SpaceX: three businesses, one balance sheet

SpaceX, Starbase, Texas, October 2026

SpaceX reported its first profitable quarter as a public company on October 20 and the stock rose about 8 percent after hours. The headline, revenue up 161 percent to $14.58 billion, is true and is also the least useful number in the report. The company is three businesses carried on one balance sheet. Connectivity, which is Starlink, made $2.37 billion of operating income on $5.25 billion of revenue and is the only segment that has been profitable in every quarter the company has reported. Super Intelligence, the compute and AI business built from the February merger with xAI, produced $7.56 billion of revenue and $431 million of operating income, and it spent $16.44 billion on capital equipment in the same 92 days. Space, which is the rockets, made $1.77 billion and roughly broke even, because three quarters of its launches carry Starlink satellites and bill no one.

The question the price asks is simple. At $147.79, the October 20 close, the company is worth about $2.0 trillion. That is about 61 times the last twelve months of revenue, 34 times the third quarter annualized, and 14 times the midpoint of the first quarter 2027 revenue guidance annualized. Whether any of those is the right multiple depends on whether the compute business, which did not exist at this scale a year ago, keeps growing at the rate management guided and keeps its two largest customers past the dates on which each is allowed to walk. This piece lays out what the filings say about each business, what the price implies, and where the company touches Texas, which is everywhere.

The author holds shares. That is disclosed at the top of this piece and again at the rating, because a rating on a stock the writer owns deserves a reader who knows it.

Rating: Buy

Street average price target: $233.50 (46 analysts, MarketBeat, October 21, 2026; range $75 to $800). StockAnalysis's narrower set of 38 analysts averages $222.71. The October 20 close was $147.79.

A rating is a judgment, so we show the judgment. We went through this company the way we would go through a business we were being asked to buy outright, asking the questions an owner asks, and we wrote down the answers before we wrote down the rating. The Street target is reported for reference. It is not our target, and we do not publish one, because a single number pretends to a precision we do not have on a company that nearly doubled its revenue in one quarter.

Do we understand how it makes money? Two of the three businesses, yes. Starlink sells a monthly connection to 13.6 million households and a growing list of airlines, ships and governments, over a network nobody else has built. Launch sells rides to orbit at a price and cadence nobody else can match. The third business is harder to read. Super Intelligence rents compute by the month to two customers whose names the company will not print, on hardware that any buyer with enough money can order from Nvidia, and sells advertising on X. We understand the arithmetic of it, so many gigawatts at so many dollars a month, but we do not yet know what the equipment is worth in year four or what a customer pays when a competitor's data center opens next door. We rate the company knowing that the fastest growing third of it is the third we understand least.

Where is the moat, and how wide is it? In launch, very wide and getting wider. SpaceX flew 165 of the world's 325 orbital launches last year, its rivals are grounded or late, one of its boosters has flown 37 times, and the gap is measured in years of engineering no one else has done. In Starlink, wide. The constellation is about 11,100 working satellites, the licenses cover 174 markets, the direct to phone spectrum portfolio is being assembled piece by piece, and the nearest competitor has 396 satellites and no customers. In compute, there is no moat in the hardware. The advantages are speed, in that the company brought 1.7 gigawatts online in a year, and power, in that it owns the sites and the pipeline to them. Those are real but they are leads, not walls. A lead can be closed by anyone with the same budget, and the hyperscalers have guided to $720 billion of capital spending this year between four of them.

Does it have pricing power, or does it have to cut? Mixed, and the direction is worth stating plainly. Starlink's consumer price has come down: average monthly revenue per account was $85 in the second quarter of 2025 and $69 now, and the lowest American plan went from $120 to $50, and to $39 in a March promotion. That is a company choosing volume over price in a market it already leads, which is what a monopolist does when it wants to stay one. Launch prices are set in competitive government awards and SpaceX keeps winning them. Compute prices are fixed in contracts through 2029 at $1.25 billion and $920 million a month, which is pricing power for as long as the customer stays.

What does the capital earn? This is where the three businesses separate. Connectivity spent $1.5 billion on equipment in the quarter and earned $2.4 billion of operating income, so it pays for its own growth and sends the rest upstairs. Space spent $1.6 billion and earned nothing, by design, because its customer is the other two segments. Super Intelligence spent $16.4 billion in the quarter and $40 billion in nine months and earned $431 million of operating income, which is about 4 percent annualized on the equipment bought this year alone. The segment's adjusted EBITDA, $4.95 billion in the quarter, annualizes to a far better looking figure, about 50 percent on the year's spending, but that number adds back depreciation on hardware that loses most of its value in a few years and $1.2 billion of stock paid to acquired employees. An owner would count the depreciation. We do.

Is management allocating capital like an owner? The founder owns about 42 percent of the company and controls 82 percent of the vote, so he is one. His record on the two older businesses is the reason the moats exist. The questions are about the newest one. The company paid $42.5 billion in purchase accounting for Cursor and will expense $23.4 billion more in stock to keep its people, for a business that contributed $608 million of revenue in seven weeks. It leases AI equipment from a firm whose founder sits on the board, which is $12.9 billion of the $38.8 billion of debt. And it is spending at a rate, $83 billion to $88 billion this year, that the operating cash flow, $13.1 billion in nine months, does not approach. None of this is hidden; all of it is in the filings. It is a management that has earned the right to be trusted with a lot of capital and is now testing how much.

Can the balance sheet carry a mistake? Yes, for now. $89.9 billion of cash and securities against $38.8 billion of debt, with $2.4 billion of principal due in 2027 and bonds priced in June at a weighted 5.855 percent. The guided fourth quarter can be paid from cash on hand and would leave about $60 billion. The company has said it is talking to lenders, and the rating agencies have it one and two notches above the bottom of investment grade. A missed year would be expensive. It would not be fatal.

What would we pay, and what is the price? We tried the owner's exercise of valuing the parts at what comparable businesses sell for. Starlink's annualized EBITDA of $14 billion at the 8 times a wireless carrier gets is $112 billion; at 15 times, which a business growing 69 percent a year can argue for, it is $210 billion. The compute segment's annualized $19.8 billion at the 21 times CoreWeave trades at is $416 billion. The rockets, roughly breakeven, carry a value we cannot calculate and will not pretend to. Add $51 billion of net cash and the parts come to somewhere between $600 billion and $700 billion plus the rockets. The market price is $2.0 trillion. The difference is the growth the company has guided to: at the midpoint of first quarter 2027 revenue, $35 billion in a quarter, and a 60 percent EBITDA margin, which is about what the third quarter showed, the company would be earning $84 billion a year before depreciation, and 21 times that is $1.76 trillion. So the price today is next spring's guidance at a cloud multiple. A buyer at $147.79 is paying for a quarter that is five months away and partly under contract, and is paying nothing for what comes after it. That is the whole valuation case in two sentences, and a reader should decide which half of it they believe.

What breaks it? Three things, in order of how fast they would show up. A termination notice on either compute contract, which both customers are free to send after the windows close. A quarter in which capital spending does not fall when guided and cash falls faster than $60 billion. And a Starship that cannot be caught, reflown and refueled on something near the schedule, because the lunar contract, the V3 constellation, the Louisiana spaceport and the Mars ambition all sit on it.

Our rating: Buy. We arrive there the way an owner would, by weighing the two businesses we would buy on their own against the one we are being asked to pay for in advance. Starlink alone is a business we would own at the price the parts imply; the moat is real, the growth is 69 percent, and it pays for itself. Launch is the strongest position in any industry we cover. The compute business is the bet, and what tips us to Buy rather than Hold is that the bet is written down: the two contracts are in the prospectus at fixed fees, the equipment is already running, the balance sheet pays for the next quarter in cash, and the customers, by the chief executive's account and by the hyperscalers' own capital budgets, are short of exactly what SpaceX is building. We hold the rating with the valuation objection stated in full above, and we expect to be wrong about the timing more than the direction. The rating goes to Hold on a termination notice from either compute customer, on year end cash below $55 billion, or on a Starship cadence that has not reached weekly by the end of 2027. It goes to Sell if Super Intelligence posts an operating loss after this quarter's profit or if the company issues stock to pay for compute.

The author holds shares, bought before this rating was written. The rating would be the same if he did not; readers will have to take that on the evidence above.

What the company is now

The SpaceX that listed on June 12, 2026 is not the rocket company of the prospectus era. The sequence matters. On March 28, 2025 xAI absorbed X, the social network. On February 2, 2026 SpaceX acquired xAI in an all stock exchange, converting xAI's preferred shares into about 1.99 billion SpaceX shares the company valued at $37.5 billion in its equity statement. In April the company took an option on Cursor, the coding software maker, at an implied equity value of $60 billion, exercised it on June 16, and closed on August 14 by issuing about 389 million Class A shares. Under purchase accounting the consideration was $42.5 billion, with $23.4 billion of the equity awards treated as compensation for future service rather than purchase price.

The IPO priced on June 11 at $135 a share. The company sold 638.9 million Class A shares including the full overallotment, raised $85.7 billion net, and listed on the Nasdaq Global Select Market and on Nasdaq Texas under the ticker SPCX. It is a Texas corporation with its principal office at 1 Rocket Road, Starbase, Texas 78521. The first day close was about $161. The all time high is $225.64 on June 16. The low is $104.83 on August 3, the day before the second quarter report. The stock joined the Nasdaq 100 on July 7 under a rule change that allows entry 15 trading days after listing. It is not eligible for the S&P 500 until it has traded twelve months and shown GAAP profit in the most recent quarter and across the trailing four, which puts the earliest date in June 2027.

Control sits with the founder. Class B shares carry ten votes and the prospectus put Elon Musk at about 82.4 percent of the voting power. He also holds about 1.3 billion restricted Class B shares that vest on market capitalization milestones plus two operational conditions written into the filing: one tranche requires a permanent human colony on Mars with at least one million inhabitants, the other requires data centers off Earth delivering 100 terawatts of compute per year.

Three operating segments report separately. The company renamed the Artificial Intelligence segment "Super Intelligence" in the third quarter release. Nothing changed in what it contains.

The quarter in numbers

The table is the company's segment disclosure, in millions of dollars. First quarter figures are derived by subtracting the second and third quarters from the nine month totals.

	Q3 2025	Q1 2026	Q2 2026	Q3 2026
Space revenue	1,666	619	962	1,767
Connectivity revenue	3,097	3,257	4,291	5,248
Super Intelligence revenue	813	818	2,561	7,560
Total revenue	5,576	4,694	7,814	14,575
Space operating income	259	(662)	(542)	(7)
Connectivity operating income	1,190	1,188	1,656	2,367
Super Intelligence operating income	(1,722)	(2,469)	(1,257)	431
Total operating income	(273)	(1,943)	(143)	2,791
Adjusted EBITDA	2,169	1,127	3,538	8,809
Capital expenditures	6,880	10,107	18,369	19,593
Net income	(1,054)	(4,276)	(541)	4,404

Two things in the third quarter deserve more attention than the revenue line.

First, the net income. Operating income was $2.79 billion. Net income was $4.40 billion, or 35 cents a diluted share on 12.48 billion weighted shares. The difference is interest income of $873 million on the cash raised in June, other income of $1.03 billion, and a tax benefit of $390 million, less $675 million of interest expense. The business earned $2.8 billion. The balance sheet and the tax line contributed the rest. Anyone annualizing 35 cents should know that.

Second, the compensation. Share based compensation was $2.36 billion in the quarter, of which $1.22 billion related to the Cursor acquisition. Adjusted EBITDA of $8.81 billion adds all of it back. The Super Intelligence segment's adjusted EBITDA of $4.95 billion, a 66 percent margin, sits next to a 6 percent operating margin for the same segment. Both numbers are real. The gap is depreciation on $89.5 billion of net property and equipment, which will grow with every quarter of capital spending, and stock paid to people the company bought.

Connectivity: the business that pays

Starlink is the profitable core. Connectivity revenue was $5.25 billion in the quarter, up 69 percent from a year earlier and 22 percent from the second quarter, with operating income of $2.37 billion, a 45 percent margin. The company said gross margin was 55 percent including about three points of one time benefit from tariff duty recoveries.

Consumer accounts reached 13.6 million at September 30, up 1.6 million in the quarter and up from 7.2 million a year earlier. Average revenue per consumer account was $69 a month, up from $66 in each of the first two quarters and down from $83 a year earlier and $85 in the second quarter of 2025. The company attributes the sequential rise to customer mix in developed markets. The decline from last year is pricing: the lowest residential plan in the United States fell to $39 a month in a March promotion and standard list is now $55, $85 and $130 by speed tier. Thirteen point six million accounts at $69 for three months is $2.8 billion, which is most of the $3.12 billion of consumer revenue; the rest is hardware and other charges.

The faster line is enterprise and government, $2.13 billion in the quarter, up 91 percent from a year earlier, and the president said the company remains on pace for it to exceed consumer revenue. Three pieces drive it. Starshield, the government business, won more than $6 billion of multiyear Space Force contracts in the second quarter, the largest a $2.29 billion fixed price order for the Space Data Network backbone with a prototype due by the end of 2027. Aviation reached more than 3,800 aircraft equipped and roughly 12,000 under contract, which the company puts at almost a quarter of the available fleet, with Lufthansa, Swiss, Copa, Korean Air, Austrian and Asiana flying their first equipped aircraft in the quarter and agreements with three of the four major American carriers. Starlink Mobile, the direct to phone service, passed 22 million monthly users at quarter end, up 37 percent from the prior quarter.

Mobile is where the spending is going. The company closed the purchase of EchoStar's AWS 4 and H Block spectrum in May, about 65 megahertz, and on October 8 agreed to buy Grain Management's nationwide 800 megahertz low band portfolio, up to 14 megahertz paired, with closing expected in the first half of 2027. The company has not disclosed a price. Reuters, Bloomberg and the Wall Street Journal reported about $8 billion in cash. On October 7 the FCC authorized up to 15,000 direct to device satellites at about 330 kilometers, half due in orbit by 2032. The president said the low band spectrum is what penetrates buildings, that the next generation Mobile satellites begin flying in 2027, and that the integrated service turns on in early 2028. That is the timeline against which the mobile carriers' public coolness should be read: the three national carriers have said they will not sign wholesale deals with SpaceX, and EchoStar's Boost Mobile will use the next generation service.

The constellation itself is about to change. Starship Flight 14 on September 28 put 26 Starlink V3 satellites in orbit, each designed for a terabit per second of downlink, about ten times what a Falcon 9 batch adds. The president's estimate is that a few hundred V3 satellites, achievable by the middle of 2027, is the point at which customers notice. The company has about 11,100 satellites working according to independent trackers, and it changed its reported definition this quarter to include satellites approved for deorbit, which would otherwise have removed about 400 from the count.

The competitive field is thin. Amazon's Leo, formerly Kuiper, had 396 satellites deployed as of mid October, has missed its mid 2026 service target, has not published consumer pricing, and depends in part on United Launch Alliance's Vulcan, which has flown once in 2026 and has been held off national security missions since February. Eutelsat's OneWeb has about 654 satellites and has ordered 340 more. China's two constellations had 186 and 238 satellites in orbit as of July.

Super Intelligence: the business that grows

The compute business is where the revenue growth and the capital spending both live. Segment revenue was $7.56 billion, up 830 percent from a year earlier and 195 percent from the second quarter. Of that, $7.02 billion was "solutions and infrastructure," which is renting compute, and $538 million was advertising on X. Cursor contributed $608 million from its August 14 closing. Nameplate compute reached 1.7 gigawatts at September 30, from 1.4 at June 30, 1.0 at March 31 and 0.6 a year earlier.

Two contracts in the prospectus explain most of the revenue. Anthropic signed in May for capacity across the Colossus sites, about 325,000 Nvidia GPUs, at $1.25 billion a month through May 2029, ramping in May and June at a reduced fee. Google signed on June 5 for about 110,000 GPUs plus supporting hardware at $920 million a month from October 2026 through June 2029, ramping through September at a reduced fee. At full rate the two are $2.17 billion a month. Super Intelligence revenue in the third quarter was $2.52 billion a month. The second quarter filing showed one customer at 19.5 percent of total company revenue, in this segment only, and another at 18.3 percent across all three segments. The company has not named either.

The termination clauses are the part of the prospectus that a reader of the revenue line needs. After the first three months, either party to the Anthropic agreement may terminate on 90 days' notice. After December 31, 2026, either party to the Google agreement may do the same, and Google could have terminated or taken a pro rata reduction if GPUs were not delivered by September 30. Three year terms with 90 day exits are rental agreements, not take or pay contracts, and the company's backlog of $63 billion at September 30, cited by the chief financial officer on the call, should be read with that in mind. The chief executive's description of demand was that the company is "hounded relentlessly by every hyperscaler and by dozens of smaller companies to rent more of our compute."

The guidance is the growth case stated by the company. The chief financial officer said nameplate compute will reach nearly 2.2 gigawatts by the end of October and 2.5 to 3 gigawatts by the end of December, that fourth quarter capital spending will be $35 billion to $40 billion against $19.6 billion in the third, that first quarter 2027 revenue will be $33 billion to $37 billion, and that the company is on track for $100 billion of annualized revenue measured on December. The chief executive said he is confident in adding 10 gigawatts of compute in 2027, would be surprised by less than 13, is aiming for 20 gigawatts of power and cooling infrastructure, and that his comments imply annual recurring revenue "somewhere between $500 billion and $1 trillion" by the end of 2027. For scale, the four largest hyperscalers have guided to a combined $720 billion to $745 billion of capital spending in calendar 2026. SpaceX's own 2026 capital spending, on the fourth quarter guidance, will be $83 billion to $88 billion.

Space: the business that carries the story

The rockets are the reason the company exists and the smallest line in the report. Space revenue was $1.77 billion in the quarter, $1.13 billion from launch services and $637 million from launch and development work, up 6 percent from a year earlier and 84 percent from the second quarter on customer mix. Operating income was a $7 million loss after losses of $662 million and $542 million in the first two quarters. The reason is in the operating metrics: 35 launches in the quarter and 113 for the year, of which 84 carried Starlink satellites and recorded no revenue in this segment. By mass, 83 percent of what the company put in orbit this year was its own. In 2025 SpaceX flew 165 of the world's 325 orbital launches and about four fifths of the mass, by BryceTech's count; the chief executive put the current share at almost 90 percent.

Falcon 9 is a mature product: one booster has flown 37 times, 157 of 165 Falcon 9 launches in 2025 used a flown booster, and on October 1 the company launched three missions from three sites in 13 hours. The customer pipeline added in the quarter is government heavy: 18 National Security Space Launch missions from Vandenberg worth $1.6 billion, three more NASA crew rotations worth $946 million that take the crew contract to $5.92 billion across 17 missions, three commercial missions for Starfall, the company's new cargo return capsule, and a lunar surface delivery on Starship.

Starship is the variable. Flight 12 in May was the first of the V3 vehicle and the first from the second pad. Flight 13 in July deployed 20 V3 satellites on a suborbital path. Flight 14 on September 28 reached orbit, deployed 26 operational V3 satellites, lost an engine on each stage during ascent, ended the ship's mission after three hours instead of ten, and brought the ship down upright in the north Pacific before it fell and exploded. The company counts it as the first revenue producing Starship mission because the satellites it carried are now in service. Flight 15 is planned for November with an attempt to catch both the booster and the ship at the tower; the president said the FAA license is expected to include that, which means it had not been issued at the time of the call. The stated goals are weekly flights in the first half of 2027, two or more a week by the end of it, a reflown second stage by early 2027, and an orbital propellant transfer demonstration in early 2027.

NASA's revised Artemis plan makes Starship's schedule a national one. Artemis III, in mid 2027, is now a low Earth orbit docking test with one or both commercial landers and no lunar landing. Artemis IV, targeted for early 2028, is the first landing, and NASA has said lander readiness decides which provider flies. The Starship lander contract is $2.89 billion plus a $1.15 billion option. The propellant transfer demonstration the landing depends on has not yet flown.

The competitors are grounded or late. Vulcan has launched once in 2026 and remains under investigation. Blue Origin's New Glenn has launched once, left its payload in an unusable orbit, and then lost its next vehicle in a May static fire that damaged the pad. Rocket Lab moved Neutron's first flight to 2027 on October 14.

The government customer

The United States government is SpaceX's largest customer across all three segments, and the direction of federal spending is the clearest external tailwind in the report. The Space Force's budget request for fiscal 2027 is $59.2 billion, more than double its $26.1 billion appropriation for 2026, with a further $12.1 billion proposed through reconciliation. The reconciliation package carries $17.1 billion for Golden Dome, the missile defense architecture, including $4.5 billion for space programs that could pay for a layer of space based interceptors. The two largest unclassified Space Force programs in the request are the Space Based Air Moving Target Indicator at $7.9 billion and the Space Data Network at $4.5 billion. SpaceX holds the largest award in each: $4.2 billion on the targeting constellation and $2.29 billion on the data network backbone, the two Space Force orders that made up the more than $6 billion of Starshield contracts the company announced in the second quarter. The backbone contract calls for a working prototype by the end of 2027 and is described by the Space Force as the communications layer of Golden Dome.

The launch side is as concentrated. The 18 National Security Space Launch missions added in the third quarter are worth $1.6 billion and fly from Vandenberg by the end of 2027, carrying the same sensing and targeting satellites; the Space Force awarded them on what its program office called a two month acquisition timeline. In the heavier lane of the same program, SpaceX was assigned five of the seven fiscal 2026 missions for $714 million, United Launch Alliance two, Blue Origin none. In 2025 SpaceX flew 11 of 12 national security medium and heavy missions. When Vulcan was held off the manifest in February, a GPS satellite moved to Falcon 9 and launched in April. NASA's commercial crew contract now stands at $5.92 billion across 17 missions after the September award of three more, and the Starship lunar lander contract is $2.89 billion plus a $1.15 billion option. The National Reconnaissance Office's 2021 contract for a network of imaging satellites was reported by Reuters at $1.8 billion.

The logic of the demand is that the military now treats orbit as a theater. Missile warning, targeting, communications and the proposed interceptor layer are all satellite programs, and each needs launch at a cadence only one provider currently offers, a constellation bus that is already in mass production, and a secure network that already exists. Starshield sells all three. The risk in that position is the one every defense contractor carries: budgets are proposed, not passed, and the fiscal 2027 request sits with a Congress that has not yet enacted either the reconciliation bill or the appropriation. The company's second quarter filing listed one customer at 18.3 percent of revenue across all three segments. The filing does not name it.

Mars

The company was founded in 2002 to make humanity a multiplanetary species, and that purpose is in the filings as something more than a mission statement. One tranche of the chief executive's 1.3 billion restricted shares vests only on a permanent human colony on Mars with at least one million inhabitants. Starship, the vehicle that now carries Starlink V3 satellites and the Artemis lander contract, was designed for Mars first; Starlink exists in large part to pay for it. The schedule has moved. The company had planned five uncrewed Starships toward Mars in the window that opens in November 2026, and in February it set that aside. In July the chief executive wrote that people would be on Mars "in roughly 5 to 7 years" with an uncrewed lander a few years sooner. No company has sent a vehicle of this size beyond Earth orbit, and the propellant transfer that any Mars flight requires has not been demonstrated. The point for a shareholder is narrower than the ambition: every capability Mars needs, a reusable heavy vehicle, orbital refueling, a high flight rate, a launch site built for it, is also the capability that the satellite, government and lunar businesses pay for today. The company is building toward Mars with other people's missions.

The balance sheet and the cash

At September 30 the company held $51.3 billion of cash and $38.6 billion of marketable securities, $89.9 billion together, against $38.8 billion of debt and finance leases, for about $51 billion of net cash. The debt is in three pieces. The $25 billion of bonds sold in June across five maturities from 2031 to 2056 carry a weighted average coupon of 5.855 percent; the order book was about $90 billion and the proceeds repaid a $20 billion bridge loan. A $5 billion revolving facility maturing in 2031 was undrawn at June 30. And $12.9 billion, a third of the total, is related party debt, most of it an April 2026 equipment lease with Valor Equity Partners, whose founder sits on the board, that the auditors treat as a financing. Moody's rates the company Baa1, S&P BBB and Fitch BBB+, each with a stable outlook, each in the lower reaches of investment grade, and Moody's cited the capital spending, the negative free cash flow and the concentrated voting control in its reasoning.

The cash flow is the arithmetic that matters. Operating cash flow for the nine months was $13.1 billion, about $9.6 billion of it in the third quarter. Capital spending was $48.1 billion, so the business consumed about $35 billion of cash this year before financing, which was covered by $85.7 billion from the IPO and $51.8 billion of debt proceeds. On the fourth quarter guidance of $35 billion to $40 billion of capital spending, and assuming operating cash flow holds near the third quarter's level, cash and securities would end the year near $60 billion to $65 billion without new borrowing. The chief financial officer told an analyst on the call that "there are a number of available pools of debt capital that we are talking to at the moment," and listed operating cash flow, debt, and partnerships along the supply chain as the sources for 2027. Scheduled debt maturities are small, $944 million for the rest of 2026 and $2.4 billion in 2027. The constraint is not the debt that exists. It is the rate at which the company wants to spend against the cash it generates.

Deferred revenue was $17.5 billion at September 30, $10.8 billion of it current, which means customers have paid for a meaningful amount of service not yet delivered.

What the price implies

Market capitalization at the October 20 close of $147.79 was about $2.01 trillion on about 13.57 billion shares. Enterprise value, subtracting the net cash, is about $1.96 trillion. The multiples below use the company's own figures and are arithmetic, not opinion.

Measure	Figure	Multiple
Revenue, last twelve months	$32.0 billion	61 times EV
Revenue, Q3 annualized	$58.3 billion	34 times EV
Adjusted EBITDA, Q3 annualized	$35.2 billion	56 times EV
Revenue, Q1 2027 guidance midpoint annualized	$140 billion	14 times EV
Annualized revenue target, December 2026	$100 billion	20 times EV
Chief executive's 2027 recurring revenue range	$500 billion to $1 trillion	2 to 4 times EV
Diluted EPS, Q3 annualized	$1.40	106 times price
Consensus 2027 EPS after the report	$2.25	66 times price

Each row is a different belief about the future. The first two describe the company as it has been. The fourth describes the company management has guided to for the quarter after next. The sixth describes a company that would be larger by revenue than any that exists; Walmart's fiscal 2026 revenue was $713 billion. The price sits between them.

For comparison, as of October 21 Nvidia trades at about 17 times trailing revenue and 26 times trailing EBITDA, Tesla at 14 times revenue and 135 times EBITDA, Palantir at 74 times revenue, Oracle at 7 times revenue and 15 times EBITDA, CoreWeave at 11 times revenue and 21 times EBITDA, and T Mobile at 3 times revenue and 8 times EBITDA. SpaceX's three segments map onto all six of those companies in some way, which is why the multiple has no clean peer.

The sell side has settled on a view. MarketBeat counts 46 analysts over the past year: 33 buy equivalents, 9 holds, 4 sells. Targets run from $75 to $800. On October 21 JPMorgan went to $500 from $240 on 20 times its 2028 earnings estimate of $24.41 a share, which is a bet that earnings grow about seventeen fold from the annualized third quarter in two years, and Goldman went to $245 from $230. The dispersion is the point. When published targets for one stock span more than ten to one, the analysts are not disagreeing about the quarter. They are disagreeing about which of the rows in the table above describes the company.

Short interest was 181 million shares at September 30, 1.3 percent of shares outstanding and about 7 percent of the float.

The supply of shares

The float is the number most price discussions skip. Of 13.57 billion shares, about 2.56 billion trade. The rest are held by insiders and pre IPO holders under a staggered lockup: about 912 million shares came free on August 6, two trading days after the first report, with further releases of roughly 320 million to 330 million on September 9, September 24 and October 9, and 328 million more on October 26. The largest tranche of the year, about 1.3 billion shares by the schedule Reuters published, becomes eligible two trading days after the third quarter report, which is this week. The 180 day expiry in early December releases about 342 million more. The founder's roughly 6.4 billion shares are locked until June 2027.

Insider selling has been one filing: the president sold 342,170 shares in September for about $52 million under a prearranged plan adopted in June. There has been no selling by the chief executive, who cannot sell until next June. The lockup schedule is public and the dates are known; what the market does on them is not.

Tesla

On the call the chief executive confirmed that both Tesla and SpaceX have formed special committees of independent directors to consider a combination "at the request of shareholders," that he is a conflicted party excluded from their discussions, that the committees "have not reached agreement on the terms of any transaction," and that calling a deal certain "would be overstating it." A Tesla approval would require a majority of all outstanding Tesla shares, not of votes cast. Tesla's proxy for its December 3 meeting at Gigafactory Texas carries no merger item. One director, Ira Ehrenpreis, sits on both boards. Tesla's $2 billion stake in xAI converted into about 19 million SpaceX shares. Barclays has written that a deal could be announced before the Tesla meeting with SpaceX as the acquirer in an all stock exchange; no terms exist.

The two companies already transact. SpaceX bought $329 million of Tesla Megapack batteries in the first half of 2026 for its data centers and held $506 million of Megapacks and $131 million of Cybertrucks on its books at year end. At the October 20 close SpaceX's market value was about $2.0 trillion and Tesla's about $1.5 trillion.

Texas

SpaceX is a Texas corporation, headquartered in a city it incorporated. Starbase became a Type C general law city in Cameron County on May 20, 2025, with a SpaceX vice president as mayor, a population of about 500, and an annexation of about 7,100 acres in process this year. The county's economic impact study counted 3,400 employees at Starbase in 2024 and 900 more added in 2025, 350 suppliers, and a $147 million regional supply chain. The county's ten year property tax abatement on the site ran its final year in 2025. Two state enterprise zone nominations, for the 700,000 square foot Gigabay and for the second launch pad complex, carry up to $3.75 million each in sales tax rebates against $506 million and $480 million of stated investment and 500 jobs apiece. A 2025 state law let the city's commissioners approve closures of State Highway 4 and Boca Chica Beach; the city's September 23 order closed both for Flight 14.

The next step is the Federal Aviation Administration's programmatic environmental review, opened October 14 with a scoping period through November 30 and hearings in Brownsville on November 17 and Port Isabel on November 19. The current license allows 25 launches and 50 landings a year. The company is asking for up to 100 launches and 200 landings in a first phase and up to 456 launches and 912 landings in a second, on weekdays. The weekly cadence the company guided to for 2027 does not fit under the current permit, and the review is the mechanism that decides whether it fits under the next one. An eight mile gas pipeline to the site is under construction for service in early 2027, and NextDecade's Rio Grande LNG terminal, about six miles away, is building four trains for operation between 2027 and 2030.

North of the launch site the footprint is industrial. McGregor, on more than 4,300 leased acres, tests every Falcon engine and stage and the Raptor engines before flight. Bastrop, east of Austin, builds Starlink terminals: 1,590 workers as of June, an $855 million expansion that targets 10 million kits a year and the consolidation of all kit production in Bastrop County, a $17.3 million state semiconductor grant for chip packaging, and county enterprise zone rebates of up to $3.75 million with no property tax abatement. The second quarter earnings call was hosted from the Bastrop plant. In Grimes County, between Houston and College Station, the Terafab chip project with Tesla covers more than 22,000 acres with a stated initial investment of $55 billion, up to $119 billion, and a full property tax abatement from 2027 to 2036.

Louisiana is the next spaceport. Announced August 25, Starbase, Louisiana on Pecan Island in Vermilion Parish is planned at $100 billion over time, with construction in 2027, first launch targeted for 2029, 3,000 direct jobs at an average of $92,600 by 2035, and an incentive package the Louisiana Illuminator priced at about $27.5 billion over 25 years, most of it property tax exemption. The chief executive said the site could eventually hold 20 or more launch towers. Texas keeps the headquarters, the engine testing, the terminal plant and the first spaceport; Louisiana gets the one designed for thousands of flights a year.

On the exchange question, the company listed on Nasdaq and on Nasdaq Texas, Nasdaq's Dallas venue, and the Governor's office marked the dual listing on IPO day. It has not listed on the Texas Stock Exchange, which began corporate trading on October 5 with Energy Transfer, Sunoco, USA Compression and Dillard's and has since added Texas Capital Bancshares, Origin Bank, CECO Environmental and International Bancshares. The largest Texas company by market value trades in Dallas through Nasdaq's exchange, not the new one.

Risks

Capital spending exceeds operating cash flow by a wide margin, $35 billion in nine months, and the fourth quarter guidance roughly doubles the quarterly rate. The company has said it will borrow. Each dollar of new debt or equity changes the per share arithmetic.
The two compute contracts that account for most of segment revenue can be terminated by the customer on 90 days' notice, one of them from January 1, 2027. Revenue visibility is shorter than the three year terms suggest.
Two customers were 38 percent of second quarter revenue between them. Neither is named.
Starship has reached orbit once. The launch cadence guided for 2027, the Artemis IV landing in early 2028, the V3 constellation and the Louisiana spaceport all depend on a vehicle that has not yet been caught, reflown or refueled in orbit, and on an FAA permit that does not yet allow the cadence.
The chief executive controls about 82 percent of the vote, is chief executive of Tesla, and is the subject of a possible merger between the two companies in which he is a conflicted party. Moody's named the governance structure in its rating.
A third of the debt is owed to a related party, and the equipment behind it depreciates fast.
The mobile business needs FCC approval of the 800 megahertz purchase, satellites that fly in 2027, and a 2028 service start, against three national carriers that have said they will not wholesale it.
Share based compensation was $2.36 billion in the quarter and the Cursor awards alone add $23.4 billion of future expense. The release shows 12.48 billion diluted shares against 2.94 billion a year ago; between the two dates the company split its stock five for one and issued shares for xAI, for Cursor and in the IPO.
About 1.3 billion locked shares become saleable this week and 6.4 billion more next June.

What we would want to see

First quarter 2027 revenue inside the $33 billion to $37 billion range, with the Super Intelligence segment's revenue reconciled against the contracted monthly fees in the prospectus and any new named customers.
Fourth quarter capital spending inside $35 billion to $40 billion and the year end cash figure, with the terms of whatever new debt is raised disclosed in the 10 K.
Operating income in each segment above zero for two consecutive quarters, with Space's internal launches still excluded.
Starship Flight 15 with both stages caught, and a published FAA license for it.
Consumer accounts above 15 million with average revenue per account holding at or above $69.
The Google and Anthropic agreements still in force on April 1, 2027, after the first open termination window.

SpaceX is the largest company headquartered in Texas and the most consequential one for the counties this publication covers, from Cameron to Bastrop to Grimes. It reported a quarter in which the profitable business grew 69 percent, the new business grew to the point where its two contracts rival the whole of what the company earned last year, and the rocket business did the work for both of them without charging. The price pays for the next two of those continuing. The filings say what has to happen for that to be true, and this piece has listed it. The rating is Buy, the Street average target is $233.50, and the two tests the company fails, cash generation and valuation, are the two we will be watching when the fourth quarter report arrives. The author holds shares and will keep reading the filings as a shareholder as well as a publisher.

This is general commentary about a publicly traded company, published for a general readership on a regular schedule. It is not individualized advice and not an offer to buy or sell anything; a rating here is the publication's opinion, not a recommendation to any particular reader. The author holds shares of SpaceX. See the disclosures page.

Sources

Space Exploration Technologies Corp. (Q3 2026 results release, Q2 2026 results release, Q3 2026 earnings call, Q2 2026 Form 10 Q, June 2026 prospectus, June and August 2026 Form 8 K filings)
U.S. Securities and Exchange Commission
Nasdaq
S&P Dow Jones Indices
Moody's
S&P Global Ratings
Fitch Ratings
Federal Aviation Administration
Federal Communications Commission
NASA
U.S. Space Force Space Systems Command
BryceTech
Office of the Governor of Texas
Texas Department of Licensing and Regulation
Cameron County
Louisiana Economic Development
Louisiana Illuminator
Texas Tribune
Elgin Courier
Reuters
Bloomberg
Wall Street Journal
CNBC
Motley Fool
Benzinga
Light Reading
Broadband Breakfast
Latham & Watkins
Air & Space Forces Magazine
NASASpaceflight
SpacePolicyOnline
Spaceflight Now
MarketBeat
StockAnalysis
MarketScreener
Yahoo Finance
24/7 Wall St.
Teslarati
Electrek
IFR