SpaceX IPO: Correcting the Record — The "Conglomerate Discount" Explained
Yesterday, I posted a deep dive into the SpaceX S-1. It was well-received, but it also triggered a crucial piece of feedback from a sharp reader who knows their way around SEC filings.
He was right to push back.
In an IPO this complex, "close enough" financial summaries can be dangerous. The previous post was a strong "red flag memo," but it glossed over some math and leaned too hard on dramatic narrative. A public offering deserves precision.
Here is the correction. Here is where the S-1 actually says the numbers, and why the "Buy Starlink, not SpaceX" thesis is even harder to execute than it first appeared.
1. The "80% Starlink" Myth
Correction: Starlink is ~61% of revenue, not 80%.
The previous post claimed Starlink accounted for "80% of SpaceX's entire revenue." This was mathematically incorrect.
According to the S-1:
- Connectivity Revenue: $11.39 billion
- Total Revenue: $18.67 billion
- Share: ~61%
The 80% figure could only be argued as a "value hypothesis" (i.e., Starlink should be worth 80% of the enterprise value). But as a revenue share, it was a mistake.
Why this matters: If Starlink is 61% of revenue and valued at a healthy 5x multiple ($57B value), it accounts for roughly 33% of the $1.75T IPO valuation. The remaining ~67% of the valuation rests entirely on the Space segment, the AI/xAI segment, and the Mars narrative. That is a much riskier ratio than the 80% narrative implies.
2. The "Profitable" Space Segment
Correction: Space was actually operating at a loss.
I described the Space segment (launch services) as "profitable" because the Segment Adjusted EBITDA was positive at ~$653 million.
However, looking deeper at the S-1 financials reveals a more nuanced picture. The Operating Income for the Space segment in 2025 was approximately -$657 million.
Why this matters: Rocket manufacturing and launch operations are brutally capital-intensive. Relying on EBITDA here is misleading because it paper-over massive depreciation and amortization. The Space business is not a cash cow; it is a loss leader that keeps the Starlink constellation in orbit. If you strip out the Space segment, the core business is less profitable than it appeared.
3. The $45B Anthropic "Promise"
Correction: It is a 180-day lease, not a bank deposit.
The previous post highlighted the $1.25 billion/month compute deal with Anthropic as a $45 billion revenue pillar.
The Reality:
- Termination Risk: The contract allows either party to terminate with 90 days' notice.
- Musk's own words: Reuters reported Musk stating the effective commitment was closer to a 180-day lease, not a guaranteed multi-year stream.
Valuation Impact: If we assume a maximum of 6 months of execution at $1.25B/month, the contract is worth ~$7.5B. Discounting this for termination risk (let's say 50%), the expected value of this "revenue stream" to the IPO valuation is closer to $3–4 billion, not $45 billion.
It is a positive signal for AI infrastructure demand, but it is not the "savings account" the headline implied.
4. The $10B Cursor Fee
Correction: Read the legal boilerplate.
The previous post framed the Anysphere (Cursor) deal as: *"If SpaceX doesn't buy Cursor, they owe $10 billion."*
While the potential exposure is indeed that high (a $1.5B termination fee + $8.5B deferred services), legal reality is messier. In M&A, "termination fees" are heavily constrained by definitions of "Material Breach," "Timing," and "Delivered Services." A simple "change of mind" usually triggers a negotiated settlement, not a automatic $10B write-off.
Verdict: It is a very aggressive contract designed to lock SpaceX in, but treating it as a literal $10B debt on day one is a misreading of M&A mechanics.
5. Revised Valuation Logic: The "Conglomerate Discount"
This brings us to the real investment problem.
If Starlink is the "real" business—a $11.4B revenue asset with 61% margins—what is it worth?
- Conservative multiple (3x): $34B
- Aggressive multiple (5x): $57B
Even at 5x, that is roughly 3.3% of the $1.75T valuation.
The remaining 96.7% of the price is being paid for:
- xAI: A company with ~$3.2B revenue but a ~$6.4B operating loss.
- Launch Services: A business operating at a loss.
- The "Mars Narrative": The $28.5T TAM claim, which relies on AI replacing white-collar labor (a speculative macro theory).
- Governance Risk: The absolute control Musk holds over the entity.
The Verdict:
When private markets price a company, they can ignore "operating losses" because they have infinite time to wait for an exit. But public markets have daily P&L pressure.
By taking SpaceX public, Musk is effectively forcing retail investors to subsidize xAI's losses using Starlink's cash flow.
The "Conglomerate Discount" applies here because you are buying:
- A high-quality asset (Starlink)
- At a massive discount (because you are stuck with the AI and Governance baggage).
But the discount isn't deep enough. At $1.75T, the market is pricing in a perfectly executed AI revolution, not just a satellite company.
Final Thought for Investors
The feedback I received was spot on: *"SpaceX IPO is not a 'buy SpaceX'. It is a bundle."*
If you believe Starlink is the most valuable asset on the planet, you are right. But at this IPO price, you are not just buying Starlink. You are buying the entire ecosystem, including the governance red flags, the AI money-burn, and the Musk-controlled governance structure.
"Buy Starlink, not SpaceX" remains the right investment philosophy. The problem is, there is no such thing as a pure "Starlink" stock on the Nasdaq yet. And until there is, you're paying a premium for the privilege of owning the whole messy package.