Research Finder
Find by Keyword
Which SpaceX Did Investors Buy: Launch, Broadband, or Compute?
Revenue and EPS beat, AI segment EBITDA turned positive, and the capital allocation now points toward compute rather than launch. Connectivity carries the revenue, AI carries the capital, and the August 6 unlock tests whether the story survives new supply.
08/06/2026
By the Numbers
- Revenue: $7.81B, +92% YoY from $4.07B, +66% sequentially from $4.69B; beat LSEG consensus of $6.93B (Bloomberg consensus was $6.81B)
- Loss per share: $(0.09) basic and diluted, versus $(0.26) expected (LSEG)
- Net loss: $541M, a $467M improvement from $1.01B in Q2 2025
- Adjusted EBITDA: $3.54B, +191% YoY, against roughly $2.0B consensus (Bloomberg, via Yahoo Finance)
- Connectivity: revenue $4.29B, +66% YoY and +32% sequentially; income from operations $1.66B, +79% YoY; segment adjusted EBITDA $2.60B, +64% YoY
- Connectivity mix: consumer $2.49B (+44% YoY), enterprise and government $1.81B (+108% YoY)
- AI: revenue $2.56B, +247% YoY and +213% sequentially; operating loss narrowed to $1.26B; segment adjusted EBITDA positive at $1.15B, versus $(276)M a year ago
- Space: revenue $962M, +29% YoY and +55% sequentially; loss from operations $542M; segment adjusted EBITDA $(205)M
- Starlink: 12.0M subscribers, doubled YoY and up 1.7M sequentially; ARPU $66 per month, flat sequentially, down from $85 a year earlier
- Capex: $18.37B total versus a $13.22B average analyst estimate (FactSet, via CNBC); AI $15.83B, Connectivity $1.37B, Space $1.17B
- Nameplate compute: 1.4 GW, up from 1.0 GW in Q1 2026 and 0.4 GW in Q2 2025; management targets over 2 GW exiting 2026
- Contracted sales: $14.1B of Cloud Services Agreements signed, contributing $1.6B of incremental AI infrastructure revenue in the quarter
- Balance sheet: $100B in cash, equivalents and marketable securities; $47.5B backlog; debt and finance leases of $36.8B, up from roughly $22B three months earlier
- Capital markets: $85.7B net IPO proceeds (June 15 close), $25B inaugural investment grade issuance across five tranches at a 5.855% weighted average rate
- Forward markers (management, not formal guidance): $100B annualized revenue run rate exiting December including Cursor; $6.7B of additional cloud services revenue contracted in the first weeks of Q3; internal $1T revenue target pulled forward to 2030 from 2031
The News
SpaceX reported second quarter 2026 revenue of $7.81 billion, up 92% year over year, with a net loss of $541 million and a loss of $0.09 per share, its first results as a public company following the June listing. Revenue and loss per share both cleared consensus, and adjusted EBITDA of $3.54 billion ran well ahead of what the sell side had modeled. Total capital expenditures of $18.37 billion exceeded estimates and sent shares lower in extended trading. The signal worth isolating is where the capital went: the AI segment consumed $15.83 billion of that spending and turned segment adjusted EBITDA positive for the first time, while the Space segment stayed in loss. Full release: SpaceX Reports Second Quarter 2026 Results.
Analyst Take
We have followed SpaceX for years, through the reuse debates, the Starlink constellation build, and now the compute pivot. After all that, this is the first quarter in which the company has had to argue its case in public filings rather than in private rounds. The print itself landed well. Revenue and loss per share both cleared the bar, and the AI segment crossed into positive adjusted EBITDA earlier than most models implied. It arrived into hostile positioning, with press reports before the release describing short interest near a third of the public float and a massive lockup expiration two days out. Yet the number that reframes the company sits on the capital line, not the revenue line. There, AI absorbed the overwhelming majority of the quarter's spending and the launch business, the thing most investors still name first, took a single digit share. Our read is that the more underpriced risk sits inside Connectivity, where subscriber growth appears to be purchased at a declining average revenue per user.
What the Numbers Mean
The transcript is more revealing than the tables. Bret Johnsen described compute capital as behaving almost like cost of goods sold, citing a "less than one-year payback on our new capital deployments for compute," the argument that has to hold for the whole allocation framework to make sense. If it holds, the spending is working capital in disguise. If the payback stretches, it becomes a fixed asset base depreciating against contracts signed at today's prices. That tension is visible in the segment reconciliation itself. AI crossed into positive adjusted EBITDA while still posting an operating loss, and the entire distance between the two is non-cash, principally depreciation. Because the depreciation tied to this quarter's build lands in later periods, we expect that gap to widen before it narrows. Elon Musk supplied the supply-side case, arguing that memory output is growing far more slowly than demand, which points toward firm pricing rather than compression.
On architecture, management removed ambiguity: "We're exclusive to Nvidia," with Vera Rubin named as the basis for terrestrial deployment and for the planned Starmind orbital design, a concept that has drawn criticism from researchers on orbital debris grounds and carries regulatory exposure not yet visible in the financials. Musk also indicated that roughly 10% of compute would be reserved for internal Grok training, with the balance directed to inference and third-party rental. That converts the AI segment from a model shop into a landlord. Google and Anthropic were both named as capacity customers, so concentration disclosure in the Form 10-Q deserves a close read. The year-end run rate marker management put forward also assumes contribution from Cursor, an acquisition that has not yet closed, which makes that figure partly an M&A artifact.
Now the contrary take on Connectivity. The ARPU decline is deliberate, a function of localized pricing as the service opens new markets, and enterprise and government revenue is growing considerably faster than consumer, while Gwynne Shotwell noted the company has "never lost an enterprise customer." Credible. It is also untested across a full competitive cycle, and consumer remains the larger of the two Connectivity revenue lines.
Market Analysis
The peer question is unresolved, and the resolution is uncomfortable for the space cohort. We think the capex table is the real read on peers. Judged by capital deployed rather than revenue recognized, SpaceX belongs in the conversation with the hyperscalers that reported over the past two weeks, not the launch providers it is usually bracketed with.
That cohort had a difficult season. Alphabet spent $44.9 billion on capital investment in the quarter and posted negative free cash flow for the first time as a public company, and the shares were marked down on a raised spending guide. Amazon lifted its 2026 capital plan to roughly $220 billion, citing memory pricing, and drew a similar reaction. Microsoft and Meta told versions of the same story. The pattern is consistent: demand is described as supply constrained, capital plans keep rising, and investors are no longer granting the benefit of the doubt on payback. SpaceX arrived into that mood, which helps explain why a beat still traded lower after hours.
The space names offer contrast rather than comparison. Rocket Lab guided its June quarter to a range that SpaceX's AI segment outspends many times over in ninety days. To be fair, that is a statement about capital scale, not execution. Three prints next week should sharpen the read. Rocket Lab and AST SpaceMobile both report on August 10, CoreWeave on August 11. CoreWeave is the cleaner read-across, since backlog conversion and margin trajectory test the same neocloud economics SpaceX is now underwriting. AST SpaceMobile matters for the direct-to-device thesis the FCC-approved EchoStar license transfer is meant to serve, particularly given Shotwell's description of cellular base stations mounted on customer Starlink dishes, a lighter capital path into mobile than a conventional carrier build.
Looking Ahead
The key trend we'll be monitoring is conversion, specifically whether contracted compute translates into recognized revenue at anything close to the payback management described. Three markers arrive quickly. The August 6 lockup expiration makes up to roughly 911.5 million insider shares eligible for sale, more than the current public float, though Musk's own block is understood to remain restricted until June 2027, which caps the genuine worst case. The August 10 and August 11 peer prints should indicate whether neocloud economics are holding or compressing. And Starship Flight 14, with its attempted tower catch of the upper stage, would move Starship from demonstration toward the operational cadence the V3 Starlink deployment depends on. Musk called the upper stage Starship heat shield problem solved, a claim we treat as provisional until a recovered vehicle is inspected. We also need to see a successful catch (or controlled splashdown) of the crucial V3 Super Heavy first stage.
The verdict on this first public print: connectivity funds, launch excites, but AI defines it.
Stephen Sopko | Analyst-in-Residence – Semiconductors & Deep Tech
Stephen Sopko is an Analyst-in-Residence specializing in semiconductors and the deep technologies powering today’s innovation ecosystem. With decades of executive experience spanning Fortune 100, government, and startups, he provides actionable insights by connecting market trends and cutting-edge technologies to business outcomes.
Stephen’s expertise in analyzing the entire buyer’s journey, from technology acquisition to implementation, was refined during his tenure as co-founder and COO of Palisade Compliance, where he helped Fortune 500 clients optimize technology investments. His ability to identify opportunities at the intersection of semiconductors, emerging technologies, and enterprise needs makes him a sought-after advisor to stakeholders navigating complex decisions.