Over the past two days, there has been a noteworthy shift in the technology sector, with the strongest gains seen not just in traditional chip stocks, but along the entire cloud services, data center, and AI infrastructure line. The market is sending a clear signal that the AI computing power theme is far from over, but has simply entered a new phase. In the past, the market primarily bought into computing power based on "capital expenditure" and "GPU shortages." However, following the recent volatility, capital has clearly adopted a new set of criteria. It is no longer only about the scale of computing power, but rather a deeper concern for whether that computing power can ultimately translate into customer orders, deliverable projects, and sustainable revenue. In other words, the AI rally is moving from a "pick-and-shovel seller" narrative to a phase focused on "who can consistently collect rent."

This change is critical. In the earlier stage, the market was willing to give valuations based on "how much computing power was planned," "how many resources were locked in," and "being positioned in the AI trend." But after a period of volatility, capital has become more demanding. It no longer just listens to growth potential or looks at concepts; it starts asking more practical questions: Are the orders real? Can the projects be delivered? Can the computing power be operated? Can the revenue be recognized? Can profits and cash flow keep up? This is what the market values most right now. Previously, when discussing AI, the market was willing to hear about potential; now, the market prefers to see delivery. Order size, customer quality, delivery pace, contract duration, revenue recognition, and cash flow conversion are becoming the core metrics for pricing this sector. A company with only a concept but no delivery will struggle to make progress in a recovery rally; but a company with orders, projects, and revenue will see its valuation recover much faster once market sentiment turns positive.

Following this logic, when looking at Hong Kong-listed computing power operators, the value of some companies becomes clearer. One company worth examining separately offers a perspective not just on "how much computing power is planned," but on how much verifiable business foundation has already been established. According to public information, GBA AI COMP has on-hand orders exceeding 30 billion yuan, has delivered over 15 billion yuan in cumulative orders, and operates more than 50,000 P (FP16 dense) of computing power. Over 95% of newly disclosed orders are long-term contracts with a five-year duration. When these numbers are viewed together, their meaning is different. On-hand orders indicate that demand is not imaginary but has been solidified through contracts, projects, and customer relationships. Delivered orders show the company is not just capable of taking orders but also has the ability to execute projects. Operating over 50,000 P (FP16 dense) of computing power demonstrates that this is not an asset still under construction, but infrastructure already in the operational phase.

More importantly, this is only the part visible in current public information. Business in the computing power industry is often delivered in batches, billed in stages, and revenue is released in cycles. In other words, the published data is just a snapshot, and the actual scale of future business, revenue release pace, and earnings flexibility still hold significant room for imagination. This is a key differentiator for GBA AI COMP (01396). Many computing power companies talk about "how much they will build in the future," but the market now wants to see "how much has already been delivered, run, and collected." Computing power is not a number on a PPT; true implementation involves a complex set of steps including equipment, data centers, networks, scheduling, operations, maintenance, and customer adaptation. If any single link gets stuck, plans can hardly turn into billable assets. Therefore, the core competence of a computing power operator is not just telling a big story, but successfully running the entire value chain.

From this perspective, GBA AI COMP offers not just a simple "computing power concept," but a delivery sample that goes from orders to delivery, from delivery to operations, and from operations to revenue. The market is currently re-screening AI companies. Those without orders will be questioned; those with unclear deliveries will be discounted; and those whose revenue is slow to materialize will be revalued. Conversely, companies that can secure long-term customers, consistently deliver projects, and enter a revenue release phase will be re-identified during the sector's recovery. Therefore, this rebound in the computing power sector is not just an emotional rebound, but more like a re-confirmation from the capital market of the prosperity of AI infrastructure. The previous sector adjustment has already released many concerns, and market confidence in computing power supply-demand dynamics and AI investment returns is being restored. However, after the restoration, capital will not buy indiscriminately as it did in the first round, but will place greater emphasis on execution ability.

The computing power market hasn't stalled; it has just entered a more demanding phase. Going forward, the market will not only reward those who tell stories, but will favor those who can deliver. Orders continuing to come in, projects continuing to be delivered, and operations continuing to scale up are the most valuable main themes of the current computing power sector. In the AI era, computing power is not a one-time business, but a new type of infrastructure. Whoever can build, manage, operate, and lease out this infrastructure, generating stable returns, is more likely to navigate through volatility. This is the underlying logic for the revival of the computing power sector.