AI Relay Stations: A Business So Profitable It's Scary
Right now, the hottest topic in the crypto space is AI relay stations. People say this business is insanely profitable, and it's no exaggeration. Just look at who's in this game—famous figures like Brother Sun from the crypto world are involved, the former boss of Cheetah Mobile, Fu Sheng, is in on it, and the most unbelievable part is that even the Trump family has recently jumped in, showing just how lucrative this venture is.
But how should we view AI relay stations?
The business model of AI relay stations essentially revolves around the “middleman model” that exploits mismatches in supply and demand. It survives on information asymmetry and resource disparities, offering short-term profits but with significant long-term risks. It has elements of both legitimacy and gray areas, which can be objectively viewed from three points:
Existence of Legitimacy: Domestic users currently face many barriers when trying to access high-quality foreign AI (like Claude and GPT)—network restrictions, payment barriers, IP bans, and high subscription costs with complicated operations. Relay stations purchase computing power in bulk, integrate multiple models, and create user-friendly tools, allowing ordinary folks to access global AI at low costs and with ease. This “filling the gaps” positioning is the core reason for its rapid rise and has generated real market demand.
Layered Profit Models and Chaos: The more legitimate ones earn reasonable price differentials, buying computing power in bulk to get discounts, splitting official accounts for sharing, making “hard-earned arbitrage money.” However, many relay stations take a gray route, even resorting to switching out models and using low-quality factory models to impersonate the latest Claude and GPT versions, relying on users' limited AI understanding to get by.
Risks and Limitations Not to Be Ignored: Upstream model manufacturers (like Anthropic) have strengthened risk control, cracking down on behaviors like interface wrapping and account sharing. In early 2026, there was a situation where 80% of relay service providers got their accounts banned, crippling their operations. Relay stations relying on gray paths are extremely prone to collapse. On the other hand, issues like model switching and data leaks not only violate user rights but also potentially breach unfair competition laws and related regulations. The technical barriers are quite low, as multiple open-source frameworks can be set up, leading to intense homogenous competition. With tightening compliance policies, long-term survival space is limited, making it more like a “short-term business” born from a temporal windfall rather than a sustainable model.
Right now, the hottest topic in the crypto space is AI relay stations. People say this business is insanely profitable, and it's no exaggeration. Just look at who's in this game—famous figures like Brother Sun from the crypto world are involved, the former boss of Cheetah Mobile, Fu Sheng, is in on it, and the most unbelievable part is that even the Trump family has recently jumped in, showing just how lucrative this venture is.
But how should we view AI relay stations?
The business model of AI relay stations essentially revolves around the “middleman model” that exploits mismatches in supply and demand. It survives on information asymmetry and resource disparities, offering short-term profits but with significant long-term risks. It has elements of both legitimacy and gray areas, which can be objectively viewed from three points:
Existence of Legitimacy: Domestic users currently face many barriers when trying to access high-quality foreign AI (like Claude and GPT)—network restrictions, payment barriers, IP bans, and high subscription costs with complicated operations. Relay stations purchase computing power in bulk, integrate multiple models, and create user-friendly tools, allowing ordinary folks to access global AI at low costs and with ease. This “filling the gaps” positioning is the core reason for its rapid rise and has generated real market demand.
Layered Profit Models and Chaos: The more legitimate ones earn reasonable price differentials, buying computing power in bulk to get discounts, splitting official accounts for sharing, making “hard-earned arbitrage money.” However, many relay stations take a gray route, even resorting to switching out models and using low-quality factory models to impersonate the latest Claude and GPT versions, relying on users' limited AI understanding to get by.
Risks and Limitations Not to Be Ignored: Upstream model manufacturers (like Anthropic) have strengthened risk control, cracking down on behaviors like interface wrapping and account sharing. In early 2026, there was a situation where 80% of relay service providers got their accounts banned, crippling their operations. Relay stations relying on gray paths are extremely prone to collapse. On the other hand, issues like model switching and data leaks not only violate user rights but also potentially breach unfair competition laws and related regulations. The technical barriers are quite low, as multiple open-source frameworks can be set up, leading to intense homogenous competition. With tightening compliance policies, long-term survival space is limited, making it more like a “short-term business” born from a temporal windfall rather than a sustainable model.
