There was some looney stuff on the earnings call, even by Elon Musk standards.
For example, he claimed there would be 100 billion in recurring revenue by the end of 2026, and 1 trillion by 2030 maybe by 2029. Revenue is currently about $30 billion annual run rate. 3x by the end of the year is…a lot. 30X in 3.5 years is a CAGR of 140%. Good luck with that.
Musk said they are anticipating having “close to 10 GW” compute capacity next year, up from 2 GW by the end of this year. I asked Gemini how much it would cost to build 6 GW of compute (2+6 = close to 10). It spit out $210B – $360B. SpaceX spent $15 billion on AI capex last quarter, so $60 billion annual run rate. They would have to radically expand the capex. And of course, data center construction is already constrained by labor and power shortages, not to mention crowds of angry villagers. No way.
I’ll say it is pretty much out of the question. The main obstacle for space-based Internet and to a lesser extent phone service is bandwidth. Each Starlink satellite has limited radio spectrum and must share that capacity with everyone underneath it. As more users are added to a given area, each user’s share of capacity declines. Now imagine trying to supply Internet to NYC or Tokyo where you have a very high concentration of users. It would require a very high concentration of satellites. Fiber has enormous bandwidth and it is easy to add more. And unlike satellites which need to be refreshed regularly, once fiber is installed the maintence costs are very low.
Now, Starlink is great for rural/underdeveloped areas, maritime uses, aircraft, that sort of thing. But a majority of the world’s population can’t afford the hardware and subscription fees.
And finally, Musk touted the prowess of SpaceX’s terrestrial data center team, saying that SpaceX vs Google building terrestrial data centers is like “the New York Yankees going in and playing a little league team.”
Keep in mind, Google invented the way modern data centers work, and been building and operating them since the late 1990s. Virtually their entire business, Youtube, search, Gmail, etc. uses data centers. Google designs their own chips and builds their own networks. XAI only became part of SpaceX this year, and has only been in business for two years. It isn’t plausible that in some short period of time they cracked some code that puts them leaps and bounds ahead everyone to the extent Musk claims.
The Tesla earnings call was pretty sedate, but this one was borderline unhinged.