What assumptions about AI computing, satellite networks and future cash flows are embedded in SpaceX’s public valuation?

Advocate A’s weakest move is treating the existence of a real launch-and-Starlink operating business as a “valuation floor.” That is directionally true only at much lower prices. At a $1.77 trillion IPO value, the disclosed current business does not function as a floor; it functions as a small down payment on a much larger set of future assumptions. Using A’s own numbers, SpaceX is being valued at roughly 95 times 2025 revenue and 269 times adjusted EBITDA. Those are not ordinary infrastructure-company multiples, and they cannot be justified by saying Starlink is already scaled. A cash-generating network can still be overcapitalized if the price requires it to become a dominant global telecom, defense, cloud, and orbital-compute platform simultaneously.

The first counter-evidence is that consolidated profitability is weaker than A’s framing implies. The prospectus says 2025 revenue was $18.674 billion, but also reports a $2.589 billion operating loss. Connectivity was profitable, but the Space segment produced only $653 million of segment adjusted EBITDA while funding $3.004 billion of Starship R&D, and in Q1 2026 that Space segment was adjusted-EBITDA negative. So investors are not simply buying a mature launch monopoly plus broadband annuity; they are funding a capital-intensive development machine whose value depends on future cost curves and launch cadence.

Second, Starlink’s economics are promising but not yet proven as telecom-like free cash flow. Connectivity generated $7.168 billion of segment adjusted EBITDA in 2025, but the same filing attributes pressure to lower-priced international expansion, higher depreciation from more Starlink flights, ground operating costs, and international expansion costs. Q1 2026 subscriber growth of 104.7% was offset by a 22.9% ARPU decline. That is not fatal, but it means the key question is replacement capex and lifetime economics per satellite, not segment EBITDA before all capital intensity is settled.

Third, the AI/orbital-compute option is not marginal to the IPO story. The prospectus says SpaceX acquired xAI in February 2026, expects substantial capital allocation to compute infrastructure, and reported AI segment revenue of $3.201 billion against a $6.355 billion operating loss in 2025; Q1 2026 AI capex alone was $7.723 billion. It also says orbital AI compute satellites may begin deploying “as early as 2028,” which is inherently forward-looking. Independent technical evidence supports caution: the GAO says space data centers require further engineering development, including solar arrays larger than any launched and assembled in space as of April 2026, while the World Economic Forum highlights severe radiative cooling limits.

Finally, market-structure evidence supports the liquidity-event thesis. Axios reported the $75 billion raise and $1.77 trillion value, and its newsletter cited Bloomberg that everyday investors submitted more than $100 billion in purchase orders; Axios also reported that up to a quarter of shares were earmarked for smaller retail traders. That demand is not proof of fundamental value. It is evidence that scarcity, retail access to a Musk-controlled asset, and AI-era liquidity are embedded in the price.

A’s rebuttal concedes the central issue and then tries to relabel it. A now agrees that the $1.77 trillion valuation is not supportable by conventional capitalization of current revenue or EBITDA, that consolidated GAAP profitability is weak, and that Starlink segment EBITDA is not the same as steady-state free cash flow after replacement satellites, ground infrastructure, spectrum compliance, and expansion capex. That concession matters because the debate is not whether SpaceX has a real business; B has never denied that. The question is what assumptions are embedded in the public valuation. Once A admits current cash flows do not justify the price, the remaining value must come from aggressive future monetization, option value, scarcity/liquidity demand, or some combination of those. That is B’s thesis.

The specific weakness in A’s logic is the claim that because orbital AI is technically unproven and currently loss-making, it must be “secondary upside” rather than a major driver of the IPO price. Negative present EBITDA does not show investors are not paying for AI; in high-growth markets it often shows the opposite, because investors accept near-term losses only if they believe future scale economics will be large. The prospectus says the AI segment had a 2025 operating loss of $6.355 billion, 2025 segment adjusted EBITDA of negative $1.237 billion, and Q1 2026 AI capex of $7.723 billion. It also says SpaceX expects to launch orbital data centers and faces a multi-year horizon before the AI segment reaches sustained positive adjusted EBITDA. That is not a small side project; it is a capital allocation priority large enough to dominate near-term investment needs.

A’s vertical-integration point also does not solve the valuation problem. Internal launches can reduce deployment costs, but they also demonstrate that the launch segment is partly an internal capex engine rather than a fully monetized third-party revenue business. The prospectus says launch capacity is significantly allocated to Connectivity and expected to be significantly allocated to AI. That means reported launch cadence cannot be valued like pure external launch revenue; much of its value depends on whether Starlink and future orbital AI earn attractive returns on the assets deployed.

Starlink remains the best fundamental case, but A’s own evidence leaves the free-cash-flow bridge incomplete. The prospectus says Starlink subscribers grew rapidly, but ARPU is expected to keep declining as lower-priced international plans expand. Connectivity capex rose to $4.178 billion in 2025 and $1.332 billion in Q1 2026, driven by satellite and ground equipment costs. Therefore the correct test is not segment EBITDA but post-replacement, post-expansion economics over satellite lifetimes.

Finally, liquidity evidence remains material. Axios reported more than $100 billion of retail purchase orders and a record $75 billion offering, while broader commentary warned that SpaceX, OpenAI, and Anthropic listings could absorb more than $240 billion of risk-on capital. A says this is not decisive; fair. But combined with valuation multiples A concedes are extraordinary, it strongly supports B’s view that scarcity, Musk premium, AI-era demand, and future-option assumptions are embedded in the price.

A’s strongest position is that SpaceX is not an empty AI story: it has a disclosed, scaled operating base in reusable launch and Starlink, with 2025 revenue of $18.674 billion, adjusted EBITDA of $6.584 billion, operating cash flow of $6.785 billion, and Connectivity segment adjusted EBITDA of $7.168 billion. A is also right that unresolved orbital-compute engineering constraints make AI hard to treat as the near-term “valuation floor,” while launch cadence and internal deployment give SpaceX a vertical-integration advantage that ordinary telecom or launch peers lack. (justtrade.com)

B’s position is still stronger because the debate asks what assumptions are embedded in the public valuation, not whether SpaceX has real assets. The valuation reported by Axios was about $1.77 trillion after a $75 billion raise, which means current disclosed revenue and adjusted EBITDA imply roughly 95 times sales and 269 times adjusted EBITDA. (axios.com) Those multiples are too high to be explained as a normal launch-plus-broadband infrastructure valuation. Therefore, a large share of the price must depend on future assumptions: Starlink sustaining high returns after replacement capex, defense and enterprise demand expanding, orbital AI becoming technically feasible, or public-market scarcity supporting a premium for a Musk-controlled asset.

A’s best evidence, Starlink EBITDA, is useful but incomplete. Segment adjusted EBITDA does not settle whether Starlink can produce durable telecom-like free cash flow after satellites are replaced, ground equipment is expanded, depreciation rises, spectrum obligations are met, and lower-priced international customers dilute ARPU. If the valuation requires Starlink to become a global broadband utility with unusually high margins despite ongoing satellite replenishment, then investors are already underwriting assumptions beyond the currently proven cash-flow base.

The AI evidence is also more material than A allows. The prospectus data cited in this debate show large AI operating losses and very large AI capex, while outside technical assessment from GAO says space data centers still face unresolved power, thermal-management, communications, radiation, collision, and orbital-interference issues. (justtrade.com) A loss-making segment can still drive valuation if investors believe the eventual addressable market is enormous; that is common in AI-era growth pricing. SpaceX’s own proceeds narrative, reported by the Los Angeles Times/Bloomberg and NPR Illinois, tied the IPO to AI compute infrastructure as well as launch and satellite expansion. (latimes.com)

Finally, liquidity evidence reinforces B’s interpretation. Axios reported more than $100 billion of everyday-investor purchase orders, and CoinDesk described concerns that the SpaceX, OpenAI, and Anthropic listing wave could absorb risk-on capital and create index-flow effects. (axios.com) That does not prove the entire valuation is irrational, but it does show that scarcity and AI-era capital-market demand are part of the price formation. I would be convinced by A only if audited filings and allocation records showed most valuation support came from conservative Starlink-plus-launch free-cash-flow models rather than orbital-AI assumptions or price-insensitive liquidity demand.