SpaceX's First Public Earnings: Starlink's $11.4B Profit Machine Is Bankrolling a $6.4B AI Bet

On June 12, 2026, Space Exploration Technologies Corp completed the largest IPO in history, raising roughly $85.7 billion at a $135-per-share offering price and opening trading at $150 — an 11% first-day pop. Buried inside the S-1 that made that possible was the first-ever public look at SpaceX's actual financial statements, and they tell a story far more complicated than "the world's dominant rocket company goes public": full-year 2025 revenue grew 33% to $18.7 billion, but the company swung from a $791 million profit in 2024 to a $4.9 billion net loss in 2025 — driven almost entirely by a newly merged AI segment (built around X and xAI) that lost $6.4 billion on just $3.2 billion of revenue while Starlink, the satellite-internet business, quietly threw off $4.4 billion in operating profit.

The Headline Numbers​

SpaceX's S-1 discloses three years of audited income statements (FY2023–FY2025) and two years of audited balance sheets (FY2024–FY2025), plus an unaudited Q1 2026 stub — the last full quarter before the IPO.

Gross margin actually improved sharply — Starlink's fixed satellite-network costs are spreading across a fast-growing subscriber base. But that improvement gets swallowed twice over further down the income statement: research and development spending more than doubled, from $3.46 billion to $8.64 billion, and the company booked $1.67 billion of net financing and other costs (interest expense, minus interest income, plus a small impairment). A company that grew revenue by a third and improved gross margin by 6.5 points still lost $4.9 billion. That gap is the whole story of this filing.

Revenue Deep Dive: Three Segments, One Very Uneven Split​

Since 2025, SpaceX reports through three segments — Space (launch), Connectivity (Starlink), and AI (X and xAI, following their mergers into the company) — and the segment operating-income table is the real headline the consolidated numbers hide.

Connectivity ($11.4B, +50% YoY): This is the business. Starlink crossed 10.3 million active subscribers across 160 countries as of March 31, 2026, more than double the 4.6 million it had at the end of 2024, and it now generates a 38.8% operating margin — a software-like margin on a satellite constellation. Connectivity alone now accounts for 61% of SpaceX's total revenue and effectively all of its consolidated operating profit.

Space ($4.1B, flat operating income near breakeven): The original SpaceX business — Falcon 9, Falcon Heavy, Starship launches — is barely profitable despite dominating the global commercial launch market. Research and development on Starship's third-generation vehicle (Starship V3, which flew successfully for the first time on May 22, 2026) pushed the segment to a $657 million operating loss in 2025, up from roughly breakeven the year before.

AI ($3.2B, −$6.4B operating loss): The newest and by far the most expensive segment. Revenue grew 22% year over year, but the operating loss nearly quadrupled from $1.56 billion to $6.36 billion, and R&D inside the segment alone hit $5.06 billion in 2025 — more than the segment's entire revenue. This is the segment absorbing nearly all of the company's incremental spending, and it is the reason a company with a 39%-margin core business reports a consolidated net loss.

The Margin Story​

Gross margin has been climbing steadily for three years as Starlink's unit economics mature — that part of the business is working exactly as a satellite-internet bull case would predict. Everything below gross profit is a different story: operating and net margins have been negative in three of the last four periods shown, and Q1 2026 — the freshest data point, and the last quarter before the IPO — was the worst yet, with a net margin of −91%. FY2023's large loss was driven by a one-time $3.78 billion impairment tied to the original X integration; FY2025's and Q1 2026's losses are not one-time charges — they are ongoing operating spend.

The One Big Question: Is SpaceX Now an AI Infrastructure Company?​

Look at where the capital is actually going, not just where the losses are booked:

The AI segment consumed $12.7 billion of capital expenditure in 2025 — 61% of the company's total, and four times the segment's own revenue. In the first quarter of 2026 alone, AI capex hit $7.72 billion, already 61% of all of FY2025's AI spend in a single quarter, and more than the entire Space and Connectivity segments' capex combined. Depreciation on that infrastructure is only going to grow from here, meaning the AI segment's losses are more likely to widen than narrow over the next several quarters before any of that compute translates into revenue.

The honest framing: SpaceX is currently two businesses sharing one balance sheet. Connectivity is a mature, highly profitable satellite-internet company. Space is a capital-intensive but roughly breakeven launch business subsidizing its own next-generation rocket program. AI is a bet — funded by the other two segments and by $85.7 billion of fresh IPO capital — that X and xAI's compute buildout eventually produces a business with anything like Starlink's margins. Nothing in the numbers yet proves that bet will pay off; nothing yet disproves it either.

What Wall Street Thinks​

Six major banks initiated coverage in the first days after the post-IPO quiet period lifted (July 7, 2026), and the dispersion in their targets says as much as their averages:

• Raymond James (Brian Gesuale) initiated at the Street-high $800 target, implying roughly 425% upside from the stock's early-July trading level, betting heavily on the AI and Starship optionality.
• Morgan Stanley (Adam Jonas) set a base-case target of $300, about 87% above the July 7 price of $160.42.
• Goldman Sachs (Eric Sheridan) initiated at Buy with a $205 target.
• Morningstar was the outlier, initiating at Sell with a $63 target, flagging the AI segment's cash burn against the company's $1.9+ trillion valuation.
• Consensus across 27 analysts: Strong Buy, with an average target of $245.96 against a $115–$800 range — an unusually wide spread for a newly public mega-cap, reflecting genuine disagreement over how to value the AI segment.

The stock itself has already lived through that disagreement: it hit an all-time high of $225.64 on June 16, four days after the IPO, then fell to an all-time low of $145.07 on July 10 — down roughly 36% from peak — before stabilizing near $145–152 and a $1.93 trillion market cap.

Tracking a $1.9 Trillion Company in Plain Text​

One of the most clarifying exercises for any investor is to model a company's finances in Beancount, the open-source double-entry accounting system — it forces every dollar to reconcile, which is especially useful for a company as unusual as SpaceX: a mezzanine preferred-stock structure from a decade of private funding rounds, a bitcoin treasury, and three segments with wildly different economics all rolled into one balance sheet.

We modeled SpaceX's FY2024–FY2026 Q1 income statements and balance sheets in plain-text Beancount files, sourced entirely from the S-1's audited and unaudited financial statements. The currency unit is MUSD (millions of USD). Note the Beancount convention: Income accounts carry negative (credit) balances, Expenses carry positive (debit) balances:

; FY2025 Income Statement — fiscal year ended December 31, 2025
; 1 MUSD = USD 1,000,000  |  All figures in millions USD
; Check: −18674 + 9451 + 8643 + 2644 + 487 + 1668 + 718 + (−4937) = 0 ✓
 
2025-12-31 * "Space Exploration Technologies Corp" "FY2025 Income Statement"
  Income:Revenue                         -18674 MUSD  ; revenue earned (credit)
  Expenses:CostOfRevenue                   9451 MUSD  ; cost incurred (debit)
  Expenses:ResearchAndDevelopment          8643 MUSD  ; cost incurred (debit)
  Expenses:SellingGeneralAdministrative    2644 MUSD  ; cost incurred (debit)
  Expenses:Restructuring                    487 MUSD  ; workforce-reduction charges (debit)
  Expenses:OtherNet                        1668 MUSD  ; net interest + other expense + $38M impairment (debit)
  Expenses:IncomeTax                        718 MUSD  ; cost incurred (debit)
  Equity:Adjustments                      -4937 MUSD  ; net loss offset (RE set by balance assertion)

The balance sheet needed one account the standard chart doesn't have: Equity:RedeemableConvertiblePreferredStock, for the roughly $38.8 billion of pre-IPO preferred stock still outstanding at the end of FY2025 — a mezzanine equity structure sitting between liabilities and common equity, standard for a decade-old private company but unusual once you're used to modeling public companies. By Q1 2026, that balance had collapsed to $7.0 billion as preferred converted to common ahead of the IPO — the clearest single number in the ledger for "this company just went public":

; FY2025 → Q1 2026: preferred stock converting out ahead of the IPO
2025-12-31 balance Equity:RedeemableConvertiblePreferredStock -38752 MUSD  ; $38.8B, 2,046 shares outstanding
2026-03-31 balance Equity:RedeemableConvertiblePreferredStock -7049 MUSD   ; $7.0B, 135 shares outstanding

The complete ledger — three years of income statements, two full balance sheets, and the pre-IPO quarter — is open and auditable:

The Multi-Year Arc: From Rocket Company to Three Companies​

Two things compound at once across this arc. First, revenue is growing at a healthy clip — roughly 33% in FY2025 alone — almost entirely on Starlink's back. Second, the AI segment's operating loss has been on a wild ride: a $3.97 billion loss in FY2023 (mostly a one-time impairment during the initial X integration), a much smaller $1.56 billion loss in FY2024, and then a fresh acceleration to $6.36 billion in FY2025 and an annualized pace well above that in Q1 2026. This isn't a segment converging toward profitability — it's one whose losses are currently getting larger, not smaller, even as the rest of the company matures.

The Verdict: Bull vs. Bear​

The Bull Case:

• Starlink is a genuinely excellent business on its own numbers: $11.4 billion revenue, 38.8% operating margin, subscribers more than doubling in fifteen months to 10.3 million across 160 countries — a satellite-internet company with software-like economics that alone could justify a large fraction of the current valuation
• Starship V3's first successful flight (May 22, 2026) is designed to carry up to 60 next-generation Starlink satellites per launch, roughly 20x the network capacity of a Falcon 9 mission — if it reaches operational cadence in H2 2026 as guided, both launch costs (helping the Space segment) and satellite deployment economics (helping Connectivity) improve simultaneously
• The $85.7 billion raised in the IPO funds the AI buildout without near-term debt or dilution pressure, buying the company years of runway to prove out the thesis
• Raymond James's $800 target and Morgan Stanley's $300 target both imply the market is still pricing meaningfully less than some sophisticated investors think the combined entity — particularly Connectivity — is worth
• Gross margin expansion (42.9% → 49.4% → 49.1% across FY2024, FY2025, and Q1 2026) shows the core cost structure is improving even while losses widen further down the P&L, meaning the losses are a spending choice, not a structural weakness

The Bear Case:

• The AI segment lost $6.36 billion in FY2025 on $3.20 billion of revenue and is capital-consuming at an even faster pace: $7.72 billion of capex in Q1 2026 alone, more than 60% of all of FY2025's AI capex in a single quarter, with no disclosed timeline for the segment to approach breakeven
• The Space segment — the original, market-dominating rocket business — is itself unprofitable (−16.1% operating margin in FY2025), meaning Connectivity is currently the only segment carrying its own weight, let alone subsidizing the other two
• Net losses are accelerating, not narrowing: $791 million profit in FY2024 became a $4.94 billion loss in FY2025, and Q1 2026 alone lost $4.28 billion — nearly matching the entire prior year's loss in a single quarter
• Long-term debt and finance leases grew from $13.4 billion (FY2024) to $28.7 billion (Q1 2026), more than doubling in five quarters, with interest expense in Q1 2026 alone ($664 million) already 42% of all of FY2024's interest expense
• The 36% collapse from the stock's June 16 all-time high to its July 10 all-time low, and the $63-to-$800 analyst target spread, both signal that professional investors have not converged on how — or whether — to value the AI segment at all

Our Take: SpaceX's IPO filing reveals a company that is really two very different businesses bolted together by a shared balance sheet and a shared brand. Connectivity is as good a business as the market believes SpaceX to be: high-margin, fast-growing, globally scaled. Space and AI are not — Space is roughly breakeven while it reinvests in Starship, and AI is burning cash faster than any of SpaceX's historical launch-business losses ever did, with capex outpacing segment revenue by four to one. That doesn't make the AI bet wrong; xAI and X give SpaceX a call option on a second trillion-dollar market, funded by a proven cash generator instead of debt. But it does mean the $1.9 trillion valuation currently prices in that the AI segment eventually looks something like Connectivity does today. Nothing in three years of financial statements shows that happening yet — the losses are getting bigger, not smaller. The next few quarters of AI-segment revenue growth, not Starship's launch cadence, are what will actually settle the argument between Raymond James's $800 and Morningstar's $63.