SpaceX (SPCX) · Deep analysis

Report previewSPCX.USReport date Last updated
← Back to reports
Report contents
01

Business model

Business model: a subscription engine funding two capital pits

SpaceX sells three things: launch missions for commercial satellite operators and government agencies; Starlink broadband subscriptions to households, airlines, shipping carriers and defense customers across 167 countries; and AI compute rented to enterprises and government buyers under multi-year Cloud Services Agreements. Starlink is the largest segment and the profit engine. AI now contributes roughly a third of revenue and absorbs most of the capital spending. That makes SpaceX an expansion-stage conglomerate: a subscription business bolted to a launch manufacturer, with a young compute landlord attached. The subscriptions matter most, because they are the only revenue that can plausibly fund the other two.

MetricLatest
Starlink subscribers (million)12.0
Starlink monthly ARPU (USD)66, from 85
Starlink satellites in orbit10,200
AI compute nameplate capacity (GW)1.4, from 0.4
AI Cloud Services Agreements (USD bn)14.1
H1 2026 launches / mass to orbit (t)78 / 1,041
H1 2026 operating cash flow vs. capex (USD bn)3.47 vs. 28.5
Connectivity pays; the other two spend

Connectivity is the only segment producing revenue and profit at the same time. Space, the original business, shrank 2% in the first half of 2026 as external launches and mass to orbit both fell; its segment Adjusted EBITDA turned negative while Starship research spending rose 55%. AI is growing fastest from a small base and takes the capex. In its first results as a public company, published in August 2026, SpaceX reported revenue up 92% and Adjusted EBITDA up 191%, still at an operating loss. That growth mixes three qualities of revenue: recurring subscription fees, contracted compute, and lumpy third-party launch work. Segment profitability is reported only at the Adjusted EBITDA line, so the profit mix has to be inferred, and the inference is that Connectivity carries the other two.

Starlink: more subscribers, lower revenue per subscriber

Subscribers doubled to 12.0 million, but average monthly revenue per user (ARPU) fell from $85 to $66 as growth moved into lower-priced international markets. The volume is real. SpaceX says its reusable boosters give it the lowest cost per kilogram to orbit, so adding capacity for a new market costs less than a rival would pay. What the disclosure does not settle is whether cost per subscriber falls faster than price. There is no subscriber-level gross margin, and the 10,200-satellite fleet is a depreciating asset that must be replaced on a cycle the filings do not break out, so replacement capex arrives whether or not ARPU recovers. Growth won through geography is still growth; it is not yet evidence of pricing power.

Bargaining power differs by buyer. Households can switch, and Amazon's Leo is the best-funded alternative at roughly 300 satellites, a small constellation but a credible one over time. Airlines and carriers sign multi-year capacity deals, and governments buy through procurement competitions that set their own terms. SpaceX's leverage is highest where terminals are installed and contracts run for years, which is why the enterprise and government share deserves more attention than the household count.

Compute and launch: two different pricing tests

The AI segment signed $14.1 billion of Cloud Services Agreements while nameplate capacity, the installed compute SpaceX can power and cool, rose from 0.4 to 1.4 gigawatts. Capacity plus contracts is not billed revenue: multi-year agreements become sales only as equipment is delivered and invoiced, and depreciation starts when the equipment does. SpaceX does not name the counterparties or disclose their concentration. If a handful of buyers account for most of that $14.1 billion, the segment's durability rests on their budgets and credit; a broad customer market is not yet evident.

Launch has the opposite problem: an old, observable customer base and a new competitor. External missions are the only part of Space that tests third-party pricing, because Starlink's own launches are internal cost transfers. First-half external launches declined, which reads as the manifest shifting inward rather than customers defecting. Rocket Lab's Neutron targets Falcon 9 economics, with a first flight aimed at the fourth quarter of 2026, so the low end of the market could face genuine price competition within about two years. An internal launch saves a cost; an external one earns a price.

What would change the judgment

Enterprise and government now supply 42% of Connectivity revenue, including more than $6 billion of multi-year Starshield contracts (SpaceX's defense satellite business). Those commitments run for years and are less price-sensitive than households, so a rising share is the most plausible offset to falling consumer ARPU. The counterweight is capital intensity: first-half operating cash flow of $3.47 billion against $28.5 billion of capex, with Adjusted EBITDA of $3.54 billion sitting on an operating loss, means much of the reported profitability is depreciation added back on assets still under construction. The judgment would break if AI capacity is billed at high utilization while the enterprise mix climbs. If ARPU keeps sliding and compute contracts sit underused, growth is consuming more capital than it returns.

Questions worth following

  • Can Starlink grow subscribers without cutting consumer prices in its largest markets?

    Pricing power decides whether Starlink's growth is demand-driven or bought. If ARPU holds while subscribers rise, subscription revenue can repay terminal and satellite costs; if growth requires discounts, revenue trails users and the capital burden compounds.

  • Will SpaceX fill its 1.4 GW of AI compute capacity?

    AI supplies about a third of revenue but most of the capex, on the strength of $14.1 billion of Cloud Services Agreements as capacity rose to 1.4 GW. Revenue depends on billed utilization, and the counterparties are not disclosed, so delivery decides whether the buildout earns a return.

  • Can Starshield and airline deals lift Connectivity as consumer prices fall?

    Enterprise and government supply 42% of Connectivity revenue, including over $6 billion of multi-year Starshield contracts, and those customers are less price-sensitive than households. Whether they keep renewing on similar terms decides if the mix can offset consumer ARPU declines and keep funding the rest of the company.

02

Financial quality

Financial quality: operating cash covers a third of the buildout

SpaceX lost money in four of its five most recent reported periods and still produced operating cash in every one of them. That split defines the financial quality here: the cash engine works, the earnings base does not yet exist, and the gap between what the business generates and what it spends is met with outside capital. In the June 2026 quarter, operating cash flow of $3.47 billion met capital spending of roughly $28.48 billion, leaving free cash flow at –$25.01 billion. Financing of $100.29 billion filled the hole, mostly equity: total equity rose from $41.58 billion to $127.22 billion in three months.

The company listed on June 12, 2026, so only three complete fiscal years are public. The annual record starts at FY2023 and there is no longer history to test how these ratios behave through a downturn. Quarterly disclosure is thinner still, at three periods.

USD bn unless notedFY2023FY2024FY2025Q1 FY2026 (3 mo.)Q2 FY2026 (3 mo.)Assessment
Revenue10.3914.0218.674.697.81Strong growth
Gross margin41.2%42.9%49.4%49.3%55.3%Strong
Operating income0.510.74–2.59–1.94–0.14Weak
Operating margin4.9%5.3%–13.9%–41.4%–1.8%Weak
Net income–4.630.79–4.94–4.28–0.54Very weak
Net margin–44.6%5.6%–26.5%–91.3%–6.9%Very weak
Return on equityNot disclosedNot disclosed–14.7%Not disclosedNot meaningfulVery weak
Operating cash flow4.525.786.781.053.47Adequate
Capex (derived)4.4111.1720.7310.1128.48Weak
Capex/revenue42%80%111%216%365%Very weak
Free cash flow0.11–5.39–13.95–9.06–25.01Very weak
CashNot disclosed11.3824.7515.8593.52Strong
Total assetsNot disclosed57.0692.08102.09192.77Context
Liabilities/assetsNot disclosed54.8%55.1%59.3%34.0%Adequate
Interest expense1.691.581.940.660.63Adequate
Interest/operating cash flow37%27%29%63%18%Adequate
USD bnRevenueNet incomeTotal assetsTotal liabilitiesFixed assetsOperating cash flowCapex (derived)Free cash flowAssessmentTrend
FY2025 Q1 (3 mo.)4.07–0.53Not disclosedNot disclosedNot disclosed0.734.14–3.41WeakBaseline
FY2026 Q1 (3 mo.)4.69–4.28102.0960.51Not disclosed1.0510.11–9.06Very weakDeteriorating
FY2026 Q2 (3 mo.)7.81–0.54192.7765.55Not disclosed3.4728.48–25.01WeakEarnings up, burn worse
Cash conversion carries the quality case

Operating cash flow has risen in each of the three fiscal years, from $4.52 billion to $6.78 billion, even as the net result swung from a $4.63 billion loss to a $0.79 billion profit and back to a $4.94 billion loss. Depreciation explains most of that resilience: FY2025 EBITDA of $4.29 billion against negative EBIT of $2.41 billion implies about $6.70 billion of depreciation and amortization. Cash is being produced, but a large share of it is the accounting recovery of money already spent.

That is why the cash-flow statement and the balance sheet tell different stories. Capital spending was 3.1 times the implied depreciation charge in FY2025 and far above it in 2026, so this is expansion rather than replacement. Total assets climbed from $57.06 billion to $192.77 billion in eighteen months while operating income fell from a $0.74 billion profit to a quarterly loss of $0.14 billion. Returns on the newly added assets are not yet visible in the numbers, and the interest line that eventually has to be earned against them runs at 18–63% of quarterly operating cash flow.

Margins improved at the gross line only

Gross margin widened from 41.2% in FY2023 to 55.3% in the June 2026 quarter. Nothing comparable reached the operating line: the FY2025 operating margin was –13.9%, and the second-quarter recovery to –1.8% is one quarter against a full year of losses. The spread between the two lines is operating expense and depreciation, and FY2025 shows what happens when both grow faster than revenue. FY2024 was the only profitable year, at $0.79 billion, and it included a $0.55 billion tax benefit and $1.58 billion of interest expense, so the sole positive year clears zero narrowly. Return on equity of about –14.7% in FY2025 is the more honest return figure; for the June 2026 quarter it is meaningless because equity tripled on issuance.

Leverage fell because equity rose, not because debt shrank

Liabilities-to-assets dropped from 59.3% to 34.0% between March and June 2026, which looks like deleveraging and is not. Liabilities actually rose $5.04 billion while assets rose $90.68 billion. Cash of $93.52 billion covers current liabilities of $21.12 billion more than four times over, so refinancing is not the near-term constraint. The constraint is that the cash has to become earning capacity before it is spent again. Interest expense of $1.94 billion in FY2025, against $0.93 billion of short-term debt and no disclosed long-term debt line, remains unexplained by the borrowings on the balance sheet, and that is worth resolving before treating the obligation as small.

What needs checking
  • Dilution. No share count appears in the financial statements. The June raise added $85.64 billion to equity in one quarter, so the dilution is already large; whether financing continues at that scale is the test.
  • The interest charge. It absorbed 29% of FY2025 operating cash flow and 63% in the March 2026 quarter, yet the debt lines do not account for it. Finance leases, other funding arrangements, or non-cash accretion would each change the cash timing.
  • Disclosure gaps. Inventory, receivables, and net PP&E are not broken out, so receivable days, inventory turns, asset turns, and capex-to-depreciation computed from the balance sheet are unavailable; capex here is derived as operating cash flow minus free cash flow.
  • The EBITDA bridge. FY2025 EBITDA exceeds EBIT by about $6.70 billion, while in the June 2026 quarter the same gap is roughly $0.25 billion. That does not fit an asset base that nearly doubled, so the reconciliation deserves a look before the margin improvement is taken at face value.
  • Goodwill. At $11.64 billion it is flat since FY2024 and only about 9% of equity, so impairment is a contained risk rather than a leading one.

What would change the reading is operating cash flow growing faster than capital spending for several quarters, with segment disclosure showing which activity earns a return. So far the evidence shows a business that converts revenue into cash reliably and converts cash into assets expensively.

Generated by DeepVal AI. Data and models may contain errors or be outdated. For research only; not personalized investment advice.

Go beyond the preview

Continue reading the full report

  1. Key risksWhat could break the investment case
  2. Valuation and margin of safetyWhat the current price assumes, and how much room for error it leaves
  3. SummaryOur overall judgement and what to watch next

Free to sign up · 20 research points · unlocking this report uses 10

20 points cover 2 full reports. No card needed.

Research reflects its generation date. Preview the first two modules.

More on Space Exploration Technologies Corp. Class A Common Stock

Space Exploration Technologies Corp. Class A Common Stock

Business, moat and key questions

Open company page