Question: Given that SpaceX itself isn't a good right-out-the-gate buy, what other public companies does the S-1 + S-1/A reveal as investable — across AI-compute, power, and the (unfamiliar) space supply chain?

Method: Took the combined SpaceX+Starlink+xAI S-1 read (companion note, full section-by-section pass) + the priced 6/3 S-1/A, extracted every named/implied third party, fanned out across AI-compute+power / space industrial base / connectivity+value-transfer, then coverage-checked every candidate against our watchlists.

Why SpaceX (SPCX) isn't the trade at the open — priced 6/3 S-1/A: 555,555,555 Class A @ $135 = ~$75B, all primary (enormous new float). Musk = 82.4% of voting power post-IPO (81.1% from Class B), elects a majority of the board with no sunset → full "controlled company." Founder lock-up is 366 days with ZERO early release while everyone else is on a staggered/price-triggered 180-day ladder. Deep losses ($4.3B net in Q1'26), ~$10B/qtr capex, IPO proceeds must first repay a $20B bridge. The interesting money is in the suppliers, not the stock.

The headline insight: the S-1 ratifies theses we already own

The investable supply chain the filing implies is ~80% already in our watchlists. The S-1 is best read as a demand-side ratification of ai-power-bottleneck, optical-supercycle, and memory-supercycle — now with SpaceX/xAI disclosed as a named buyer spending ~$20B/yr (FY25 capex $20.7B; AI capex alone $12.7B FY25 + $7.7B Q1'26; a 1.0 GW compute build on gas turbines; 23,000+ inter-satellite lasers; GB200/GB300 = HBM). It is confirmation, not a new thesis — which is exactly the disciplined read (cf. the @InTheAssembly/@ALEXEIMARTOV "Aschenbrenner recycler" lesson: when our coverage already includes the names, the filing is a cross-validation, and the edge is in the few genuine gaps + the second-order reads).

Caveat that shapes everything below: SpaceX is famously vertically integrated — it builds its own engines, satellites, terminals, lasers, ground stations — so it names very few direct suppliers. "SpaceX supplier" is mostly NOT an investable theme. The real themes are (a) the broader AI-compute/power capex wave the filing quantifies, (b) upstream materials/components that go into all rockets/datacenters, (c) competitors in niches SpaceX can't fill, and (d) value-transfer read-throughs (who holds SpaceX equity, who sold it spectrum).

---

The cleanest reads (start here)

1. SATS (EchoStar) — a synthetic long on the SpaceX IPO at a fixed strike. EchoStar is selling SpaceX its AWS-4/H-block/AWS-3 spectrum for $19.6B, of which ~$11.1B is ~262M SpaceX Class A shares at a FIXED $42.40 (pre-split). With the IPO priced at $135, that consideration is struck far below market → EchoStar holds a large embedded gain on a defined share block (deal FCC-approved 5/12/26, closes ~Nov 2027). It's effectively a long-dated, fixed-strike call on SPCX bolted onto the EchoStar telco. (Already in our space watchlist — but not framed this way. Verify the split mechanics: the dollar consideration is fixed, share count adjusts.)
2. SEI (Solaris Energy Infrastructure) — the ONLY company NAMED as a power-infra partner. Stateline Power JV provides off-grid gas-turbine power to the xAI datacenter. Directly named, small-cap, in ai-power. The physical reason it wins: behind-the-meter gigawatt power for AI is exactly its product.
3. NVDA / TSM — the silicon bottleneck. Every COLOSSUS GPU is an NVIDIA purchase-order (the Anthropic deal alone ≈ 325k NVIDIA GPUs); GB200/GB300 can only be packaged at TSMC CoWoS. Both already covered (ai-infra/semis). Confirmation of the demand, not new.
4. TSLA + GOOGL — balance-sheet mark-up holders. Tesla holds 18.99M SpaceX Class A shares (+ $2B Series E converted, + Megapack/Cybertruck/Terafab ties); Alphabet is a named investor with registration rights. The IPO marks up both. Cleanest because the S-1 names them explicitly.

---

The layered map (with coverage flags)

[in: …] = already in our watchlist(s); [NEW] = coverage gap. Direction = how SpaceX's success cuts.

AI compute (mostly already covered — confirmation)

Optical & networking (inter-sat laser + cluster fabric → our optical-supercycle)

Power generation & supply (the sharpest gap cluster)

Electrical / cooling (covered)

Space / launch industrial base (the part you said you don't know)

Plain version: SpaceX builds almost everything itself, so the public "space supply chain" is really (1) competitors (some hurt by SpaceX), (2) upstream materials that go into every rocket regardless of who wins, and (3) defense primes riding government space budgets.

Private — don't chase: Blue Origin, ULA (BA/LMT JV), Relativity, Arianespace, Spectrolab (Boeing), Precision Castparts (Berkshire). Ticker corrections: MAXR taken private 2023 (drop), AJRD absorbed into LHX (drop standalone).

Connectivity — winners (partners) and losers (disrupted incumbents)

---

What's genuinely NEW vs already-tracked (the actionable shortlist)

Everything high-conviction (NVDA/TSM/AVGO/semis, the optical six, SEI/GEV/ETN/VRT/BE/HWM/ATI/CRS in ai-power, RKLB/FLY/SATS in space, the defense primes) is already covered — so the S-1 mostly validates existing positioning. The coverage gaps worth a price/fundamental screen:

• Natural-gas-to-datacenter layer: KMI, DTM (pipeline/midstream), AR (low-cost E&P). The S-1 is explicit that the 1 GW build runs "significantly on natural gas" + a $2B mobile-turbine buy — and our AI-power coverage has the turbines (GEV/SEI) but is thin on the gas-supply bottleneck. Best new angle. — RESOLVED 2026-06-04: screened (price + fundamentals). Not a buy-watch — DTM/HWM are extended leaders (+36/+42% 1Y), KMI is fully priced, AR is cheap only on a peak-cycle EPS quarter (value-trap signature). Basket repurposed as a rotation tell (co-surge = money leaving crowded AI-infra), folded into the SpaceX-IPO standing-watch row. Don't re-propose as a laggard-buy.
• HXL (aerospace carbon fiber) and MOOG (actuators) — real space/aero component bottlenecks, not in any list.
• RMBS (memory-interface IP) — memory-supercycle adjacency that's surfaced before and still isn't tracked.
• SPXC (transformers) — the 2-yr-lead-time power-delivery bottleneck.
• The "Starlink disruption losers" basket (VSAT, TSAT, LUMN, CHTR) — not currently framed as a short-watch set; could seed one if we want the bear side.

Discipline / what this is NOT

• Candidates, not buy calls. No prices, multiples, or returns computed here (price-truth comes from the scan summaries, not the filing). Direction labels (winner/loser) are qualitative reads from the S-1.
• Producer-side only — no watchlist edits, no trades. Recommended next step is the user's call: (a) add the 4-6 gap names to a watchlist and refresh the scan summaries, or (b) a deeper trace on the gas-supply + space-materials layers.
• Verify before sizing: MYNA (solvency/listing), BKSY (ticker), foreign ADR depth (SIEGY/KDDIY/VOD/RCI), and the EchoStar fixed-share mechanics.

Sources

• 2026-06-04-spacex-s1-full-read-section-notes (the full S-1 read).
• SEC EDGAR: SpaceX S-1 (5/20, acc 000162828026036936) + S-1/A priced (6/3, acc 000162828026040364, 555,555,555 sh @ $135, Musk 82.4% voting).
• Coverage check run against * (2026-06-04).
• Thread 2026-06-02-ai-mega-ipo-fast-entry-liquidity-regime; cross-ref 2026-06-02-xai-s1-segment-crosswalk-read.