SpaceX Targets $2T IPO at 278x Earnings on Starlink Alone

The Deal

SpaceX is heading to public markets in approximately two months at a proposed $2 trillion valuation. Two detailed reports from The Information in four days have provided unprecedented financial transparency: Starlink generated $7.2 billion in EBITDA in 2025 — and it's the only profitable segment. The rocket launch business doesn't produce cash. xAI is the biggest money-loser. Orbital data centers are pre-revenue concepts.

At 278x Starlink's EBITDA, this isn't a valuation — it's a conviction premium on Elon Musk's ability to execute across four capital-intensive frontiers simultaneously. The analyst framing is blunt: investors face a "very real chance they will end up losing their money."

What This Tests

This IPO is a referendum on whether public markets have the same risk tolerance as late-stage venture. If it succeeds at $2T, expect every cash-burning space-tech and AI-infrastructure company to rush the IPO window within 6 months. If it stumbles, the repricing will cascade through late-stage private valuations across both sectors.

A $2T IPO for a company where 3 of 4 segments burn cash tests whether public markets will price optionality at venture-fund levels — the answer sets the ceiling for every tech IPO behind it.

The Starlink Standalone Case

Starlink at $7.2B EBITDA has global monopoly characteristics in satellite broadband. Even at generous SaaS multiples (15-20x EBITDA), a standalone Starlink is worth $108B-$144B. That leaves roughly $1.85-$1.89 trillion of value assigned to cash-burning rocket launches, xAI (competing with OpenAI, Anthropic, and Google with inferior positioning), and a concept for orbital data centers. The gap between Starlink's defensible value and the headline number is where all the risk lives.

Portfolio Implications

This IPO has second-order effects across your portfolio regardless of whether you participate. It sets a valuation ceiling for space infrastructure, resets late-stage private AI lab comps (xAI's implied valuation within SpaceX), and tests LP appetite for narrative-priced mega-deals. Cross-reference with the AI revenue credibility concerns in this briefing: if markets accept $2T for one profitable segment, they'll accept anything — and if they don't, the repricing touches every inflated growth name.

What's New

The OpenAI-Anthropic competitive narrative shifted from strategy to financial credibility this week. OpenAI's CRO Denise Dresser alleged in a leaked internal memo that Anthropic inflates ARR by $8 billion through gross revenue accounting that includes cloud partner pass-through to AWS, Microsoft, and Google. This was independently confirmed: the $8B represents what OpenAI would add to its own run rate under gross reporting. The apples-to-apples comparison:

On a normalized basis, OpenAI likely leads by ~$3B regardless of which accounting method you use. If you're allocating based on headline ARR, you're overweighting Anthropic and underweighting OpenAI. Both methods are GAAP-compliant — but the ~27% implied cloud partner take-rate on Anthropic's revenue suggests meaningful margin pressure that won't show up until S-1 filings force standardized disclosure.

Three Additional Signals Reshaping the Competitive Map

Stargate execution risk: Three senior OpenAI executives behind Stargate defected to Meta this week — signaling that OpenAI's flagship infrastructure initiative faces management continuity risk at the exact moment it needs execution certainty. Meta, with deep pockets and zero compute trade-offs, is the structural beneficiary.

Pentagon supply-chain risk: The Pentagon formally labeled Anthropic a supply-chain risk for restricting military use of Claude. This creates a quantifiable displacement opportunity in defense AI procurement — the vendor field just thinned. Paradoxically, Anthropic is simultaneously building government relationships at the White House level over its Mythos model, and hired Ballard Partners as its lobbyist — building the regulatory moat that matters when formal frontier model licensing arrives.

Vercel IPO benchmark: Vercel signaled IPO readiness at $340M ARR with 30% of apps on its platform now agent-generated. At 20-30x revenue, expect a $6.8-$10.2B valuation — establishing the comp set for every developer platform and AI infrastructure company going public in the next two years.

The first S-1 filing from either OpenAI or Anthropic will force revenue accounting convergence — position on the normalized numbers before that catalyst, not the headline numbers after.