The computing power sector is back and entering a more demanding stage

The trend of technology stocks in the last two days has been very interesting.

It is not just traditional chip stocks that have risen the most, but also cloud services, data centers, and AI infrastructure. The signal given by the market is very direct: the main line of AI computing power is not over; it has just entered a new stage.

In the past, when the market bought computing power, it was more about “capital expenses” and “GPU shortages”; but after the recent round of fluctuations, capital clearly began to change the yardstick: not by looking at the scale of computing power, but more concerned about whether this computing power can eventually be turned into customer orders, deliverables, and continuous revenue. In other words, the AI market is moving from a “seller” narrative to a stage of “who can collect rent stably”.

This change is critical.

In the previous stage, the market was willing to value “how much computing power is planned,” “how many resources are bound,” and “standing on the AI cusp.” But after a round of fluctuations, capital became more picky. It doesn't just listen to space, it doesn't just look at concepts, it starts asking a few more realistic questions: is the order real? Can the project be delivered? Can computing power be operated? Can revenue be confirmed? Can profits and cash flow keep up?

This is also what the current market values the most.

In the past, when talking about AI, the market is willing to listen to space; now, when talking about AI, the market is more willing to look at cash. Order size, customer quality, delivery pace, contract cycle, revenue recognition, and cash flow conversion are becoming core indicators of sector pricing. If a company only has a concept and no delivery, it is difficult to get far in repairing the market; however, if it has orders, projects, and revenue, once market sentiment picks up, valuation repairs tend to come faster.

Following this logic, if you look at Hong Kong stock computing power operators, you will find that the value of some companies is beginning to become more clear.

Among them, there is one company worth checking out alone.

Its focus is not simply “how much computing power planning”, but how much verifiable business foundation has been formed. According to public information, Guangdong-Hong Kong Bay Intelligent Computing's current orders have exceeded 30 billion yuan, the cumulative number of orders delivered has exceeded 15 billion yuan, and the operating computing power exceeds 50,000 P (FP16 intensive). Of the new orders disclosed, more than 95% were long-term contracts for a period of up to five years.

When you look at these numbers together, the meaning is not the same.

Ongoing orders indicate that demand was not conceived out of thin air; rather, they have settled down through contracts, projects, and customer relationships; orders delivered indicate that the company is not only accepting orders, but has the ability to implement the project; the operating computing power exceeds 50,000 P (FP16 dense) indicates that this is not an asset that has remained in the construction period, but an infrastructure that has already entered the operation stage.

More importantly, this is only the part that can be seen in the current public information. Many businesses in the computing power industry often deliver in batches, pay in stages, and release revenue in cycles. In other words, the published data is only a facet. There is still plenty of room for imagination about the actual scale of the business, the pace of revenue release, and profit elasticity in the future.

This is also where Guangdong-Hong Kong Bay Intelligence (01396) is quite recognizable.

Many computing power companies talk about “how much they will build in the future,” but what the market now wants to look at is “how much they have paid, how much they have run, and how much they have collected now.” Computing power is not a number on PPT; actual implementation requires a complete set of complex processes such as equipment, computer room, network, scheduling, operation and maintenance, and customer adaptation. If any link gets stuck, it's hard for the plan to become a billable asset.

Therefore, the core ability of a computing power operator is actually not to tell a big story, but to run the chain from this perspective. The Guangdong-Hong Kong Bay Intelligent Computing provides not a simple “computing power concept,” but a sample of fulfillment from order to delivery, from delivery to operation, and from operation to revenue.

Currently, the market is actually re-screening AI companies.

If there is no order, it will be questioned; if delivery is not clear, it will be discounted; if revenue is delayed, it will be re-valued. In contrast, companies that can acquire long-term customers, continue to deliver projects, and enter a revenue release period will be re-identified in sector repair.

Therefore, this round of recovery in the computing power sector is not just a rebound in sentiment, but more like a reconfirmation of the prosperity of AI infrastructure in the capital market. Early sector adjustments have already released many concerns, and the market's confidence in the supply and demand of computing power and the return on AI investment is being repaired. However, after the restoration, capital will no longer be bought as indiscriminately as in the first round, but will place more emphasis on ability to cash out.

The computing power market has not stopped; it has only entered a more critical stage.

Next, the market will not only reward storytellers, but will prefer those who can deliver. Orders continue to come, projects continue to be delivered, and operations continue to expand. This is currently the most valuable main line in the computing power sector.

In the AI era, computing power is not a one-time business, but a new infrastructure. Whoever can build, manage, use, and rent infrastructure and generate stable returns is more likely to overcome fluctuations.

This is also the underlying logic of the current hashrate sector becoming popular again.

This article is reprinted from the “Liu Xiaobo Talks About Finance” official account, Zhitong Finance Editor: Liu Jiayin.