoralize the team and affect future financing.
In conclusion, Siliky Flow’s struggles reflect the challenges faced by independent AI intermediaries: even with strong technology, they cannot overcome the pressure from giants and price wars. Listing may provide a temporary solution, but the key to survival lies in finding a sustainable business model. The fate of such companies often depends on external factors—market conditions, capital availability, and new competitors.