Operating Profit Arrives and the Q1 Guide Runs 58% Above the Street, but the Capex Bill Doubles With It

Results vs. Consensus

This is SpaceX's second quarter as a public company, and the first with a clean three months of public-company share count. Coverage has matured since August and the dispersion has narrowed: published revenue estimates clustered between $12.7B and $12.9B, and EPS estimates ran from $0.10 to $0.16. The Cursor acquisition closed on August 14, so the quarter carries about seven weeks of an acquired software business, which we separate below wherever the release allows.

Q3 2026 Scorecard

The pattern is the inverse of August. In Q2 every revenue line beat and capex overshot the Street by 39%, and the stock gave back a 9% rally after hours. In Q3 total and SI revenue beat by more, Connectivity beat by less, and capex landed on the number. The surprise this time sat in the forward guidance rather than the spending, and it ran in both directions at once.

Year-over-Year Comparison

Sequential Comparison

Revenue Assessment

The composition changed more in one quarter than in the previous four. SI was 51.9% of revenue, against 14.6% a year ago, and is now the largest segment. Connectivity fell to 36.0% of the mix while growing 69.5%, and Space was 12.1%. The headline growth rate of 161% is therefore mostly a statement about one business line: SI solutions and infrastructure revenue went from $359M to $7,022M in a year.

The sequential shape matters more than the annual one. Revenue rose 86.5% from Q2 while nameplate compute rose only 21%, from 1.4 GW to 1.7 GW. The difference is price and utilisation, not capacity, and it is the first hard evidence that the $30 to $50 per watt that management guessed at in August describes new deals rather than the installed base. It also means Q3 revenue is not a run-rate: data halls came online in early October, and the Q1 2027 guide says the exit rate is far above the quarter's average.

Margin Assessment

Adjusted EBITDA margin of 60.4% against 45.3% in Q2 and 38.9% a year ago is the cleanest evidence yet that compute rental at current prices is a high-margin activity. The mechanism is visible in the SI segment, where cost of revenue rose $1,569M sequentially against a $4,999M rise in revenue. The gap between adjusted EBITDA and operating income was $6,018M, made up of $3,644M of depreciation, $2,363M of share-based compensation including the Cursor awards, and $11M of acquisition and restructuring costs.

"We also expect total company gross margins in Q4 to be more consistent with the second quarter given some onetime benefits and customer mix that we highlighted in Q3." — Bret Johnsen, Chief Financial Officer

Two things will compress the reported margins from here, and neither means the economics are worse. Depreciation rose 27.9% sequentially and will accelerate as $35 to $40B of Q4 capex goes into service. And the Cursor retention awards add roughly $2B a quarter of stock compensation for about three years. Adjusted EBITDA excludes both. Investors anchored on it will keep seeing a better business than the income statement shows, which was our warning in August and remains true now that the gap is in profit rather than loss.

EPS Assessment

Diluted EPS of $0.35 beat a $0.15 consensus by $0.20, on 12,475M diluted weighted-average shares. Basic EPS was $0.37 on 11,942M. The share count is now stable enough to compare: the June-quarter weighted average of 5,864M was distorted by a mid-quarter IPO and preferred conversion, and the Q3 count includes the 391M Cursor shares for about half the quarter.

The below-the-line items deserve more scrutiny than they got. Interest income of $873M on roughly $95B of average cash and securities is a durable 3.7% yield, and it now exceeds interest expense of $675M. Other income of $1,025M is not durable on the evidence available: the nine-month figure is still a $937M expense, and the release labels $561M of the quarter's other income as related-party without saying what it is. The $390M tax benefit on $4,014M of pre-tax income is also unexplained. Our run-rate EPS for modelling purposes is closer to $0.24 than $0.35.

Segment Performance

SI: The Step-Function Repeats, and This Time the Operating Line Followed

In August we wrote that the AI segment's EBITDA flip was "100% depreciation arithmetic" because the segment still lost $1,257M at the operating line. That caveat no longer holds. SI earned $431M of operating income in Q3 after absorbing $1,217M of Cursor acquisition stock compensation, or $1,648M before it. Depreciation rose $588M sequentially and revenue rose $4,999M. The segment now now earns an operating profit after depreciation, and its adjusted EBITDA covers that depreciation twice over ($4,953M against $2,473M), which is the test that matters for a business whose assets wear out in five or six years.

"For SI Solutions and Infrastructure, revenues were $7 billion, of which $608 million came from the Cursor acquisition closed roughly mid-quarter. And even excluding Cursor, SI Solutions and Infrastructure revenues grew 192% sequentially, driven primarily by the activation of new cloud services agreements." — Bret Johnsen, Chief Financial Officer

Advertising also turned. Revenue of $538M rose 46.6% sequentially and 18.5% year on year, the first year-on-year quarterly increase since we began coverage, and management credited the ad-tech platform rolled out in Q2. Nine-month advertising revenue is still 5.7% below last year, so the turn is one quarter old.

Assessment: SI is now the reason the stock is valued where it is, and Q3 answered the question we said mattered most. It did not answer the durability question. The release does not update customer concentration (one AI customer was 19.5% of consolidated revenue in Q2), and the CEO said the company prefers short-term contracts at spot rates. That is the right strategy in a shortage and the wrong one after it. We now underwrite SI at a multiple of forward EBITDA rather than as an option, and we keep that multiple below hyperscaler levels for exactly this reason.

Connectivity: Strong Growth, Two Asterisks

Connectivity grew 69.5% and nearly doubled its operating income, and on the surface the quarter reverses two things we worried about in August: ARPU rose 5% sequentially to $69, and operating margin jumped 650bp in a quarter. Both deserve an asterisk. The margin includes a one-time benefit the CFO put at roughly three points of segment gross margin, mainly tariff refunds, and excluding it adjusted EBITDA margin expanded about two points year on year rather than 4.6. Consumer revenue rose 25.6% sequentially while subscribers rose 13.3%, and consumer revenue per average subscriber reached about $81 a month against a disclosed ARPU of $69, a wider gap than the $74 against $66 we noted in Q2.

"We strategically managed our channel distribution and prioritized delivery of kits to retail channels ahead of the holiday season, which we expect will lead to increased activations in Q4." — Bret Johnsen, Chief Financial Officer

That sentence is the explanation for the gap. Kits shipped into retail are recognised as consumer revenue before they become subscribers, so Q3 borrowed some hardware revenue from Q4 activations. It is a timing effect, not a quality problem, but it means Q3's consumer revenue per subscriber should not be extrapolated. The underlying growth is still excellent: 1.6 million consumer net adds, 174 markets, and enterprise and government revenue up 91.4% on Starshield and aviation.

"Starlink is now contracted for roughly 12,000 planes, almost 1/4 of the available aircraft, and we're expanding quickly with both new and existing customers." — Gwynne Shotwell, President and Chief Operating Officer

Assessment: the segment that funded everything in Q2 now ranks second to SI on adjusted EBITDA, though it is still the largest earner at the operating line, and the change is a statement about SI rather than a weakness here. Twelve thousand contracted aircraft against 3,800 equipped at quarter end is the most visible multi-year backlog in the company outside government launch. We would model consumer revenue per subscriber near Q2's level rather than Q3's, and treat the segment's underlying operating margin as low-40s rather than 45%.

Space: Starship Reaches Orbit, Falcon Cadence Keeps Falling

Space revenue rose 83.7% sequentially to $1,767M on a richer customer manifest: a Falcon Heavy launch of NASA's Roman Space Telescope, six national security missions, and launch services revenue of about $94M per customer launch against $65M in Q2. The segment came within $7M of operating breakeven. The operating statistics went the other way. Total launches fell 20.5% year on year to 35, mass to orbit fell 9.3%, and nine-month launches are down 11.7% while nine-month Space revenue is up only 2.2%.

"During the final week of the quarter, we flew Starship Flight 14 to orbit, as Elon talked about, followed just after on October 1 by a triple header." — Gwynne Shotwell, President and Chief Operating Officer

Flight 14 was the quarter's real event: Starship's first orbital flight, carrying 26 operational V3 Starlink satellites. The release puts the downlink capacity added at roughly ten Falcon 9 Starlink launches, and the CEO said the vehicle was only about half loaded. The commercial pipeline also grew, with three more crewed Dragon missions, 18 national security launches, three Starfall missions and a lunar-surface delivery contract on Starship.

Assessment: Space is a development programme with a launch business inside it, and Q3 was a good quarter for both halves. But the Falcon cadence decline is now two quarters old and it is not a mix story: internal Starlink launches fell 20.7% year on year in a quarter when Starlink added 1.6 million subscribers. The bridge is Starship, and Q3 proved it can carry a commercial payload to orbit. The cost of not yet having a reusable upper stage is visible in the cadence statistics, which is why the Flight 15 catch attempt matters more to Connectivity than to Space.

Key Topics & Management Commentary

Overall Management Tone: expansive beyond anything on the Q2 call, and also more accountable. The forward claims grew by an order of magnitude, but for the first time the company gave a quarterly revenue range, a capex range and a gross-margin direction that can be graded. Analyst Q&A was brief and deferential, and the only pressure on funding came in one two-part question; the answers on capital sources were framework-level.

1. The Operating Line Turns Positive, With a $1.2B Charge Inside It

Our August upgrade rested on an operating loss of $143M that had come within 1.8% of margin of breakeven. Q3 cleared it by a wide margin: operating income of $2,791M on revenue of $14,575M, a 19.1% margin, while absorbing $1,217M of stock compensation tied to the Cursor acquisition. That charge sits inside SI operating expenses, which helps explain why SI SG&A rose 138% sequentially.

"The combination of our launch leadership, recurring Starlink revenue and the significant expansion of our SI infrastructure business have combined to drive meaningful operating leverage in the business." — Bret Johnsen, Chief Financial Officer

Total costs and expenses rose 101.5% year on year while revenue rose 161.4%. Every segment improved its operating line sequentially: Space by $535M, Connectivity by $711M and SI by $1,688M.

Assessment: the inflection we upgraded on has happened, and it happened faster than our model assumed. Our 3Q26 estimate was a 7.0% operating margin; the reported margin was 19.1% after an acquisition charge we had not modelled. This removes the "can it ever earn money" question from the bear case. What remains is whether profits can keep pace with the capital the company plans to spend.

2. Below the Line: $1.4B of Net Income We Do Not Capitalise

The call's opening read-out led with net income, which is the less demanding number this quarter, just as management led with adjusted EBITDA when the operating line was negative.

"revenues of $14.6 billion, up 161% from $5.6 billion; net income of $4.4 billion, an improvement of $5.5 billion from a net loss of $1.1 billion and adjusted EBITDA of $8.8 billion, up 306% from $2.2 billion." — Andrea Williams, Head of Investor Relations

The bridge from $2,791M of operating income to $4,404M of net income is $198M of net interest income, $1,025M of other income and a $390M tax benefit. Of the other income, $561M was with a related party, and the cash-flow statement shows $532M of unrealised gains on investments in unconsolidated affiliates for the nine months. Neither the release nor the call explained the related-party item. The tax benefit on a quarter with $4,014M of pre-tax income is equally unexplained, and nine-month deferred taxes swung to a $455M benefit.

Assessment: roughly $1.4B of the $4.4B comes from other income and a tax benefit we do not expect to recur, about a third. We do not capitalise it. The operating story is strong enough on its own, and leading with net income in a quarter like this invites a question management did not need to create.

3. The First Formal Guide: Q1 2027 Revenue of $33 to $37 Billion

SpaceX does not guide quarters. It did on this call, and it skipped the quarter in between.

"Given the increased visibility into our compute contract pipeline over the next 6 months, we have the opportunity to provide additional look-forward guidance on the acceleration we see in our business in Q1 2027. We now see total company revenues in the range of $33 billion to $37 billion for Q1 of 2027." — Bret Johnsen, Chief Financial Officer

The midpoint of $35B is 2.4 times Q3 revenue and 4.5 times Q2. It is 58% above the roughly $22.2B the Street carried for that quarter before the print, and 90% above our own model. Annualised, the range is $132B to $148B, against FY2027 consensus near $110B. On the monetisation arithmetic above, it is consistent with about $28 to $38 per watt on 3.0 to 3.5 GW of average capacity, so the guide is a statement about price as much as volume.

Assessment: this is the most consequential number in the report, and it is the CFO's number rather than the CEO's, which matters for how much weight it can bear. The guide is built on a six-month contract pipeline, so it is closer to contracted than aspirational. The absence of a Q4 revenue range alongside it is notable, and we discuss it below. We model the low end.

4. Pricing: New Deployments Above $50 per Watt

In August the CEO's $30 to $50 per watt was explicitly "a guess." In October the CFO put a market price on it.

"What we're seeing in the market is extraordinary. At current pricing, new compute deployments are generating a payback of less than a year. Pricing is currently tracking slightly above the high end of the $30 to $50 per watt range we were discussing just last quarter, which further supports the attractiveness of these terrestrial SI compute investments." — Bret Johnsen, Chief Financial Officer

The CEO added that the company's short-term contracts are renewing at higher dollars per watt, and that older accelerators are renting for more than they did when new. If a watt of new capacity earns $50 a year and costs roughly $40 to build, the sub-one-year payback claim requires margins and pricing near the top of what we can observe. The nine-month evidence is more modest and still good: SI capex of $40.0B has lifted annualised SI adjusted EBITDA by $22.2B since Q1, a payback of about 1.8 years at the EBITDA line, with some of that capex in data halls that only came online in October.

Assessment: the payback claim moved from unverifiable to directionally supported in one quarter. We still apply a haircut, because spot pricing in a shortage is the peak of the cycle by definition, and because the evidence is one quarter long.

5. Capex Doubles Again: $35 to $40 Billion in Q4

Q3 capex of $19,593M was 6.7% above the "very similar" level the CFO guided in August, and on consensus. The forward guide was not.

"Because of the additional capacity coming online this quarter, we expect fourth quarter capital expenditures to now be in the range of $35 billion to $40 billion for this quarter." — Bret Johnsen, Chief Financial Officer

That takes FY2026 capex to roughly $83B to $88B, against the ~$65B implied in August and against nine-month operating cash flow of $13,097M. Nine-month free cash flow was negative $34,972M. Cash and marketable securities fell from about $100.0B at June 30 to $89,855M at September 30, and debt and finance leases were $38,844M, leaving net cash of $51,011M. A Q4 at the guided level will consume roughly half of that after operating cash flow, before any new financing. Customer prepayments are already part of the funding: deferred revenue rose $5,404M in nine months to $17,520M.

Assessment: the ratio we said to audit in August, SI capex against SI adjusted EBITDA, fell from 13.8 to 3.3. The spending is converting to earnings faster than we assumed. But the absolute requirement is now large enough that 2027 is a debt-market year, and the CFO's answer on funding named categories rather than amounts. We address it in the Q&A section.

6. Compute Targets: 2.5 to 3 GW This Year, 13 GW Next

Nameplate compute reached 1.7 GW at September 30, with "nearly 2.2 gigawatts" guided for the end of October and 2.5 to 3 GW by year end, against the "over 2 GW" promised in August. The 2027 target rose more.

"At this point, I would say … I'm fairly confident of being able to bring an incremental 10 gigawatts of compute online in 2027. That would be in addition to the roughly 3 gigawatts, maybe 2.5 to 3 gigawatts that we'll have online by the end of this year." — Elon Musk, Chief Executive Officer

In August the CEO said "closer to 10 than 5" gigawatts by the end of 2027. The new figure is about 13, with a plan for 20 GW of incremental power and cooling and a "non-zero possibility" of more than 20 GW of compute. The prepared remarks put the deployment rate at roughly 200,000 GB300s a month; one answer in Q&A said 20,000, and only the larger figure is consistent with adding about a gigawatt in a quarter. The proposed edge is vertical: SpaceX is casting turbine blades and vanes in-house because suppliers quote 12 to 18 months for them.

Assessment: the near-term targets are credible: the company added 0.7 GW in two quarters and had more data halls online in early October. The 2027 figure is a different kind of statement. At the roughly $40B per gigawatt the company spent in the first half, 10 GW is about $400B of capital in one year. We model 5.5 to 6.5 GW at the end of 2027, about half the stated target, because GPU supply and funding bind before power does.

7. Half the Fleet Stays Home: The 50/50 Split and Spot-Rate Contracts

In August the CEO said Grok training would use "maybe 10%" of compute over time, which we read as 90% of capacity available to rent. This quarter the framing changed. The goal is now to split compute roughly evenly between internal use and rental, with internal use including inference for Grok Bot and Cursor, not just training. The CEO also said the company prefers to rent at spot rates rather than sign long-term deals.

"We are also, frankly, hounded relentlessly by every hyperscaler and by dozens of smaller companies to rent more of our compute. So we could rent out our compute several times over if we said yes to everyone." — Elon Musk, Chief Executive Officer

The two disclosures are not contradictory: training will still be a small share, and inference for the company's own products is a revenue use of compute. But the economics differ. Rented capacity earns a market price the CFO can quote. Internal inference earns whatever Grok subscriptions, Grok Bot and Cursor can charge, and those businesses are months old.

Assessment: the shift raises the share of SI value that depends on SpaceX's own software winning customers, and lowers the share that depends on a counterparty's credit. Both the Q1 guide and our valuation lean on rental pricing. The internal half is the part to watch, and the release gives one early datapoint below.

8. Cursor: $608M of Revenue, $23.4B of Retention Stock, and a $36.8B Goodwill Jump

Cursor closed on August 14 for 391.04 million Class A shares at an implied $60.0B equity value. In August we estimated the deal would cost about 522 million shares at the then-depressed price; the stock's recovery before closing cut the issuance by a quarter. The accounting is more expensive than the share count suggests.

"Double-clicking on the share-based compensation expenses, of the $60 billion equity value for Cursor when we purchased it, $23 billion was attributed towards retaining key employees and replacing outstanding equity awards, including revesting agreements for key executives. These equity awards will be recognized as stock-based compensation over roughly the next 3-year vesting period." — Bret Johnsen, Chief Financial Officer

The release puts GAAP purchase consideration at $42.5B and post-acquisition service awards at $23.4B. Goodwill rose from $11,645M at June 30 to $48,423M, almost entirely on Cursor, and is now 19.1% of total assets; intangibles rose $4,806M in nine months. Straight-line, the retention awards cost about $1.95B a quarter, and Q3 booked $1,217M in about seven weeks, a faster pace. Cursor contributed $608M of revenue, about $1.2B on a full-quarter basis.

"Grok bot, our Agentic Assistant, launched in August, and we already had paid weekly users of 800,000 earlier this month, more than doubling over the last month." — Gwynne Shotwell, President and Chief Operating Officer

Assessment: the goodwill risk we retired in August has returned in a different form. The xAI combination added no goodwill; the quarter's acquisitions added about $36.8B, nearly all of it for Cursor, a business with roughly $4.6B of annualised revenue. Cursor and Grok Bot are the products that must monetise the internal half of the compute fleet, so their growth now matters to the SI valuation directly. Grok Enterprise customers above $1M of run-rate revenue rose 50% sequentially on a pro forma basis, a useful but unquantified datapoint.

9. Starlink V3: "A Few Hundred" Satellites Instead of a Thousand

In August management said roughly 1,000 V3 satellites were needed before users would notice the capacity, around the second quarter of 2027. This quarter the threshold fell to "a few hundred" by the middle of 2027. Flight 14 carried 26.

"Through the end of the year and throughout 2027, we expect to deploy V3 broadband satellites in meaningful volume with each Starship launch ramping towards a payload that delivers over 20x the downlink capacity of the current generation of Falcon 9 launches today." — Bret Johnsen, Chief Financial Officer

The CEO's longer claim is 100 times the bandwidth of today's constellation within three years, from about 100,000 satellites with ten times the capacity each. Starlink Mobile reached 22 million monthly users, up 37% sequentially, the FCC approved 15,000 Mobile V2 satellites, and the company agreed to buy up to 14 MHz of paired 800 MHz spectrum, expected to close in the first half of 2027.

Assessment: the lower threshold is good news if it reflects V3 capability rather than a redefinition of "feel the impact." The Mobile service date moved from the end of 2027 in August to "early 2028" on this call, a small slip we note for the record. Connectivity's 2027 growth now depends on Starship cadence, which ties the two pillars together more tightly than our model does.

10. Starship: Orbit Achieved, the Catch Moves to Flight 15

In August management scheduled a ship catch for the end of that month, subject to regulatory approval. The next flight came in the last week of September, went to orbit rather than to the tower, and the catch of both stages is now the objective for Flight 15.

"For Flight 15, which we expect to happen next month, we're going to attempt to catch both the booster and the ship." — Elon Musk, Chief Executive Officer

The rest of the roadmap is dated: refly of a second stage by early 2027, an on-orbit refuelling attempt early in 2027, weekly flights at some point in the first half of 2027, and twice-weekly by the end of it. The company also announced Starbase Louisiana, with the release targeting first launch in 2029 and the CEO aspiring to 2028.

"We'll attempt an on-orbit refueling early next year, and we'll begin flying Mobile V2 and Starmind satellites in 2027." — Gwynne Shotwell, President and Chief Operating Officer

Assessment: the August catch date slipped, and the orbital flight with a real payload is a bigger milestone than the catch would have been. In-orbit refuelling, which our S-1 initiation named as the binding constraint on lunar work, now has a date for a first attempt. We raise the Starship option to $75B to $100B, from $75B in our S-1 table, and would raise it again on a successful dual catch.

11. Tesla: Special Committees, and a Founder Who Cannot Be in the Room

The call confirmed that both companies have formed special committees of independent directors to consider a combination at shareholders' request, and that the CEO is recused as a conflicted party. He said he thinks a merger "would be a good idea" and listed the overlaps: super-intelligent software, chip design, Terafab, robotics and the Starmind constellation.

"I can just say at this point that the special committees have not reached agreement on the terms of any transaction. And obviously, there's no guarantee that they will." — Elon Musk, Chief Executive Officer

He also noted that Tesla would need a majority of all outstanding shares, not just of votes cast, which he called "quite a high hurdle."

Assessment: a combination of two companies controlled by the same founder is the largest related-party transaction either could undertake, and SPCX holders have no visibility into the exchange ratio, the governance terms or which side would pay a premium. The committee structure is correct. We assign no value to the deal either way, and we raise the governance risk because the question is now formally open.

Guidance & Outlook

SpaceX gave its first quantitative revenue guidance as a public company, for Q1 2027 rather than Q4 2026, and restated most of its August commitments at higher levels.

The CFO's commitments are specific and gradeable. The CEO's are not, and the gap between them widened on this call.

"So I mean in a nutshell, what this implies for SpaceX annual recurring revenue by the end of '27 is somewhere between $500 billion and $1 trillion." — Elon Musk, Chief Executive Officer

Against the CFO's Q1 range, which annualises at $132B to $148B, the CEO's figure requires revenue to grow another 3.4 to 7.6 times during 2027. At $30 to $50 per watt, $500B of annual revenue needs 10 to 17 GW of rented compute, against roughly 6.5 GW of rented capacity if half of a 13 GW fleet stays in-house.

Implied ramp: the $100B December target requires about $8.33B of revenue in December, 1.7 times Q3's monthly average of $4.86B. The Q1 midpoint requires $11.67B a month. Both are consistent with the October capacity additions and the October start of the July cloud contracts, and the Q1 guide makes the December target look conservative rather than ambitious.

Street at: Q1 2027 consensus near $22.2B on six estimates, FY2027 near $110B, FY2026 near $44B. FY2026 consensus implies Q4 revenue of about $17B, which looks low against a $100B December run-rate. Our own model carried 1Q27 at $18.4B and FY2027 at $91.0B and is now stale by a wide margin.

Guidance style: improving. A quarterly revenue range two quarters out, a capex range and a gross-margin direction are more than this company has ever given. The habit of stating multi-year destinations without a bridge persists, and the CEO's 2027 run-rate is the largest example yet. Grade the CFO's numbers; read the CEO's as conviction.

Analyst Q&A Highlights

The Terafab Timeline and the EUV Question

The first question asked whether SpaceX has ordered EUV lithography tools and how a first-principles approach would change chip fabrication. The answer gave internal targets with explicit error bars and did not confirm an order directly, although it thanked ASML for accelerating equipment deliveries.

Q: "So, first, can you confirm if SpaceX has placed orders for EUV machines? If so, when do you expect delivery?"
— Adam Jonas, Morgan Stanley

A: "But it's what we're aiming for internally is to make our first sort of test chip functional in approximately May or June of next year. That's just a kind of a SRAM chip something that basically test the line. That's at the research fab in Austin. … Now these are very aggressive dates, and we're very appreciative of ASML and all of our other suppliers in accelerating equipment deliveries in order to achieve this outcome."
— Elon Musk, Chief Executive Officer

Assessment: a test chip in mid-2027 and volume production "around Q3 of '28" at Gibbons Creek are aspirations, and the CEO labelled them as such. Terafab has no revenue or cost line in the release, and it is the one programme on the call with no disclosed capex. It belongs in the long-dated option bucket, not in any 2027 model.

What Makes 13 Gigawatts Credible

Asked what gives management comfort in 13 GW by the end of 2027 given power and cooling constraints, the CEO pointed to the deployment rate already visible and then to a supply-chain edge: in-house casting of turbine blades and vanes, the component suppliers deliver last.

Q: "Elon, can you talk more about just how you're working through those constraints in power cooling and infra and just what gets you comfortable with the 13 gigawatts ending '27?"
— Douglas Anmuth, JPMorgan

A: "So we're casting the blades and vanes for the turbines. … And when we talk to the turbine suppliers, they tell us that the blades and vanes are 12 to 18 months behind the rest of the turbine, meaning they can deliver everything except the blades and vanes 12 to 18 months before they can deliver the blades vein. … So that means that we can bring power online 12 to 18 months faster than anyone else."
— Elon Musk, Chief Executive Officer

Assessment: a specific, checkable claim about a real bottleneck, and the most persuasive answer on the call. It addresses power. It does not address accelerators, which the company buys exclusively from one supplier, or capital, which the second half of the question raised.

Funding the Build

The same question's second part asked how the company thinks about funding that capacity. The answer listed three sources in order and quantified none of them.

Q: "And then, Bret, if you could also help us understand how you're thinking about the funding requirements just as you look out to that kind of capacity?"
— Douglas Anmuth, JPMorgan

A: "So you have your operational cash flows, first and foremost. You have multiple pools of capital available in the debt market, especially because of our investment-grade status and strong balance sheet. And then you also have the opportunity from what we've seen to partner with, I would say, the whole value chain related to, on the vendor and customer side."
— Bret Johnsen, Chief Financial Officer

Assessment: the right three sources and the right order. Operating cash flow was $9.6B in Q3, about half of Q3 capex and a quarter of the Q4 guide. "Partner with the whole value chain" means vendor financing and customer prepayments, both already visible in the balance sheet. The question was about a 13 GW plan that could need several hundred billion dollars, and no figure for 2027 capex or new debt was offered. That is the gap we would most like closed on the Q4 call.

A Combination With Tesla

The only governance question asked for the pros and cons of a Tesla combination and any roadblocks. The CEO described the special-committee process, his recusal and the Tesla vote threshold, and said he favours a deal.

Q: "But if we take a step back, how do you think about the pros and cons of a potential combination? And as part of that framework, are there any roadblocks that you would need to address before embarking on it?"
— Edison Yu, Deutsche Bank

A: "So both SpaceX and Tesla have formed special committees to discuss this idea at the request of shareholders. And these special committees are consist of independent directors. And because I'm a conflicted party, I can't actually be in the discussions of the special committees. … And particularly on the Tesla side, we will need a 50% vote of all shares, not just those who vote, but of all shareholders."
— Elon Musk, Chief Executive Officer

Assessment: the answer was careful and procedurally correct, and the roadblock it named sits on the Tesla side. For SPCX holders the risk is on the other side: a controlled company negotiating with its controller's other controlled company. The cons were not addressed.

Satellite and Terrestrial Roles in Starlink Mobile

A question on how the satellite and terrestrial layers would split traffic, and whether the focus is retail wireless or physical AI, produced a spectrum answer and a promise of a separate presentation.

Q: "So what roles do you expect the satellite and terrestrial parts of the network will play? And how should we think about the traffic mix between them and delivering a 5G-like experience in the U.S.?"
— Kutgun Maral, Evercore

A: "As far as the mix goes, I think we expect that the S-band spectrum will provide the majority. And then obviously, you need the low-band spectrum, the 800 megahertz spectrum to basically penetrate buildings."
— Gwynne Shotwell, President and Chief Operating Officer

Assessment: the answer explained the role of the new 800 MHz spectrum clearly and left the traffic split, the terrestrial build and its cost unaddressed. The CEO's follow-up, "a clear path to several hundred billion dollars a year in connectivity revenue," was explicitly deferred to a future presentation. We would rather have the presentation before the claim.

Internal Use Versus Rental, and Contract Length

The most model-relevant question asked how management balances compute between cloud customers and its own products, and how that mix monetises over the next 12 to 18 months. The answer replaced the August "maybe 10%" training figure with a roughly even split, and set out a deliberate preference for short contracts.

Q: "When you think about a framework for applying that compute to the external opportunity that presents itself from cloud services relative to the internal needs of that compute for scaling Cursor and Grok bot, how has your view evolved about striking the right balance in that mix?"
— Eric Sheridan, Goldman Sachs

A: "Our goal is to kind of split it 50-50, more or less, where half of the compute is used internally and half of it is rented to others. But that number could shift 60-40 one way or the other. … And obviously, our approach is to rent at the, effectively the spot market rate as opposed to sign these long-term deals."
— Elon Musk, Chief Executive Officer

Assessment: the most useful answer for modelling, and it cuts both ways. Spot pricing in a shortage is accretive now and is exactly the exposure our August debate warned about if supply catches up. The 50/50 split means half the fleet's value depends on products that launched this year. We keep the SI multiple below hyperscaler levels on both counts.

When V3 Capacity Reaches Users

A retail investor question asked how many V3 satellites are needed before users notice. The answer lowered the threshold from August's figure and attached a date.

Q: "So when will capacity gains from V3 satellites actually show up as meaningfully higher speeds and lower latency for users? How many V3 sats are needed before the network feels noticeably better in high-demand areas?"
— Investor question from the company's Q&A platform, read by Andrea Williams, Head of Investor Relations

A: "I think we'll need a few hundred V3 satellites for people to really feel the impact because right now, with very few satellites, you'll get an impact and then 10 minutes later, it won't be there. So a few hundred satellites, and I think we can achieve that by mid next year."
— Gwynne Shotwell, President and Chief Operating Officer

Assessment: a few hundred satellites by mid-2027 at 26 per half-loaded flight is roughly ten Starship flights in about eight months, which fits the stated cadence if Flight 15 succeeds. This is the dated commitment that links Starship execution to Connectivity revenue in 2027.

What They're NOT Saying

1. Customer concentration. The Q2 10-Q disclosed that two customers were 37.8% of consolidated revenue, one of them AI-only at 19.5%. The Q3 release and call did not update either figure, and the Q3 10-Q had not been filed when we wrote, although last quarter it arrived the same day as the release. With SI now 51.9% of revenue, concentration is likely higher, and it is the first number we will look for.
2. The $561M of related-party other income. The income statement labels it as related-party and nothing else in the release or on the call explains it. With related-party interest expense of $351M in the same quarter, the company now has material related-party flows on both sides of the income statement.
3. The $390M tax benefit. A tax benefit on $4,014M of pre-tax income needs an explanation, most often a valuation-allowance release linked to acquisition accounting. None was given.
4. A Q4 revenue range. Management guided Q1 2027 but not the quarter in progress, other than restating the December run-rate target. Q4 is the quarter that will carry the gross-margin reversion, a full quarter of Cursor retention stock and the October contract ramp, which makes it the harder quarter to guide.
5. 2027 capex and financing. The CEO set a 13 GW target for the end of 2027 and the CFO declined to size the capital behind it. On the company's own first-half cost of about $40B per gigawatt, the stated plan implies several hundred billion dollars of 2027 spending.
6. Falcon cadence. The prepared remarks opened with launch leadership and "the vast majority of payload mass to orbit." Total launches fell 20.5% year on year and mass to orbit fell 9.3%, the second straight quarter of decline in both, and neither comparison was mentioned.
7. The size of the Q3 one-offs. The CFO quantified the Connectivity tariff refund at "roughly 3%" of segment margin. The Q4 gross-margin guide cites "some onetime benefits and customer mix" at the company level, and the other one-time items were not identified or sized.
8. Retail channel inventory. Management said it prioritised kit deliveries to retail ahead of the holidays and did not say how much Q3 consumer revenue that represented. The widening gap between disclosed ARPU and revenue per subscriber suggests it was material.
9. The October 22 unlock. Up to 1.3 billion shares become sellable on the second full trading day after this release, followed by 328.4 million on October 24 and up to 797.6 million on December 8. The call did not mention the schedule.

Market Reaction

• Pre-print setup: the shares closed at $147.79 on October 20, down 0.11% on the session on 89.4 million shares, slightly below the prior 30-session average of about 93.0 million. The stock was down 3.2% over the trailing thirty days and 14.0% below its October 6 closing high of $171.92, but up 36.5% from the post-Q2 closing low of $108.27 on August 5 and 9.5% above the $135 IPO price. It remained 26.8% below the June 16 closing peak of $201.80. The S&P 500 was up 12.7% year to date. Options implied a move of about 8%.
• After-hours move: shares rose about 2% on the release and 3.5% to 4% over the following ninety minutes. Other quotes ranged from $155.40 (+5.15%) early in the session to $157.59 (+6.63%). The move was attributed to the revenue and earnings beat and to capex landing on consensus rather than above it.
• Supply calendar: up to 1.3 billion shares, 9.6% of the 13,579 million outstanding, are released from lock-up on Thursday, October 22. With the October 24 and December 8 releases, up to 2.43 billion shares, 17.9% of the count, become sellable within seven weeks.

The after-hours move was small for the size of the news. A Q1 guide 58% above consensus would normally re-rate a stock by more than the options market implied, and this one moved by less. Three things capped it. The stock had already run 59% from the August low to the October 6 high, so a good print was partly in the price. The Q4 capex guide doubled the spending the market had modelled, and the CFO's funding answer gave no numbers. And the largest unlock yet, bigger than August's 911.5 million shares, arrives in two sessions, so every holder released on Thursday now has a higher exit than they had at Tuesday's close.

The comparison with August is instructive. Then, the stock rallied 9.4% into the print and sold off on capex. Now, it drifted 14% lower from its October high into the print and rose modestly on a far larger beat. Positioning was lighter, and the reaction was correspondingly more about fundamentals. We expect the next two sessions to be governed by the unlock rather than by the quarter.

Street Perspective

Debate: Is $35 to $40 billion a quarter of capex a return or a risk?

Bull view: new compute pays back in under a year at current pricing, SI adjusted EBITDA already runs at about $20B a year, and the Q1 guide shows Q3 and Q4 capex converting to revenue within two quarters. On that arithmetic the company should spend faster, and an investment-grade balance sheet with $51B of net cash makes it possible.

Bear view: FY2026 capex near $85B against nine-month operating cash flow of $13.1B, and a 13 GW plan for 2027 that could need several hundred billion dollars, is a debt-financed bet on spot pricing staying at peak levels. Accelerators depreciate over five or six years, and the pricing that justifies them is set month to month.

Our take: the bulls won the Q3 argument on evidence. SI capex fell to 3.3 times SI adjusted EBITDA from 13.8, and nine-month SI capex is paying back in under two years at the EBITDA line. The bear case has moved from returns to scale: we underwrite about half the 2027 target, and we would treat any 2027 capex guide above $150B without matching contract disclosure as a negative.

Debate: How durable is spot-priced compute revenue?

Bull view: the company is turning customers away, short-term contracts are renewing at higher prices, and older accelerators rent for more than they did new. In a structural shortage, short terms are an option to reprice upward, and a $63B backlog with $42B due within a year shows customers are committing anyway.

Bear view: the same short terms let revenue reset downward within a quarter if supply catches up, one customer was a fifth of revenue in Q2, and half the fleet will now serve internal products that launched this year. Q3 is the first quarter of spot pricing at scale, not a track record.

Our take: the bulls are right about the next year and the bears are right about the shape of the risk. We value SI at 15 to 19 times forward EBITDA, below where profitable hyperscale infrastructure trades, because the revenue can fall as fast as it rose. A quarterly disclosure of rented versus internal capacity would do more for the multiple than another capacity target.

Debate: Would a Tesla combination create value for SPCX holders?

Bull view: the two companies already share a founder, suppliers and ambitions in AI software, chips, robotics and orbital compute. A combination would pool two large capital bases behind one compute build and remove the related-party friction between them.

Bear view: both companies are controlled by the same conflicted party, the exchange ratio is the whole economic question, and SPCX holders could end up paying in shares for a business with different growth and margin characteristics. A deal would also absorb management attention during the largest capital programme in the company's history.

Our take: no terms exist and the Tesla vote threshold is high, so we assign the deal no value. We do treat its existence as a governance cost. The special-committee structure is the right one, and we would judge any proposal on whether SPCX's independent committee obtains a fairness opinion and a majority-of-minority condition.

Model Update Needed

Our model was rolled forward to Q2 actuals on September 26 and recalibrated to the consensus band on October 9. It built the second half from disclosed contracts only and treated the $100B December run-rate as unmodelled. Q3 beat that build by 28.5%, and the Q1 guide exceeds our 1Q27 estimate by 79% to 101%.

Valuation impact. At $147.79 the equity is worth $2.01 trillion on 13,579 million shares, and enterprise value is about $1.96 trillion after $38.8B of debt and $89.9B of cash and securities. That is 33.5 times annualised Q3 revenue and 55.5 times annualised adjusted EBITDA, against 46.5 and 102.7 times at our August upgrade. On the Q1 2027 guide midpoint annualised, it is 14.0 times revenue. The multiple compressed again even though the price rose 29% from the $114.93 of our August report, because revenue rose faster.

We re-derive fair value from the sum of the parts we published at the S-1, with each part moved to what this quarter showed:

We raise our fair-value range to $1.96 to $2.59 trillion, or about $144 to $191 per share, from $1.65 to $1.95 trillion ($125 to $148 on 13,176 million shares). The per-share midpoint rises 23%, to about $168. The increase comes almost entirely from SI, where the Q1 guide turns a contracted-revenue estimate into a forward earnings base. It is smaller than the change in near-term revenue because the capital needed to earn it doubled. Against the close, the range implies roughly 2% downside to 29% upside, with 13% to the midpoint. Against the $157.59 after-hours quote, the midpoint is about 6% higher. That is a narrower margin than at our August upgrade, when the range began 9% above the price.

Thesis Scorecard Post-Earnings

Scored against the standing thesis carried since our August 5 upgrade, as corrected on September 26.

Overall: thesis strengthened on operations and weakened on price and capital. All four bull pillars held, and the most important one, AI monetisation, answered the question we set in August in a single quarter. Against that, the price has re-rated into our raised range, the capital programme doubled, and the governance risk acquired a specific shape in the Tesla committees. The business is better than we thought in August; the stock is no longer cheap for it.

Action: hold existing positions and add below $145, the low end of our range, which the October 22 unlock may well provide. We would not chase the after-hours move, and we would revisit the rating if the Q4 call sizes 2027 capex above $150B without matching contract disclosure.