Space Exploration Technologies Corp. (SPCX) — Coverage File

Data as of: Q2 2026 (quarter ended June 30, 2026; 10-Q filed Aug 4, 2026); latest source 8-K Aug 14, 2026 (Cursor merger closed) File last updated: October 10, 2026 (evidence check: no new filings since Aug 14; two risk-evidence additions and one correction)

---

0. 30-second reopen

• Business: Starlink is the profit engine (61% segment EBITDA margin), launch is its in-house delivery system, and the AI segment (Grok, X, Colossus data centres) now takes 86% of a ~$18B-a-quarter capex budget.
• Core debate: Is this a Starlink compounder whose cash and launch monopoly fund real options, or an AI-infrastructure spending vehicle priced on management's forward claims?
• Base thesis: Starship's V3 satellites step up Starlink capacity by an order of magnitude, while owned compute earns fast paybacks renting to frontier labs, so revenue scales faster than the capital it consumes.
• Main bear case: Free cash flow is -$32B on a trailing basis, compute revenue rests on a few customers who can leave on 90 days' notice, and outside holders have no governance voice.
• What to watch: Starship reaching operational payload delivery, AI infrastructure revenue against capex, cloud contract renewals, Starlink subscribers × ARPU, and dilution from Cursor, EchoStar and equity awards.
• Valuation: ~$2.2T market cap, ~93x trailing revenue, negative trailing free cash flow. The price capitalises management's $100B ARR and $1T revenue targets, not reported results.

---

1. Business snapshot

SpaceX reports three segments, recast in Q1 2026 after the February 2026 xAI merger. Because SpaceX, xAI and X were under Elon Musk's common control, every historical period is restated as if the three had always been combined, so FY2023–FY2024 include X and xAI (and a $3,775M Twitter-era impairment in FY2023) [424B4].

Connectivity (Starlink) is the core: $11,387M of FY2025 revenue, $4,423M of operating income and $7,168M of segment adjusted EBITDA [424B4, FY2025]. In Q2 2026 it did $4,291M of revenue (+66% y/y), $1,656M of operating income and $2,597M of segment adjusted EBITDA. Consumer revenue was $2,485M and Enterprise & Government (including Starlink Mobile) $1,806M, the latter up 108% [Q2 2026 10-Q]. Subscribers doubled to 12.0M while ARPU fell from $85 to $66 on international expansion and cheaper plans.

Space sells Falcon launches and government development work: $4,086M of FY2025 revenue, but a 657M operating loss because it carries Starship R&D (3,004M in FY2025) [424B4, FY2025]. Starlink launches earn no inter-segment revenue; their cost is capitalised into satellites. Only 10 of 37 Falcon launches in Q2 2026 were for customers [Q2 2026 10-Q].

AI combines Grok, X (advertising and subscriptions) and compute infrastructure. Q2 2026 revenue was $2,561M, of which $2,194M was AI Solutions & Infrastructure, including $1,600M from new cloud-services agreements renting Colossus capacity. Advertising fell 14% y/y to $367M during a platform rebuild. The segment lost $1,257M at the operating line but turned EBITDA-positive at $1,146M [Q2 2026 10-Q]. Cursor ($60B implied equity value, paid in ~391M Class A shares) closed on August 14, 2026 [8-K Aug 14, 2026].

2. Core debate

The first question is whether the AI compute build earns its cost. Management says new compute deployments pay back in under a year and that each new agreement carries better economics [Q2 2026 call]. The filings show the other side: capex of $18.4B in Q2 alone, $15.8B of it for AI. The largest compute contract (Anthropic, $1.25B a month through May 2029) can be terminated by either party on 90 days' notice after its first three months, and one AI customer was 19.5% of Q2 revenue [424B4; Q2 2026 10-Q].

The second question is whether Starship arrives on time to carry the Starlink and Mobile roadmaps. V3 broadband satellites (1 Tbps each, up to 60 per Starship launch), Mobile V2 satellites and orbital compute all depend on it [424B4]. As of Q2 2026, every Starship launch had been internal, and payload delivery to orbit was still expected to begin only in H2 2026 [Q2 2026 10-Q; 424B4].

3. Base thesis

SpaceX owns the cheapest, highest-cadence route to orbit, about 80%+ of global mass to orbit each year since 2023, and it has used that position to build the only global low-latency broadband network [424B4]. Starlink is already a scaled, high-margin service business. Starship multiplies its capacity rather than merely extending it, and the same vertical integration (launch, satellites, terminals, data-centre construction) lets SpaceX deploy AI compute faster and more cheaply than rivals. Compute is rented to frontier labs while it is scarce and used for Grok when it is not. On this view the IPO and bond cash ($100.0B at June 30) buys the build without stretching the balance sheet.

Assumptions this rests on:

1. Starship moves to operational, reusable payload delivery in 2026–27, so V3 Starlink deployment reaches the ~1,000 satellites management says gives critical mass by about Q2 2027 [Q2 2026 call].
2. Starlink revenue keeps compounding as subscriber growth outruns ARPU dilution, and Enterprise & Government keeps growing faster than consumer.
3. Compute stays supply-constrained long enough that cloud contracts are renewed or replaced on similar terms rather than terminated on 90 days' notice.
4. Connectivity keeps a segment EBITDA margin near 60% while V3 depreciation (satellite lives of 3–5 years) rolls in [424B4].
5. AI capex can be slowed or reallocated if demand softens, as the 10-Q says it can [Q2 2026 10-Q].

4. Main bear case

Before the xAI merger this was a broadband company with a launch subsidiary. After it, most of the capital goes to a business with a few large customers and short notice periods. Trailing twelve-month operating cash flow was $9.9B against capex of $42.2B, so free cash flow was about **-$32.3B**, and Q2 2026 alone was about -$16.0B (author's arithmetic from reported periods) [424B4; Q2 2026 10-Q]. Management guides the next two quarters to Q2-like capex [Q2 2026 call]. The AI segment's EBITDA turn came from $1.6B of revenue on newly ramping cloud agreements, while its D&A already runs at $1,885M a quarter.

How it breaks: compute supply catches up and prices normalise; a large customer gives 90 days' notice while short-lived servers keep depreciating; Starship slips, delaying V3 and Mobile, so Starlink growth shrinks to subscriber gains net of falling ARPU. Meanwhile ~$2.2T of market value rests on management targets ($100B ARR by December 2026, $1T revenue by 2030), and shareholders cannot change the board.

5. Business quality

Connectivity is a genuinely advantaged business. Launch cost and cadence are the moat: competitors must buy launches, often from SpaceX, while SpaceX builds satellites and terminals in house. Segment adjusted EBITDA margin was 63% in FY2025 and 61% in Q2 2026, and operating margin rose about three points y/y to 38.6% [424B4, FY2025; Q2 2026 10-Q]. Retention evidence is thin: no Starlink Enterprise customer above $750K of annual revenue has voluntarily left since 2023 [424B4]; consumer churn is not disclosed.

Space is strategically essential and financially a cost centre: customer launches and price per launch were flat in FY2025 even as total Falcon launches rose from 134 to 165 [424B4, FY2025]. AI quality is unproven. The segment is new, still being integrated, is still absorbing restructuring (X's 2022 workforce reduction was still producing charges in 2026) and depends on third-party GPUs, Nvidia exclusively from here [Q2 2026 call].

6. Key value drivers

• Starlink volume × price: subscribers (12.0M, +101% y/y) against ARPU ($66, -22% y/y). The CFO expects geographic mix to keep pushing blended ARPU down [Q2 2026 call].
• Enterprise & Government: aviation (<10% penetrated per management), maritime, Starshield. More than $6B of US government contracts were won in Q2 [Q2 2026 call].
• Capacity: V3 satellites (~10x V2 capacity, per management) on Starship. Starlink Mobile adds EchoStar's 65 MHz ($19.6B: $11.1B in stock, up to $8.5B in cash/debt payoff, expected to close November 2027) [Q2 2026 10-Q].
• Compute rental: nameplate draw of 1.4 GW (0.4 GW a year earlier), targeted at >2 GW by end-2026. About 90% of compute is to go to inference or rental, with ~10% for Grok training [Q2 2026 10-Q; Q2 2026 call].
• Margin swing factor: D&A, $2,848M in Q2 2026 versus $1,526M a year earlier [Q2 2026 10-Q].

7. Financial profile

FY2025: revenue $18,674M (+33%), operating loss $(2,589)M, net loss $(4,937)M, adjusted EBITDA $6,584M, operating cash flow $6,785M, capex $20,737M (AI $12,727M), SBC $1,947M [424B4, FY2025]. FY2024 was profitable at the operating line ($466M) on capex of $11,163M.

Q2 2026: revenue $7,814M (+92%), operating loss $(143)M, net loss $(541)M, adjusted EBITDA $3,538M. H1 2026 net loss of $(4,817)M includes $1,545M of debt-extinguishment losses from refinancing X/xAI debt that carried coupons up to 12.5% [Q2 2026 10-Q].

Balance sheet at June 30, 2026: cash and marketable securities $100.0B; debt and finance leases $39.4B, including $25B of SpaceX Notes (5.855% weighted coupon, first maturity 2031) and $13.3B of related-party Valor equipment financings recorded as failed sale-leasebacks; $5B undrawn revolver; net cash ~$60.6B. Non-cancellable commitments are $27,955M, of which $22,244M falls in 2027 [Q2 2026 10-Q]. Backlog was $47.5B (56% within a year), up from $28.4B at December 2025, though that excludes most of the value of terminable cloud contracts.

8. Valuation

There is no earnings base yet to capitalise. The stock trades at ~93x trailing revenue and ~260x trailing adjusted EBITDA with negative free cash flow; 69x Q2 annualised revenue ($31B); and ~21x management's $100B December 2026 ARR target [Q2 2026 call]. The price assumes the targets are hit and that much of the compute revenue is durable. Starlink alone ($7.2B of FY2025 segment EBITDA) does not get close; the market is paying for the AI build and Starship optionality.

Market data callout — as of market close October 9, 2026 (source: stockanalysis.com web market data, outside sources/; not company filings). Share price $162.57 (IPO price $135.00 on June 11, 2026); 52-week range $104.83–$225.64; shares outstanding ~13.57B; market capitalisation ~$2.21T. Derived using filings: net cash ~$60.6B at June 30, 2026, so EV ≈ $2.15T → ~93x TTM revenue ($23.0B), ~259x TTM adjusted EBITDA ($8.3B), TTM FCF -$32.3B. The Cursor share count implies a pre-closing 7-day VWAP of ~$154 [8-K Aug 14, 2026]. Market data is not a company source and is used nowhere else in this file.

9. Management & capital allocation

Elon Musk is founder, CEO, CTO and Chairman, and held ~82% of voting power after the IPO through 10-vote Class B shares. SpaceX is a Nasdaq "controlled company" exempt from independent-majority board and committee requirements. Class B holders elect a board majority, and the bylaws impose mandatory arbitration and forum limits on shareholder claims [424B4]. Musk's 2026 awards are 1.302B restricted Class B shares, which he can vote before vesting. 1.0B vest only on market-cap milestones from $0.5T to $7.5T and a permanent Mars colony of one million people. 302M vest on milestones from $1.065T to $6.565T and off-Earth data centres delivering 100 TW of compute per year [424B4]. In practice these are control instruments more than performance pay. The CFO's 2024 options were re-based in January 2026 from free cash flow to adjusted EBITDA, an easier metric while capex dominates [424B4].

Related-party dealing is material. Director Antonio Gracias's Valor leases AI hardware to SpaceX under three agreements with $20.2B of aggregate payments, guaranteed by SpaceX, and Valor adopted a 10b5-1 plan to distribute up to 225.9M shares. SpaceX also buys Tesla Megapacks ($329M in H1 2026) and is building Terafab and Macrohard with Tesla, with capex not yet determined [424B4; Q2 2026 10-Q].

Capital allocation in 2026 has been decisive: an $85.7B IPO; X/xAI debt carrying up to 12.5% refinanced via a $20B bridge and then $25B of investment-grade notes; Cursor bought with ~$60B of stock; EchoStar spectrum part-paid in stock. No dividends are planned [Q2 2026 10-Q; 424B4].

Claims to verify over time: <1-year payback on new compute; $100B+ ARR in December 2026 ("not a question mark"); $1T of revenue in 2030 (moved up from 2031); >2 GW of compute by end-2026 and "closer to 10 GW" by end-2027; enterprise revenue to "substantially exceed" consumer; heat shield "solved", ship catch on the next flight and daily flights within a year; Mobile service by end-2027 with a capex-efficient terrestrial build [Q2 2026 call]. None is auditable from the filings.

10. Risks ranked by damage

1. Compute capital cycle and customer concentration Damage rank: High · Permanence: High · Fixability: Medium Mechanism: ~$65B of 2026 capex, mostly GPUs on short lives, is committed before rental demand is contractually secured beyond 90-day notice periods. Evidence today: Q2 AI capex $15,828M; AI D&A $1,885M a quarter; one AI customer was 19.5% of Q2 revenue and a second customer 18.3%; cloud agreements terminable on 90 days' notice after ramp, while SpaceX's own non-cancellable commitments total $27,955M, $22,244M of it due in 2027 [Q2 2026 10-Q]. 2026 capex is the author's arithmetic from Q1–Q2 actuals plus guidance for Q3–Q4. What would confirm it: sequential decline in AI infrastructure revenue, a customer notice, or capex held at Q2 levels with no new contracts. Impact: turns the AI segment's EBITDA turn back into losses and leaves stranded, depreciating hardware.

2. Starship schedule Damage rank: High · Permanence: Medium · Fixability: Medium Mechanism: V3 Starlink, Mobile V2, lunar contracts and orbital compute all need a reusable Starship at cadence. Evidence today: Space R&D $1,076M in Q2, rising; two V3-generation Starship flights (12 and 13) described as successful; payload delivery to orbit expected H2 2026 [Q2 2026 10-Q; Q2 2026 call; 424B4]. What would confirm it: no operational V3 deployment by mid-2027, or a vehicle loss that grounds the programme. Impact: delays the capacity step that underwrites Connectivity growth and much of the valuation.

3. Control, related parties and key-person dependence Damage rank: High · Permanence: High · Fixability: Low Mechanism: one person controls the board, sets strategy across SpaceX and Tesla, and transacts with affiliated parties. Evidence today: ~82% voting control; Valor financings of $13.3B on balance sheet (from $4.5B at December 2025), which carried $327M of Q2 interest, at least ~9.8% annualised on the June balance against a 5.855% coupon on the SpaceX Notes (author's arithmetic) [Q2 2026 10-Q]; Tesla supply and joint projects; mandatory arbitration [424B4; Q2 2026 10-Q]. What would confirm it: further related-party capital commitments (Terafab), or capital moved to projects without disclosed returns. Impact: outside holders cannot correct misallocation, and terms are set by insiders.

4. Starlink monetisation and spectrum Damage rank: Medium · Permanence: Medium · Fixability: Medium Mechanism: blended ARPU keeps falling as growth shifts to lower-income markets and cheaper plans, while market access depends on country-by-country licences. Evidence today: ARPU $85 → $66 y/y; Connectivity SG&A +132% y/y on marketing and expansion [Q2 2026 10-Q]. What would confirm it: consumer revenue growth falling below subscriber growth by a widening gap, or segment EBITDA margin trending below ~55%. Impact: slows the one segment that funds the rest.

5. Dilution and share overhang Damage rank: Medium · Permanence: High · Fixability: Medium Mechanism: acquisitions paid in stock, large equity plans and staged lock-up releases. Evidence today: ~391M shares for Cursor; 261.8M due to EchoStar; 1,331M performance awards excluded from diluted counts; ~1.3B shares unlock after Q3 results; Musk's ~6.4B shares from June 12, 2027 [8-K Aug 14, 2026; Q2 2026 10-Q; 424B4]. What would confirm it: further share-funded acquisitions. Impact: per-share value lags company value.

6. X and Grok legal and regulatory exposure Damage rank: Medium · Permanence: Low · Fixability: Medium Mechanism: platform and model liability across the EU and US. Evidence today: EUR120M DSA fine under appeal; putative class actions over Grok-generated sexualised images; Clean Air Act suit over Colossus II turbines; $105M (+$67M interest) Vidstream verdict on appeal; $354M litigation accrual [Q2 2026 10-Q]. What would confirm it: an injunction on Colossus II power or a large adverse class judgment. Impact: cost and reputational drag on advertising and enterprise AI sales.

11. Metrics to monitor

12. Thesis tripwires

13. Open questions

• Who are Customers A (18.3% of Q2 revenue, all three segments) and B (19.5%, AI)? The filings do not name them. B's size matches the Anthropic agreement's run-rate, but that is inference.
• How much of the cloud contracts' headline value is enforceable, and how much of the $6.7B July contract and the Google deal is in backlog? (See §9 on payback.)
• What do a Starlink terminal and a consumer subscriber cost and earn, and what is churn? Not disclosed.
• What is Cursor's revenue and margin, and how much of the $100B ARR target does it contribute?
• What will Terafab cost, and on what terms will SpaceX and Tesla split it?
• How will the EchoStar cash leg (up to $8.5B) and Mobile terrestrial build be funded? Management declined to give capex [Q2 2026 call].

14. Update log

15. Source map