Evid Invest
← Back to Blog

Who actually supplies SpaceX? A data-driven map of the Starlink economy

·EvidInvest Team
SpaceXStarlinksupply chainAetherspace technologysemiconductorspublic-market proxies

Who actually supplies SpaceX? A data-driven map of the Starlink economy

SpaceX is private, valued around $350B in recent funding rounds, and you cannot buy a share of it. But it doesn't build rockets and satellites out of thin air. Its supply chain — and increasingly its customer base — is full of public tickers.

So we did what we always do at EvidInvest: instead of repeating rumor lists, we searched the primary sources. Using Aether, our SEC-native search engine, we queried 10-Ks, 8-Ks, earnings transcripts and press exhibits for direct SpaceX and Starlink evidence — then pulled the financials on every name that survived.

Here's what the data says. (All financials from the EvidInvest database, as of June 2026. Revenue/EPS CAGRs are 5-year unless noted.)

The single cleanest supplier link is in AMD's own earnings slides

Most "SpaceX supplier" lists are inference. This one isn't.

AMD's Q3 2024 earnings presentation, filed with the SEC as exhibit EX-99.2 (accession 0000002488-24-000161, filed 2024-10-29), names SpaceX directly: the latest-generation Starlink broadband satellites are powered by AMD's Versal AI Core adaptive SoCs — the Xilinx acquisition doing exactly what it was bought to do. An EDGAR full-text search for "SpaceX" + "Versal" returns three documents; AMD's own filings are two of them.

What does that exposure sit on top of, financially?

AMD ($AMD)
Market cap$760B
Revenue CAGR, 5Y+28.8%
Revenue growth, last 12m+34.3%
EPS CAGR, 5Y+4.9% (GAAP, acquisition-amortization heavy)

Satellite silicon is a rounding error inside AMD's AI-driven income statement — this is not a SpaceX play, it's an AI play with confirmed SpaceX content. But it matters for the map: with thousands of Starlink satellites launching on a multi-year replacement cycle, AMD is the one company where "supplies SpaceX satellites" is a statement you can footnote to an SEC filing. Run the AMD valuation →

The pure-play: Graham Corporation, the quiet compounder in the map

If AMD is the biggest confirmed name, Graham Corp ($GHM) is the most concentrated one. Its Barber-Nichols subsidiary builds rocket-engine turbomachinery, and its commercial space customer roster — per company materials and industry coverage — includes SpaceX. (Evidence label: company materials, not yet an SEC-filed supplier disclosure.)

The financial profile is the most interesting small cap in this dataset:

Graham Corp ($GHM)
Market cap$1.2B
Revenue CAGR, 5Y+18.3%
Revenue growth, last 12m+13.1%
EPS CAGR, 5Y+42.6%
Trailing P/E~79×

A defense-and-space industrial compounding EPS at 42% a year now trades at a software multiple — the market has found this story. The data question for investors is whether space/defense backlog keeps converting; that's a filings-monitoring problem, which is exactly what Aether is for. Run the GHM valuation →

The Starlink terminal bill of materials — one product, very different shareholder outcomes

Starlink's user terminal is mass-produced consumer-grade hardware: phased-array RF, power management, controllers, connectors, passives. Public teardown and FCC evidence identifies component families from a who's-who of electronics names. (Evidence label: component presence in the device family — not a disclosed supply contract or volume.)

Here's the part most articles skip: these companies are not one trade. Run the numbers and the "Starlink BOM" splits into winners and casualties:

CompanyMkt capRev CAGR 5YRev 1YEPS CAGR 5Y
Amphenol ($APH)$171B+21.8%+51.7%+28.3%
Analog Devices ($ADI)$196B+14.5%+16.9%+6.8%
TE Connectivity ($TEL)$62B+7.0%+7.9%
NXP ($NXPI)$75B+7.3%−2.7%
Texas Instruments ($TXN)$259B+4.1%+13.0%−2.1%
Vishay ($VSH)$8B+4.2%+4.5%
Skyworks ($SWKS)$11B+4.0%−2.2%−8.6%
onsemi ($ON)$46B+2.7%−15.3%−12.6%
Microchip ($MCHP)$48B−2.8%+7.1%−16.6%
Qorvo ($QRVO)$9B−1.7%−1.1%−10.7%

Two observations fall straight out of the table:

  1. Interconnect and analog are the structural winners. Amphenol — connectors, cable assemblies, the unglamorous plumbing of every phased-array terminal and data center — grew revenue 51.7% in the last year. If you want one liquid name levered to physical connectivity demand (Starlink terminals being one driver among many), the data points at APH, not at the RF names.
  2. RF front-end content ≠ RF front-end profits. Qorvo and Skyworks both appear in terminal teardowns, and both have shrinking five-year revenue and negative EPS CAGRs — their fortunes are set by smartphone cycles, not satellite dishes. Component presence in a hot product tells you nothing until you size it against the rest of the income statement.

That second point is the core discipline of this whole exercise.

Flip the telescope: Starlink's customers are louder in SEC data than its suppliers

Here's a finding from Aether that surprised us: search every indexed earnings transcript and press exhibit for "Starlink" and you get 63 distinct segments — and almost all of them are customers, not suppliers. SpaceX doesn't file with the SEC; its buyers do.

The demand-side evidence, with citations:

  • United Airlines ($UAL) — "industry-leading agreement with SpaceX to bring Starlink's Wi-Fi service to more than 1,000 of the airline's mainline and regional aircraft," free for MileagePlus members. It appears in UAL's Q3-2024 and Q4-2024 earnings releases and is still being discussed in Q3-2025 materials (SEC source). UAL: $34B cap, revenue CAGR +30.9% (post-COVID recovery base), trailing P/E ~9×.
  • Southwest ($LUV) — the Starlink Wi-Fi rollout is now material enough to appear in the company's own forward-looking-statements list in Q1-2026 earnings materials.
  • T-Mobile ($TMUS) — the direct-to-cell partnership: "with the beta launch of T-Satellite, T-Mobile now operates the only satellite network in the US that works on most modern smartphones" (Q1-2025 earnings, SEC source). TMUS: $193B cap, EPS CAGR +29.5%.
  • John Deere ($DE) — satellite connectivity for precision agriculture across regions where terrestrial coverage fails; announced January 2024, now part of the Precision Ag narrative in quarterly materials.
  • Carnival ($CCL), Royal Caribbean ($RCL), Norwegian ($NCLH) — fleet-wide connectivity deployments. The cruise recovery numbers (RCL revenue CAGR +52.0%, CCL +36.6% off pandemic lows) aren't Starlink-driven, but always-on connectivity is now part of the onboard-revenue model.
  • Axon ($AXON) — Starlink-enabled field connectivity in public-safety deployments. Axon itself: +32.5% revenue CAGR, one of the fastest compounders in the dataset.

The investable insight: Starlink is becoming infrastructure, and infrastructure shows up in its users' filings as a competitive weapon — free Wi-Fi as a loyalty driver (UAL), satellite texting as a churn reducer (TMUS), connected machinery as a subscription product (DE). You can't buy SpaceX's revenue, but you can see who's building products on top of it, quarter by quarter, in their own SEC documents.

Materials and propulsion: real exposure, thin disclosure

Reusable launch is a materials business — cryogenic fluid handling, valves, actuation, specialty alloys, composites. The public names that fit the technical requirements:

CompanyMkt capRev CAGR 5YEPS CAGR 5YEvidence level
Parker-Hannifin ($PH)$111B+7.7%+24.0%capability proxy
Woodward ($WWD)$21B+7.4%+14.0%capability proxy
ATI ($ATI)$24B+9.0%capability proxy
Materion ($MTRN)$5B+8.7%+36.6%capability proxy
Hexcel ($HXL)$7B+4.7%+29.4%capability proxy

Note the pattern: mid-single-digit revenue growth, double-digit EPS growth — aerospace suppliers in a margin-expansion cycle. None of them needs SpaceX to justify a thesis, and none should be bought because of SpaceX: vertical integration means SpaceX discloses almost nothing about material vendors, and we won't pretend otherwise. These are aerospace-cycle names where commercial launch is incremental demand.

What the data actually says

Four takeaways from running financials across the whole map:

  1. The confirmed-supplier layer is tiny. After filtering to names with direct, citable SpaceX evidence, you're left with a handful — AMD (SEC-filed exhibit) at the top, Graham/Barber-Nichols (company materials) as the pure-play. Everything else is component-presence or capability inference.
  2. The "Starlink BOM" is not a basket. Inside the same terminal you find a +51.7%-growth connector company and RF names with negative five-year CAGRs. Exposure labels don't pay; income statements do.
  3. Demand-side evidence dominates and is growing. 63 Starlink-related segments across SEC transcripts and exhibits, overwhelmingly from customers — airlines, telecom, agriculture, cruise, public safety. Each new partner that mentions Starlink in an earnings document is a free, timestamped datapoint on constellation monetization.
  4. The map needs maintenance, not belief. Labels should move when filings move — a teardown name disclosing a space-segment customer concentration, an airline quantifying Wi-Fi-driven loyalty revenue, a materials name calling out launch-related backlog. That's a search-and-alert problem.

Monitor it yourself

Every claim above traces to a primary source you can re-run:

  • Search the filings with Aether — it indexes SEC filings and earnings transcripts with hybrid semantic + keyword retrieval and returns the exact cited passage (try "Starlink" agreement aircraft or Versal adaptive SoC satellites).
  • Run the financials on any name in the map with EvidInvest: AMD, GHM, APH, UAL, TMUS.

Evidence labels used here: SEC-filed (the claim appears in an SEC document, linked), company materials (public non-SEC sources naming SpaceX), component presence (teardown/FCC evidence in the device family), capability proxy (fits the technical requirement; no direct relationship claimed). Financial research, not investment advice.

Fair Value Weekly

Get DCF breakdowns, fair value updates, and portfolio ideas for serious investors. No spam, no paywalled teasers.