Starlink is making money. In the second quarter, it had RMB 4.29 billion in revenue and RMB 1.66 billion in profit, with 12 million users. ARPU fell from 85 to 66, showing it’s using price cuts to gain scale—from “premium broadband” to “global infrastructure.” This move is right, but it will squeeze profits in the short term. Don’t forget Starshield: in the second quarter it won over RMB 6 billion in U.S. government contracts—this is the real silent money-maker.
$SPCX
AI is burning cash. Revenue was RMB 2.56 billion, up 247% year over year, but it lost RMB 1.26 billion. Customers include Anthropic and Google, and it signed RMB 14.1 billion in non-cancelable contracts. Sounds impressive, right? But in the second quarter, capital expenditures were RMB 18.3 billion, 15.8 billion of which went into AI. The money Starlink makes just isn’t enough to fill the gap. Management says the new compute platform will pay back within a year—if it can, that’s a flywheel; if it can’t, it’s a bottomless pit.
Rockets are paving the way. Second-quarter revenue was RMB 960 million and it lost RMB 540 million. The rocket business itself isn’t profitable, but it determines how fast the other two legs can run. Starship’s first orbital attempt is in September, deploying 26 V3 satellites. Each V3 satellite’s bandwidth is 10 times that of V2, and the unit launch cost fell by 18 times. This is the real killer feature—Starship isn’t built to launch rockets; it’s built to launch Starlink and AI satellites.
So you see, the question isn’t whether this company will die. It’s whether it can get to profitability fast enough before cash runs out.
With a market cap close to two trillion, the market has already priced in high growth. Next, we just need two numbers: the payback period for AI capital expenditures, and the launch cadence of Starship.
If those two get proven, the story holds. If not, it’s another matter entirely.
Personally, I still lean toward the “it won’t work” side—but that doesn’t stop the world from having miracles, because expanding into the universe is still just too romantic.
$SPCXB
$SPCX
AI is burning cash. Revenue was RMB 2.56 billion, up 247% year over year, but it lost RMB 1.26 billion. Customers include Anthropic and Google, and it signed RMB 14.1 billion in non-cancelable contracts. Sounds impressive, right? But in the second quarter, capital expenditures were RMB 18.3 billion, 15.8 billion of which went into AI. The money Starlink makes just isn’t enough to fill the gap. Management says the new compute platform will pay back within a year—if it can, that’s a flywheel; if it can’t, it’s a bottomless pit.
Rockets are paving the way. Second-quarter revenue was RMB 960 million and it lost RMB 540 million. The rocket business itself isn’t profitable, but it determines how fast the other two legs can run. Starship’s first orbital attempt is in September, deploying 26 V3 satellites. Each V3 satellite’s bandwidth is 10 times that of V2, and the unit launch cost fell by 18 times. This is the real killer feature—Starship isn’t built to launch rockets; it’s built to launch Starlink and AI satellites.
So you see, the question isn’t whether this company will die. It’s whether it can get to profitability fast enough before cash runs out.
With a market cap close to two trillion, the market has already priced in high growth. Next, we just need two numbers: the payback period for AI capital expenditures, and the launch cadence of Starship.
If those two get proven, the story holds. If not, it’s another matter entirely.
Personally, I still lean toward the “it won’t work” side—but that doesn’t stop the world from having miracles, because expanding into the universe is still just too romantic.
$SPCXB
