Bezos vs. Musk: The Atlantic breaks down the four-rung ladder to vertical integration of the space economy

Bezos vs. Musk: The four-rung ladder to vertical integration in space

SpaceX handled 80% of global launches in 2025, with the Chinese government as the next closest competitor. Blue Origin, which actually predates SpaceX by two years, is nowhere near that number. When Josh Tyrangiel visited a Blue Origin launch in April, the company celebrated reusing a booster for the third time. He opened X on the drive home to find SpaceX had just reused a booster for the 600th time.

That gap defines the competitive picture — but Tyrangiel argues the race is about to change.

The four rungs

The first twenty-five years of the space economy were about lowering launch costs. Musk won that phase, then covered additional costs by building Starlink, turning low-margin launch into a boring but highly profitable telecommunications business. Blue Origin and Amazon are now running their own satellite constellation to compete.

The third rung is data centers in space. Compute costs on Earth, NIMBYism, energy expenses, and the constant threat of throttling make orbital data centers, per people Tyrangiel reported on, effectively inevitable — with a five-to-seven-year timeline. The fourth rung follows logically: AI model development, where unfettered compute access becomes the determining factor in who can build, fine-tune, and serve models at scale. Without launch capability, access to the third and fourth rungs is structurally compromised. That's why, roughly 18 months ago, Sam Altman reportedly explored buying Stoke Space — and why neither Google nor Microsoft can easily shortcut into this stack by partnering with a Rocket Lab.

“SpaceX is responsible for 80% of what was launched in the world in 2025... Blue Origin celebrated a booster reused for the third time. As I was driving away, I opened X and SpaceX very casually posted they had just reused a booster for the six hundredth time... Bezos has a project called Prometheus — an artificial general engineer trained on physical data, trying to model things that are really hard to model. The end game, if you believe we will get there, is a machine that makes machines.”

Grok vs. Prometheus

The two companies are also targeting different kinds of AI. Musk's investment runs through Grok, which Tyrangiel describes as largely LLM infrastructure. Bezos's bet is Project Prometheus, which he co-CEOs directly. Prometheus is a world model trained primarily on physical data, aimed at designing and producing physical things — what Bezos and others call an "artificial general engineer." The end goal, if the framing is taken at face value, is a machine that makes machines. For someone who owns Amazon, needs dexterous robots at scale, and has serious lunar colonization plans, that's a coherent through-line.

Blue Origin's structural problem

Blue Origin's early years were, by one former senior NASA official's description, an academic lab — beautiful facilities, little urgency. SpaceX, by contrast, had young engineers sleeping in the office and eating what they killed. Musk's resource constraints drove pace; Bezos's Amazon wealth meant Blue Origin could drift.

Talent poaching won't close that gap quickly. These are tribal companies, Tyrangiel argues, where people are knitted into their physical and collaborative environments in ways that don't translate cleanly to a defection offer. The money also isn't at AI-researcher levels. Blue Origin raised outside capital and now has a real valuation, but raiding SpaceX isn't a realistic catch-up strategy.

Bezos's conviction

Tyrangiel didn't speak directly to Bezos for the piece, but multiple sources told him Bezos genuinely regrets not leaving Amazon earlier to focus on Blue Origin. His high school girlfriend told reporters during Amazon's early days that his goal in getting rich was to go to space. The personality contrast with Musk, Tyrangiel suggests, owes partly to generational formation: Bezos came up through D.E. Shaw, where you keep secrets and deliver results. Musk is just young enough to have no inhibitions about what he says out loud. The underlying conviction, according to the people Tyrangiel spoke to, is roughly equal — and Bezos is now trying to make up for lost time.

If the space economy closes into a true flywheel, the companies that own launch will have a structural lock on the higher rungs. That's the moment, Tyrangiel notes, when antitrust scrutiny becomes hard to avoid.