One could think of the Chinese constellations as a little like “divisions” of a single enterprise with the Chinese government being the “chairman of the board of directors,” which meets every five years.

Chinese five-year plans outline strategic goals and initiatives to guide economic and social development. The first plan (1953-57) began the transition from an agricultural to an industrial economy, and the 15th plan (2026-30), which was outlined this month, aims to achievegreater self-reliance and strength in science and technology, including coordinating the construction ofsatellite communication, navigation, and remote sensing systems, and accelerating the developmentof low Earth orbit (LEO) satellite internet.

China is developing three LEO Internet service constellations, Guowang, Qianfan, and Honghu-3, with plans for eventually having 12,992, 15,000, and 10,000 satellites, respectively. How many of these satellites will be in service by the end of the newly announced five-year plan in 2030? To answer that question, I “interviewed” six AI chatbots, and the results are shown here.

The chatbots disagree, which is to be expected because the estimates are based on assumptions about the rates of manufacturing of satellites and rockets, the development of high-capacity reusable rockets, the availability of launch sites, progress so far, demand for service (Qianfan and Honghu-3 will be serving paying customers), available funding, etc.

China has competitors in the US and Europe, so I also asked the chatbots to estimate the number of satellites Starlink and the other Western competitors would have in orbit at the end of 2030. Starlink dominates this group, but by 2030, SpaceX will have been de-orbiting as well as launching satellites for several years, and the others will be expanding their constellations.

The chatbots estimate that in 2030, Starlink will still have the largest number of satellites in orbit, but the three Chinese operators combined will have 72% of the number of satellites Starlink has. Since they will serve different markets—government, enterprise, consumer, etc., one could think of the Chinese companies as a little like “divisions” of a single enterprise with the Chinese government being the “chairman of the board of directors” which meets every five years.

In this case, I just asked the chatbots the same, simple question: “How many LEO satellites will constellation x have in orbit by the end of 2030? I want your best estimate as a single number.” One could get more insight into the causes of the variance by asking an open-ended question, conversing with the chatbot, and reading and analyzing the transcripts of those conversations, but today’s chatbots do not archive conversations.

A final note on the estimation approach. The variation among chatbots might be interpreted as invalidating this pooled-estimation approach, but I wonder if the same sort of variance might not be observed if six knowledgeable human experts were asked the same question, since they too might consider different factors, make different assumptions and refer to different sources.

A final note on the estimates. Together they predict 31,876 LEO Internet service satellites by the end of the five year plan. During that time, there will also be many new launches of smaller LEO satellites that do not have the ability to maneuver, direct-to-device satellites with big antennas, and maybe huge data center satellites. If our estimate is even close to accurate, this sounds like a recipe for collisions and debris, so technology and policies for collision avoidance and debris removal better be in place before 2030. SpaceX has the most to lose from collisions, but there is strong pressure for all of these companies to cooperate to mitigate collision risk.

It will be interesting to see which chatbot prediction is closest to accurate at the end of 2030.