Shenzhen AI Chip B300 Trading Turns Into Gambling: Prices Triple in 5 Months

Overview

According to Huxiu, the trading market for the AI computing power chip “B300” in Shenzhen had fallen into an extremely speculative state as of August 2026. The unit price of around 4.9 million yuan in March 2026 had risen to approximately 13 million yuan just five months later, in August. The rate of increase even surpasses the stock market movements of domestic chip makers.

The author is a former programmer who currently runs an AI computing power equipment brokerage business in Shenzhen. This time, he obtained a procurement channel for B300 through an acquaintance at a central state-owned enterprise and attempted to broker a deal with a buyer. However, in the process, he was confronted with the fact that the credit system in the B300 market has effectively collapsed.

Background of the Price Surge

There are two factors behind the rapid rise in B300 prices. One is the explosive growth in demand from domestic AI manufacturers. The other is that physical procurement channels are extremely limited. The author describes the situation as “as scarce as water in the desert,” with the majority of original suppliers continuing to respond that they have “no inventory.”

The procurement channels the author can contact largely overlap with those held by major manufacturers. His only advantage was a direct connection through a long-term partner at a central state-owned enterprise. This shows just how rare a “trustworthy channel” has become in the B300 market.

Collapse of the Credit System

Normal computing power equipment transactions proceed through a series of steps: confirming purchase intent, price negotiation, contract signing, inventory confirmation, payment, load testing, and delivery. The author likens this to “writing a program, like if A then B, else return error.” However, in the B300 market, almost every stage of this process has encountered problems.

Issues surrounding asset confirmation are typical. When a buyer requests proof from the supplier, and the author relays this to the consignor, the consignor responds by asking, “First, please send us your asset confirmation certificate.” In one reported case, after sending the certificate, communication ceased, and eventually the author found his certificate reposted in a scalpers’ group chat.

Consignors also often scrutinize the corporate scale of the buyer. A registered capital of around 20 million yuan is rejected as “too small,” and buyers are asked to prepare a larger legal entity. The author describes this as “not selling equipment, but searching for a marriage partner.”

Negotiations over deposits and penalty clauses also tend to drag on. Consignors demand a 30% advance payment, while buyers offer 10%. On penalty clauses, both sides refuse to yield in demanding mutual guarantees, and as negotiations drag on, prices rise and buyers withdraw. This pattern is repeated over and over.

The Biggest Obstacle: “Landing Price Increases”

The problem the author identifies as most fatal is the “landing price increase.” This refers to the consignor unilaterally raising the price after the contract has been signed and the buyer’s seal affixed.

As a concrete example, a consignor approached a buyer who had signed the contract and said, “This price was yesterday’s market price. Today it has gone up by 100,000, so why not add a bit more?” The author compares this to “a story of the mother-in-law demanding an additional bride price on the wedding day.”

The structural background to this problem is that consignors receive signed contracts from multiple buyers at the same time. They treat these contracts like playing cards and select the buyer who offers the highest price. As a result, many buyers are forced to withdraw from the transaction without even being able to verify the equipment.

Structural Problems in the Market

The author likens the B300 market to “gambling,” analyzing it as a structure in which resource setters act as the house and always profit. Outside investors enter seeking gains, but many brokers inside the industry exit with their credibility destroyed.

Every day in group chats, questions are posted such as “Does anyone know a trustworthy channel?” or “Does anyone know a supplier that won’t raise prices on landing?” But the author points out that “nobody can answer, because everyone is looking for the same thing.” That thing is “people who keep their promises.”

The author lists the minimum conditions required for a transaction to be concluded: a fixed price that is adhered to without changes, fulfillment of the contract, compliance with delivery schedules, and the provision of physical goods at a clear price. He positions these as “the bare minimum in any normal industry,” then adds, “In this year’s computing power market, they have become a luxury wish.”

Ripple Effects on AI Infrastructure Investment

The turmoil in the B300 market shows how rapidly AI infrastructure investment in China is expanding, and how far the supply system is lagging behind. The surge in demand is linked to the AI model development race among manufacturers, and GPU procurement is treated as a strategic resource.

Looking back on his 12 years in Shenzhen, the author says that the person who once wrote code has now been pushed into a position resembling a “swindler whose credibility is eroding.” The structural problems of the AI computing power market stem not only from technical challenges but also from a lack of trust and discipline, which are the foundations of commercial transactions.

Editorial Opinion

In the short term, the turmoil in the B300 market could directly affect the development schedules of AI models in China. Landing price increases and transaction uncertainty directly push up procurement costs, which in turn affects the prices and development speed of AI services. Against the backdrop of U.S. export controls, the more domestic procurement is relied upon, the more such structural distortions in the market will come to the surface.

In the long term, the challenge will be to build mechanisms to establish sound trading practices in the AI infrastructure market. The current Shenzhen market, with its low entry barriers and lack of price transparency, is fueling speculation. Market infrastructure development—such as certification of reliable suppliers, escrow mechanisms that guarantee contract fulfillment, and visualization of price information—should be pursued in the future.

The author of this article points out that “people who keep their promises” are the scarcest resource in the market. The suggestion that, amid the rapid expansion of demand accompanying advances in AI technology, not only the technology itself but also the surrounding commercial practices and the building of trust relationships are equally important—this is a broad issue that extends beyond China alone.

References

• ” B300交易沦为赌博！十赌九输的畸形信用生态成为常态，规矩在贪婪面前一文不值 ”, by 嗅友pcTce — 虎嗅网, 2026-08-09T08:54:21.000Z (ARR)
• Source URL: https://www.huxiu.com/article/4881781.html?f=rss