SpaceX's June 2026 IPO was the largest in history — $75 billion raised at a $1.8 trillion valuation. Its first earnings report as a public company showed a profitable, high-margin Starlink business bolted onto a money-losing AI cloud segment burning through $15.8 billion of quarterly capex. Its supplier list splits the same way, and one long-assumed "SpaceX supplier" stock is now the subject of a short-seller report alleging the relationship itself was oversold.
"SpaceX supplier" has become one of 2026's most reliable stock-pitch phrases, and since June it has had an unusually clean anchor: SpaceX itself is now a public company, trading under the ticker SPCX, so "as SpaceX rises" is no longer a metaphor — it is a chart anyone can pull up. But the company behind that chart turns out to be running two quite different businesses under one name, growing at different speeds with different economics, and the suppliers that get paid by one are not always the same ones exposed to the other. Sorting out which is which is the entire point of this letter.
SpaceX priced its initial public offering on June 11, 2026, raising $75 billion at a valuation of roughly $1.8 trillion — the largest IPO in history by a wide margin. Shares closed their first day of trading up 19% at $161. The company's first earnings report as a public entity, covering the second quarter of 2026, landed on August 4 and delivered a genuine beat: revenue of $7.8 billion, up 92% year on year, roughly $900 million ahead of Wall Street's $6.9 billion consensus. Shares fell more than 7% in after-hours trading anyway. The reason is the split this letter is built around.
Starlink generated $4.29 billion of that quarterly revenue, up 66% year on year, on 12 million subscribers — double the subscriber count of a year earlier — at an operating margin of 38.6%, ahead of the 35.9% analysts had modeled. This is a real, profitable, scaling consumer and enterprise broadband business, and it is the part of SpaceX that most closely resembles the "space company" investors thought they were buying.
The newer and less widely understood piece is SpaceX's AI cloud segment, which posted $2.6 billion in revenue — up 247% year on year, driven by $14.1 billion in freshly signed cloud-services contracts — but still ran a $1.26 billion operating loss for the quarter, an improvement from a $2.47 billion loss the quarter before, not a business anywhere near breakeven. Total quarterly capital expenditure reached $18.4 billion, of which $15.8 billion — well over four-fifths — went to the AI segment specifically. Annualized, that capex run-rate is roughly $73.5 billion, comfortably ahead of the $48.7 billion analysts had expected. Elon Musk used the earnings call to move up his own internal target for $1 trillion in annual revenue from 2031 to 2030, with what he called a "non-zero chance" of hitting it as early as 2029 — an aggressive acceleration riding almost entirely on the AI segment's growth, not Starlink's.
| Segment | Q2 2026 Revenue | YoY Growth | Profitability |
|---|---|---|---|
| Starlink | $4.29B | +66% | 38.6% operating margin |
| AI / cloud | $2.6B | +247% | -$1.26B operating loss |
| Total company | $7.8B | +92% | -$541M net loss (narrowing) |
The market's negative reaction to a genuine revenue beat is the clearest evidence that "SpaceX is rising" now means at least two different things, and investors need to be specific about which one they mean before picking a supplier to ride along with it. A supplier tied to Starlink's hardware build-out (user terminals, RF components, satellite structures) is exposed to a profitable, disciplined, margin-expanding business. A supplier tied to the AI segment (GPUs, fabrication capacity, data-centre buildout) is exposed to a business still losing more than a billion dollars a quarter and spending capex at a rate that already exceeds Street expectations — closer in character to the AI infrastructure story covered elsewhere in this letter series than to a traditional aerospace supply chain.
These are the names most directly tied to Starlink and launch-vehicle hardware build volume: STMicroelectronics (STM) supplies RF chips and microcontrollers for Starlink terminals; Honeywell (HON) supplies avionics and flight-control systems; Hexcel (HXL) and Carpenter Technology (CRS) supply carbon-fiber composites and specialty engine alloys; Materion (MTRN) supplies niobium and advanced materials; Redwire (RDW) supplies space infrastructure and solar arrays; Moog (MOGA) supplies precision motion-control systems; and on the Taiwan side, Wistron NeWeb (6285.TW) manufactures Starlink user terminals, with Compeq (2313.TW) and Tong Hsing Electronic (6271.TW) supplying satellite printed-circuit boards and RF modules. Britain's Filtronic (FTC.L) supplies RF/GaN amplifiers and has disclosed what it calls its largest-ever SpaceX contracts — one valued at $62.5 million, a second at £47.3 million with deliveries beginning in FY2027 — though the company's own February 2026 interim results emphasize a deliberate strategy of reducing customer concentration by winning new contracts elsewhere (a €7.0 million European aerospace deal, a £13.4 million European defence deal), which reads as an implicit admission that SpaceX had been a large share of revenue and the company is actively trying not to be defined by that alone.
These sit on the other side of the split, exposed specifically to the loss-making AI segment's build-out rather than to rockets or satellites at all: Nvidia (NVDA) for GPU and AI computing hardware, Broadcom (AVGO) for communication semiconductors, TSMC (TSM) for chip fabrication, and Intel (INTC) as a disclosed "Terafab" fabrication partner. Alphabet (GOOGL) is a different category again — a financial stakeholder, not a supplier, holding a stake in SpaceX reportedly worth roughly $150 billion after the IPO, meaning its exposure is to SpaceX's total enterprise value rather than to any specific segment's revenue.
Velo3D (VLD) has circulated widely in 2026 coverage — including in this letter series' own September 8 issue on AI-accelerated 3D printing — as a SpaceX-linked additive-manufacturing supplier. A short-seller report published by Morpheus Research makes a specific, documented case that this relationship has been substantially oversold. SpaceX was genuinely Velo3D's largest customer at the company's 2021 SPAC IPO, at 27.8% of revenue. By 2023, SpaceX had stopped ordering new printers and Velo3D's SpaceX revenue share had fallen to roughly 4% of the total. In October 2024, SpaceX paid Velo3D a one-time $8 million fee for a perpetual, royalty-free license to manufacture printers in-house — after which, excluding that one-time payment, Morpheus estimates SpaceX-derived revenue at roughly 3.4% of Velo3D's total. Despite this, CEO Arun Jeldi stated in December 2025 that "SpaceX is actually increasing their fleet and also sending us parts to print" — a claim Morpheus contrasts with Velo3D's own disclosures showing SpaceX revenue below the 10% customer-concentration reporting threshold, and with former-employee statements to the effect that SpaceX had "found an alternative" supplier and was unlikely to buy another Velo3D system. Neither Velo3D nor SpaceX has published a documented rebuttal to these specific claims as of this letter. This does not make Morpheus's allegations proven fact — short-seller reports carry their own incentive to talk a stock down — but it is a serious, specific, sourced challenge to exactly the narrative this letter's own supplier lists (and much of the broader financial press) have repeated, and it belongs here rather than being quietly omitted.
The single most useful diligence question for any "SpaceX supplier" stock in 2026 is not "does SpaceX buy from this company" — plenty of names can answer yes to that. It is "how much of this company's current revenue does that relationship actually represent, and is it growing or quietly shrinking behind a confident quote." Velo3D is this year's clearest lesson in why that second question matters more than the first.
SpaceX's own stock reaction to a genuine beat is itself the central risk this letter has tried to make legible: a market pricing the AI segment's losses and capex more heavily than Starlink's profits means any supplier levered to that segment inherits AI-infrastructure-cycle risk, not aerospace risk. Starship's own commercial reliability remains unproven at the cadence Redwire's bull case requires. And the Velo3D situation is a live reminder that "SpaceX supplier" status, once granted by financial media, tends to persist in coverage and in valuation well after the underlying commercial relationship has quietly changed — a risk that likely applies, in smaller and less documented ways, to other names on this list that this letter cannot fully verify from public disclosure alone.
SpaceX going public did not create one trade; it created at least three — a profitable satellite-broadband business, a loss-making AI cloud build-out, and a supplier ecosystem that splits cleanly along the same line once you look past the "SpaceX supplier" label itself. Favor diversified, disclosed-contract names (Filtronic, STMicroelectronics, TSMC) for measured exposure with a lower single-customer risk, treat concentrated pure-plays (Redwire) as a bet on Starship's commercial maturity specifically, and verify — don't assume — the current dollar reality behind any stock still trading on a SpaceX association it built its reputation on years ago.