Hello! I'm your financial analysis assistant. This article about the listing of Suiyuan Technology contains a wealth of information and touches on the most critical issues in the current domestic AI chip industry: "Who is paying for it?" and "Is this business actually profitable?"
To help you understand it easily, I've broken down this long article into a core summary and a detailed analysis from five perspectives.
---
📝 Core Content Summary
In one sentence:
Suiyuan Technology, the last of the "four major domestic AI chip companies" to go public, may have seen a significant increase in its stock price on the first day due to its association with Tencent, but it is essentially a company that relies on Tencent for revenue, sells low-margin inference cards, and has not yet turned a profit.
Key contradictions:
- Tencent's logic vs. Investors' logic: Tencent is buying into Suiyuan for supply chain security (it's willing to invest even at a loss to ensure it has access to domestic chips); investors, on the other hand, are looking for profitability and independent market competitiveness.
- High growth vs. Low quality: Revenue is growing rapidly (mainly due to Tencent), but the gross margin is low (because it mainly sells inexpensive inference cards, and its major customer is a shareholder with strong bargaining power), and the company is heavily dependent on this single customer.
Conclusion:
Suiyuan Technology has significant industrial value, but its current stock price has already anticipated its performance for the next few years. Ordinary investors should not rush to buy at high prices and should wait for three key signals: diversification of customers, upgrading of products, and positive profit margins.
---
🔍 In-Depth Analysis: Five Perspectives on Suiyuan Technology
1. Why is Tencent's investment increasing? — It's more about "insurance" than just stocks
Many retail investors might think that Tencent's 20% stake and an additional investment of 248 million yuan, locked in for three years, indicates a positive outlook, so they should buy in too. However, there's a big misunderstanding: Tencent is thinking strategically, while investors are thinking financially.
- Tencent's perspective: In the AI era, computing power is as essential as oil. Previously, companies relied on NVIDIA chips, but now, due to external factors, they can't get enough or the prices are too high. As a cloud giant, Tencent needs to have controllable domestic computing power to avoid cost constraints on its cloud services.
- Suiyuan's role: Suiyuan is a crucial link in Tencent's "domestic computing power supply chain."
- Capital tie: Tencent is one of the largest shareholders.
- Business integration: Cooperation started in 2019, with chips tailored for Tencent's use cases (such as WeChat voice and ad recommendations).
Governance involvement: Tencent representatives sit on Suiyuan's board and participate in decision-making.
- What does the 248 million yuan represent? The article calls it a "supply chain maintenance fee." By locking in the investment for three years, Tencent ensures a long-term partnership, providing Suiyuan with stability. For Tencent, it's a way to secure priority supply.
- Implication for investors: Tencent is willing to pay a premium for certainty, but secondary market investors need immediate returns or clear profit growth. Tencent's strategy doesn't directly apply to them.
2. The numbers look good, but the growth is largely driven by Tencent
Suiyuan's stock price soared on the first day of listing, reflecting market optimism about its growth potential. The numbers are impressive: revenue is expected to more than triple by 2025-2026, with a 279% increase in the first half of 2026.
- The truth behind the growth: 83.79% of revenue comes from Tencent. This means Suiyuan's growth is largely due to Tencent's own increasing AI demands (Tencent's capital expenditure increased by 176% in Q2 2026).
- Customer diversity: Other major customers change annually, indicating that Suiyuan hasn't yet established a solid presence outside of Tencent's ecosystem.
- Risks: This growth is passive; if Tencent decides to develop its own chips or adjusts its procurement strategy, Suiyuan's revenue could plummet.
- Implication: Just because Tencent holds a large stake doesn't mean Suiyuan's products are easily sold. Suiyuan is more of an internal supplier to Tencent rather than an independent brand.
3. High sales but no profit? — The dilemma of "inference cards"
Despite high revenue, Suiyuan lost 632 million yuan in the first half of 2026. The reason lies in what it sells and to whom.
- Product structure: Suiyuan mainly sells inference cards for AI-generated content and recommendation algorithms, which require high efficiency but have low unit prices.
- Competitors: Competitors like Muxi and摩尔线程 sell training cards for large model training, which are more expensive (up to 200,000 yuan per card).
- Result: Selling 60,000-70,000 cards, Muxi generated 1.6 billion yuan, while Suiyuan only 990 million yuan. This leads to a low gross margin of around 30%, far below the industry average (55%+ for Cambricon).
- Bargaining power: Suiyuan uses a DSA (dedicated architecture) that's not compatible with the NVIDIA ecosystem, limiting its versatility. Additionally, since Tencent is both a shareholder and the largest customer, Suiyuan has less bargaining power.
- Implication: Suiyuan's business model relies on low margins and high sales volumes, mainly to Tencent. To improve profitability, it needs to move to more expensive training cards or combined training/inference cards, but this will take time.
4. The battle of technical approaches: DSA vs. GPGPU — Which is better?
The article discusses two technical paths for domestic AI chips, which will shape future competition:
- GPGPU (general-purpose GPUs): Examples include Muxi and Moore Threads, which are compatible with the NVIDIA ecosystem and easier for developers to use.
- DSA (dedicated architectures): Examples include Huawei Ascend, Cambricon, and Suiyuan, which are more efficient for specific tasks like inference.
- Advantages and disadvantages: GPGPUs are more versatile but face intense competition and price wars. DSAs are more efficient but have a closed ecosystem and higher development costs.
- Suiyuan's choice: Suiyuan focused on DSA for inference, which makes sense given the current demand for inference tasks (e.g., AI assistants, real-time translation).
- However: Cambricon also uses DSA but has higher margins because it sells high-value training cards to non-shareholder companies like ByteDance. Suiyuan's DSA approach, due to its reliance on Tencent, limits its profit margins.
5. Investment guidance: Should you buy now? Watch for these three signals
The article offers practical advice. Although Suiyuan is valuable, its current stock price (about 170 billion yuan) already reflects high expectations for the future.
- Valuation: The price-to-sales ratio (PSR) is over 170 times, indicating that the market expects perfect performance in the next few years.
- Wait for the right signals: If you plan to hold for the long term, watch for these three key indicators. Wait until at least two of them are met before considering buying:
1. Customer diversification: Does the revenue from non-Tencent customers increase from less than 20% to over 30%?
2. Product improvement: Do the revenue from fifth and sixth-generation chips and combined training/inference products rise from 1.15% to 5%, 10%, or even 20%?
3. Profitability: Does the gross margin stop declining and approach the industry average of over 40%?
💡 Summary for Laypeople
- Is Suiyuan Technology a good company? Yes, it's an important player in domestic computing power autonomy with unique technology and a strong backing from Tencent.
- Should you buy its stock now? Be cautious. The current price reflects high expectations of continued support from Tencent and future product success. Any change in Tencent's strategy or poor new product development could lead to a stock price drop.
- Strategy: Understanding Suiyuan is more important than owning it. Use it as a window to observe the development of the domestic AI chip industry. If you're optimistic about domestic computing power, you can keep an eye on it, but wait for the signals mentioned above or for the stock price to adjust to a more reasonable valuation before investing. Don't simply follow Tencent's strategic decisions.