SpaceX — Amazing Company, Terrible Investment
Written at IPO, June 2026. Numbers from the S-1 and IPO prospectus.
SpaceX is a genuinely remarkable company. Reusable orbital rockets, Starship, the whole vision of multiplanetary civilization; it’s the kind of stuff that felt like science fiction twenty years ago. Extending the light of consciousness beyond Earth is a goal worth taking seriously.
Space has historically been a government domain precisely because there’s no revenue in the observable future. Running something like this as a private company, and pulling it off, is impressive.
Which makes it all the stranger that it just IPO’d at the seventh largest market cap in the world, up 20% on day one. So let’s talk about what you’re actually buying.
The Business Reality
The revenue picture from the S-1 is actually clear for the first time: $18.7B total 2025 revenue, net loss of $4.9B.1
| Segment | 2025 Revenue | Notes |
|---|---|---|
| Starlink (connectivity) | $11.4B | 63% EBITDA margin, 8.9M subscribers |
| Space (launches) | $4.1B | Only 8% YoY growth; mostly Pentagon/NASA |
| xAI / X | $3.2B | Lost $6.4B on operations |
Rocket launches are not the business. Starlink is. And Starlink is genuinely impressive: 50% revenue growth YoY, 63% EBITDA margins, no meaningful competitor.2
The ceiling is real but unevenly distributed. On land, Starlink only makes sense where terrestrial ISPs can’t reach cost-effectively; that’s a niche. At sea and in the air it’s a different story: legacy satellite providers can’t compete on speed or latency, enterprise churn is essentially zero, and there’s no credible competitor except Amazon’s Kuiper.9 Government and military contracts (Starshield) are similarly sticky. SpaceX claims $1.6T in Connectivity TAM in the S-110; the maritime and government pieces are real, consumer broadband on land is a niche, and the $740B mobile TAM is speculation until there’s a business model.
The S-1 frames a large chunk of SpaceX’s valuation not on satellites but on AI. xAI is already operating as an LLM provider, just not a particularly successful one. How is that going?
Colossus 1: The Compute Play
xAI’s Colossus 1 cluster exists and is profitable to lease; Anthropic is now renting the whole thing. But it was architecturally compromised from the start: a rushed mix of ~150K H100s, 50K H200s, and 20K GB200s resulted in a straggler effect during training, with reported Model FLOPs Utilization of ~11% versus the 35-50%+ typical of well-run frontier clusters.3 xAI has since moved all serious training to Colossus 2 (Blackwell-only) and is monetizing Colossus 1 as an inference fleet. Leasing it out is sensible; building an AI crown jewel on top of it, less so.
There are also plans for GPU clusters in space (GPUs in Space are dumb, right?). The TLDR: actually a sensible long-term idea. But Elon’s timeline is wildly optimistic. It won’t happen as planned.
Grok Has No Edge
The current frontier looks roughly like this: OpenAI is the most comfortable to use, the most human-feeling model. Anthropic has the best coding abilities. Google has DeepMind’s research depth and can leverage existing users across Search, Gmail, Drive, and Android to make access frictionless. Each of them has a reason to exist.
Grok’s one genuine differentiator is Twitter data access, which is useful for real-time research on very novel or breaking topics. That’s a niche. And it’s a niche that doesn’t justify the valuation of the company training it.
On benchmarks, Grok 4 looks competitive. Take that with skepticism. Benchmark contamination is trivially easy: train on a bit of the test dataset until your boss is happy with the numbers. Gemini got caught doing this before. Others probably do it too. Real-world usage patterns tell a more honest story, and there Grok trails badly on DAUs.
xAI burned $6.4B on AI operations in 2025 while generating $3.2B in revenue.1 That’s not building a moat; that’s paying to be in the conversation.
The IPO Is a Liquidity Event, Not an Investment Opportunity
SpaceX’s current valuation is fiction. The IPO mechanics are worth understanding carefully:
- Float of ~4%: Nasdaq rewrote its own rules to allow SpaceX into the Nasdaq 100 after just 15 trading days, scrapping the prior 10% minimum float requirement. Index funds will be forced to buy at whatever price prevails.4
- Retail allocation of 30%: Three times the standard for a mega-cap IPO.5 This is retail getting handed the bag at peak enthusiasm.
- Unorthodox lock-up: Instead of a standard 180-day cliff, insiders can start selling after the first earnings report; staggered tranches begin as early as late July.6
- Accounting complexity: The $18.7B revenue figure includes xAI and X via common-control consolidation, retroactively merging entities that only formally merged in February 2026.7
- No real governance: Musk holds 42% of equity and 85% of votes. Public shareholders have essentially no influence over how this company is run.
This is a textbook liquidity event: existing shareholders cash out, SpaceX raises fresh capital. Markets are at historically high P/E multiples right now. Smart time to IPO. Worst time to invest. One analyst firm puts fair value at $780B; the IPO valued it at $1.77T.8
The Tesla PsyOp, Reprised
I love this company and their mission. Getting humanity to Mars and beyond, extending the light of consciousness; that matters. But it is not a sensible business proposition at current prices.
The dynamic here mirrors what happened with Tesla (The Tesla PsyOp). Tesla is exceptionally good at PR and hype. They drove their stock price up on a future that wasn’t ready yet, which caused a massive overallocation of capital during early R&D. The long-term vision turned out to be correct. And that early overallocation was genuinely great for Tesla, and arguably great for accelerating progress on the EV tech tree in general. It just wasn’t great for investors who bought in at inflated prices.
SpaceX is running the same play. The long-term vision is probably correct. The overallocation of capital will probably accelerate humanity’s path to Mars. The paradoxical result is that we get there faster than market-efficient conditions would allow. There’s something almost beautiful about that.
But let’s not put our own money there. In a market downturn, the narrative will deflate. That’s the buying opportunity.