Smartest Thing I Have Seen This Week: Thompson of StraTECHery & Moffett of MoffettNathanson on SpaceX & Bubbles: PODCAST OF THE DAY
Craig Moffett on Musk’s possible positive-feedback flywheels, the physics of Starlink’s possible networks, SpaceX “optionality”, what 1999 has to teach us about the AI capital expenditure boom, and how you have to ignore the revenue sloshing between the layers of the bubble value chain and look only at what it produces that people outside are willing to actually pay for:
This is the best thing I have seen this week, and the best thing on SpaceX I have seen this year.
Ben Thompson and Craig Moffett talk.
Some highlights:
On SpaceX:
Ben Thomson & Craig Moffett: SpaceX & Bubbles <https://stratechery.com/2026/an-interview-with-craig-moffett-about-spacex-and-bubbles>: ‘Craig Moffett: ‘On the one hand, [Musk] says… absurd things…. But… they have achieved some extraordinary things…. Grok and the neocloud business… the AI business terrestrially… the biggest part of what SpaceX is….
The space segment is essentially a monopoly and… they’re also gaining knowledge at something like 10x the rate of anybody else, so… they’re pulling away…. The most valuable part of… Musk…. [is] you move so fast that you accumulate knowledge at a rate… completely different than anybody else…. This flywheel of generating volume, generating learning, driving down the cost curve way faster than anybody else…. Think about the difference in pacing. Between the time that [Blue Origin’s] rocket blew up and the time that they’re ready to launch the next one, SpaceX will have launched 14 times…
And:
This flywheel of generating volume, generating learning, driving down the cost curve way faster than anybody else…. That’s an extraordinary strategic position for them to have and it creates all kinds of optionality. You can identify all the businesses that they’re currently in…. You can also identify the businesses they say they will be in…. [But] it would have been intellectually wrong to say, “I’m just going to value this company solely on the things that I can see today”, because it really does have optionality…. The question is how large that optionality is…. [Our] neutral rating… [came because] we were a bit below consensus on almost all of the businesses in the here and now…. [In valuation] you’re not just predicting when it will arrive… [but] when do people think it will arrive…. Warren Buffett… brilliantly said… derisively, that most of investing is trying to guess what a bunch of other people will guess consensus will be in six months…
And:
I don’t know that [low-earth orbit] will stay a [SpaceX] monopoly…. The market is underestimating the level of regulatory and geopolitical risk…. Domestic political risk too, [as] he’s so closely identified with the Republicans…. The rest of the world feels like it is impossible to leave any Western company… controlled by Elon Musk in charge of critical technology…. China will absolutely have its own [satellite] constellation…. India, because Elon Musk has sort of declared war on the Ambanis…. The risk that there is [also] a European constellation… is much higher by virtue of the fact that the current U.S. administration has alienated so much of Europe….
The original concept for SpaceX was that SpaceX would have all these rockets going into space anyway for the U.S. government and the military…. [Thus] you could build a constellation like Starlink… sticking the satellites in the excess capacity of the payloads and so the marginal cost of putting the satellites into space would be zero. That’s a really gorgeous concept, except it never actually happened…. Every satellite in space today for Starlink was launched in a purpose-built launch….
We did our first analysis of Starlink broadband five years ago…. The real constraint is the number of customers you can simultaneously serve within a specific spot beam from one satellite…. There’s a limit to how tight you can make a spot beam when you’re sending it from 220 miles away. You’ve still got physics to work with…. If you lower the orbital, you have a tighter spot beam that increases your capacity. But the lower the orbit, the more drag from the ionosphere, and the faster the satellite will be pulled back down…. These satellites go up and five years from now, they’ve burned up in space and you need to replace them…. You’re going to have three or four launches of Starship every week just to maintain Starlink, much less all the other purposes…
And on bubbles:
The lesson I learned from 1999 is that you need to ignore all of the revenue being generated between layers of the value chain, it doesn’t matter…. I don’t really care what Nvidia is getting paid by some other company for its GPUs…. All I care about are the revenues coming into the system…. It can be enterprise, it can be government, and it doesn’t have to be today, it can be… projected…. Ignore all the revenues going on between layers of the value chain…. That’s just noise…. [What] really matters is [“is there] enough money coming into the system to support the system?[”]… it’s not that these aren’t amazing products…. It’s not that they aren’t going to be transformational…. It’s, “Is there going to be the price realization that allows them to be profitable enough that it supports the rest of the value chain?”… Right now… the market is saying, “We don’t know”….
My suspicion is that these are [many] businesses where… a cheaper model that’s two or three generations old [is fine]…. DeepSeek at the cost-per-token… that’s one-eighth or one-twelfth of frontier models…. What you can do with a thin client… able to run locally… [with] thinner and thinner and thinner models that are sufficiently good that they do most of what needs to get done for most use cases…. That puts a hell of a lot of burden on the capital investment of the really big stuff, for a relatively small part of the revenue opportunity. That’s the risk [here]…. We’re going in with our blindfold on and just wishing for the best…
Brad DeLong here: Think of it. As Craig Moffett says, between the time Blue Origin’s rocket blew up and its next launch, SpaceX will have launched its rocket fourteen times. Craig Moffett thinks that gap in learning speed is SpaceX’s real asset, which the market is by and large correctly pricing. But he also thinks the market is massively underpricing the political risk that comes with Elon Musk’s bizarre hatreds and lies.