China has a plan to build a data center in orbit. It has the filing. It has the money. What it does not have is a satellite that has flown.
Orbital Chenguang, a Beijing startup, announced on April 20 that it had secured $8.4 billion in credit lines from twelve Chinese state banks — a show of state financial backing for a project that envisions nearly 200,000 satellites in Sun-synchronous orbit as a space-based computing platform, according to SpaceNews. The company filed with the International Telecommunication Union, the UN body that coordinates orbital slots, for two constellations totaling 193,428 satellites under the names Chenguang-Tiancheng-1 and Chenguang-Tiancheng-2, SpaceNews reported separately. The filing reserves the slots. The question is whether Orbital Chenguang can fill them.
Chenguang-1, the experimental test satellite meant to prove the architecture works, was slated for launch in late 2025 or early 2026. It has not launched yet, according to SpaceNews. At China's 2025 launch rate of 93 orbital missions — a record year — loading 193,428 satellites takes more than 2,000 years. Push to 200 launches per year and pack 60 satellites per vehicle, which is aggressive even for a heavy-lift rocket, and the theoretical floor drops to sixteen years of non-stop launches with zero failures, zero weather cancellations, and zero gaps.
The $8.4 billion in credit lines is real. These are committed financing facilities from twelve state lenders including Bank of China, Agricultural Bank of China, and CITIC Bank — not capital already deployed. The milestones that unlock the full amount are undisclosed. State credit is patient. It does not manufacture launch pads.
The ITU filing is also real. ITU filings reserve orbital slots and coordinate spectrum, but they do not commit a country to a deployment timeline. Starlink holds FCC approval for nearly 12,000 satellites and has roughly 8,000 operational after seven years of launches. The gap between paperwork and working constellation is where most space-compute ambitions die.
Smaller Chinese players are further along in hardware. ADA Space and Zhejiang Lab launched twelve satellites in May 2025 for the edge-computing Three-Body constellation, and Shanghai Bailing Aerospace is developing a 100-kilowatt-class computing satellite. These are orders of magnitude below what the ITU filings contemplate, but they exist. Orbital Chenguang's filing is an ambition. Chenguang-1 is the test.
SpaceX has filed with the U.S. Federal Communications Commission for permission to launch up to one million solar-powered satellites as orbital AI data centers, Reuters reported. Elon Musk said at the World Economic Forum in Davos that solar power generation in orbit produces roughly five times more electricity than equivalent panels on the ground, making space potentially the cheapest place to run AI compute. Microsoft abandoned its own undersea data center project — an analogous extreme-environment computing effort — according to Reuters — suggesting the economics of operating compute outside normal environmental envelopes remain unsolved at scale.
Until Chenguang-1 flies, every number in this announcement is a projection backed by a press release. The credit lines are real. The launchpad is not.