Recently, the stock prices of aerospace transportation companies have been soaring fiercely, which actually reflects an interesting phenomenon—the capital market is rethinking the cost curve of commercial spaceflight. As launch frequency increases and unit launch costs continue to decrease, those low Earth orbit applications and deep space exploration projects that once seemed distant are rapidly transitioning from theoretical feasibility to actual engineering deployment.

Let's first look at the core logic: if we talk about the hottest topics in hard technology this year, space computing definitely ranks among them. But why must computing power go to space?

**Energy Constraints Have Become the New Ceiling**

In recent years, AI chip capacity has been expanding continuously, but the real bottleneck is no longer the chips themselves. The power demand for building massive data centers is skyrocketing; the power requirements for training clusters have already approached the gigawatt level—this exceeds the capacity of current power grids and generation systems. In key data center regions in the U.S., power access scheduling can take years, severely delaying the pace of computing infrastructure development.

Besides electricity, water and cooling are also major issues. Cooling clusters with hundreds of thousands of GPUs is beyond imagination, and the water cost per unit of computing power is rising sharply. Electricity, water, and land—these three factors combined have become the hard constraints on expanding computing capacity.

**Space as a New Outlet for Energy Issues**

This is why some are beginning to explore sending computing power into space. In special orbits like dawn-dusk orbits, near-continuous solar energy collection is possible, and there are no restrictions from ground land use, environmental approvals, or grid access. The energy supply is stable and sustainable. Even better, in the space environment, heat dissipation can be achieved via radiation, greatly reducing water resource dependence. Simply put, space computing opens a new path for high-energy-consuming computing to break free from ground-based energy limitations.

**How Will Space Computing Be Used?**

Here, a misconception needs to be clarified—space computing is not about competing with or replacing the entire ground-based system, but rather embedding it as a functional node within the overall computing architecture to create more complex and precise systems.

The most direct application is in-orbit data processing, known in the industry as "sky-to-sky computing." With the explosion in the number of remote sensing, communication, and navigation satellites, transmitting all raw data back to Earth would impose enormous bandwidth and latency pressures. Preprocessing, compression, or initial analysis done in space data centers can significantly improve system efficiency.

There is also a category of computational tasks that are energy-intensive but not sensitive to latency—such as model distillation, knowledge base construction, and long-term simulations. Space computing has natural advantages here because these tasks prioritize continuous operation and energy cost per unit of computing power.

From a broader perspective, space computing may also take on strategic roles like "autonomous computing and knowledge backup." Deploying key models and data on nodes that are less dependent on ground energy and network systems can enhance the resilience and autonomy of the entire system.

**Global Validation Has Begun**

The concept stage is over; space computing has now entered the phase of real, operational systems. A foreign company, after securing investment from a major chip manufacturer, has successfully launched high-end GPU-equipped computing nodes into orbit and begun operation, marking the first time space computing has entered the engineering phase with real workloads.

Meanwhile, a rocket company has announced a plan that clearly defines industry roles—rocket companies handle launch and transit, an electric vehicle company provides solar and energy storage solutions, and an AI firm supplies models and algorithms. Domestically, from multiple innovation alliances to city-level space computing competitions, all indicate that this track is entering a new stage focused on engineering validation.

**Where to Focus Investment**

To follow this trend, attention should be on companies actively engaged in commercial spaceflight and space computing. For example, a publicly listed company holds controlling stakes in satellite data center-related firms and is integrated into core management systems. This is the only A-share target explicitly linked to space computing operations. The company, which focuses on dawn-dusk orbit space data centers, has core technologies covering energy modules, cooling, radiation hardening, and data transmission between space and ground, forming a differentiated competitive edge.

A key milestone is the launch of the experimental star project scheduled for late 2025 to early 2026, aimed at validating key technologies for in-orbit computing and thermal management, laying a foundation for subsequent constellation networking. This is a technical milestone worth continuous attention.