From SpaceX's S-! it bases its valuation on the total addressable market of 3 markets, which is made up of:

1. Space operations: $370 billion;

2. Connectivitiy: $1.6 trillion, roughly split between mobile and broadband; and

3. AI: $26.5 trillion, which includes $22.4T in AI Enterprise applications, $2.4T in AI infrastructure, $760B in subscriptions (ie Grok) and $600B in digital advertising.

So immediately we see the limits of the space market. SpaceX did 170 launches in 2025. At $100M each, that's $17 billion in revenue and Falcon 9 launches just don't cost that much. That brings Starship into the picture but that program is arleady at $15 billion spent without a single dollar in revenue. So we know from the outset that only the other markets can really justify the STarship prgoram because you have to remember that STarship has to compete with Falcon 9. Or, even worse for SpaceX, a Falcon 9 competitor.

Now, as for Starlink, this one is interesting. Starlink has a big advantage when you need mobile broadband (eg planes, boats) as there's nothing really equivalent, Yes, there's 5G and that's fine in populated areas on, say, RVs and such. But the receivers are expensive.

I think Starlink is going to have a hard time competing with 5G, mainly because 5G already exists and Starlink handhelds are predicated on STarship being able to launch sufficient low-altitude V3 Starlink satellites, which themselves have a more limited life because they are lower altitude. And you're still going to have to deal with per-country licensing, marketing, regulations, etc.

I wonder how much of this is predicated on military applications.

But let's get to AI as that's the most hand-wavy (IMHO) of all this.

First, digital advertising. Well, when Elon bought Twitter it had $4.5B in ad revenuie. Now it's $1.8B. I also think Twitter is fundamentally limited so that's just not going anywhere. Twitter just isn't 10x'ing it's audience (IMHO).

Subscriptions? I think this is a dead market. For everyone. Why? There'll be a race to the bottom, hardware will get cheaper, eventually AI agents will be run locally (even on phones, eventually) and it's just not the cash show OpenAI, SpaceX or Anthropic think it is.

Let's also dispense with orbital AI data centers. It's a completely dumb idea. it's going nowhere. Don't believe anyone hyping it up. It makes zero sense and it will never amount to anything.

SpaceX seems unable to monetize their AI investments thus far. How do I know this? Because you don't lease DC space and GPUs to Google if you have a better use for them. So it not only looks like you're falling behind but you also need the cash. Also, this is dependent on getting enough GPUs. Published details about the Google deal seems to allow Google to cancel or revise the deal if SpaceX is unable to deliver so there is a huge risk here.

What none of these companies (SpaceX, Anthropic, OpenAI) seem to be addressing is what happens when future generations of hardware come out? What happens when running models locally becomes more and more viable? What if there is no expensive hardware moat? What happens when AI models get commoditized? I personally believe this last one will happen and China will make sure it happens.

So all in all there are a ton of risks here. I am skeptical about the long-term prospects of OpenAI in particular and Anthropic less so but I think xAI's (ie Grok's) future is way more uncertain.

I'm actually most confident in Google's future here because they haven't bet the company's future on AI not imploding.

> There'll be a race to the bottom, hardware will get cheaper, eventually AI agents will be run locally (even on phones, eventually) and it's just not the cash show OpenAI, SpaceX or Anthropic think it is.

That’s why we are seeing pushes for regulatory capture, safety fear mongering, large government contracts, and even potentially the government taking stakes in these companies. It’s a sprint to cash out before they get commoditized.

[dead]