Topic

SpaceX IPO: The AI Infrastructure Behind $SPCX

SPACEX-AIINFRA-ANALYST-01

Buy/sell-side AI-infra analyst reading the SpaceX S-1 for compute-supply-chain exposure, not the rocket valuation.

Audience

  • · 6-15
  • Current: Equity/Credit Analyst — compute/semis/infra coverage
  • Pain: Separating reported/expected figures from S-1-confirmed before the public filing
  • Pain: CoreWeave as the only clean public neocloud comp for Colossus-class economics

Product Needs

(none)

Channels

(none)

Competitor Lens

(none)

Fit Score weights — adjust to your priorities

30%
25%
25%
15%
5%
Top 5 for this segment
  1. 1. NVIDIA64/100
  2. 2. Anthropic60/100
  3. 3. Astera Labs58/100
  4. 4. SpaceX56/100
  5. 5. Coherent Corp.55/100

Full Persona Brief

Buy/sell-side AI-infra analyst reading the SpaceX S-1 for compute-supply-chain exposure, not the rocket valuation.

Audience Profile

  • Age / Experience: 6-15 years
  • Current role: Equity/Credit Analyst — compute/semis/infra coverage (hedge fund / asset manager / sell-side research)
  • Top pain points:
  • Separating reported/expected figures from S-1-confirmed before the public filing
  • CoreWeave as the only clean public neocloud comp for Colossus-class economics
  • Tracing the L1-L5 supplier read (power/optical/networking/GPU) from sparse public sourcing
  • Top decision blockers:
  • Thin public sourcing on the Colossus physical layer (SpaceX vs xAI vs colo)
  • One external customer is one data point — merchant-market thesis unfalsified pre-listing

What This Segment Needs

  • Information: S-1-confirmed vs reported/expected splits on the Colossus physical layer; CoreWeave read-across for Colossus-class neocloud unit economics.
  • Tools: a filing-diff tracker (confidential S-1 2026-04-01 → public S-1 ~5/20 → SPCX list 6/12) and an L1-L5 supplier-read model (power → optical → networking → GPU).
  • Services: primary-source / expert-network channel checks resolving SpaceX vs xAI vs colo ownership of the Memphis cluster and Anthropic tenant terms.

Top 5 Companies for You (Fit Score)

| Rank | Company | Score | Why | |------|---------|-------|-----| | 1 | NVIDIA | 86/100 | Data-center revenue ~$115.2B FY2025 (+142% YoY), ~88% of $130.5B; Blackwell GB200 NVL72 several-$B in first quarter (Q4 FY2025); underpins Colossus scaling 100k→200k+ GPUs. ~75% gross margin, net-cash. | | 2 | Astera Labs | 82/100 | 10-Q quarter ended 2026-03-31: revenue $308.4M vs $159.4M (+93.4% YoY), net income $80.3M; R&D 40.7% of revenue (UALink/PCIe Gen6/CXL); ~$1.18B cash, zero debt. | | 3 | SpaceX | 80/100 | Confidential S-1 filed 2026-04-01, public S-1 ~5/20, SPCX list 6/12 at $1.75-2T (~$75B raise) vs ~$350B 2025 tender; Colossus 100K+ Hopper Memphis, ~150MW+ turbines; Anthropic first external customer 2026-05-06. | | 4 | Anthropic | 80/100 | Run-rate ~$1B→$5B+ during 2025 (~5x), $13B Series F at $183B post (2025-09-02); Deloitte ~470k + Cognizant ~350k seats standardized, not pilots. Tri-cloud infra. | | 5 | Coherent | 74/100 | Q3 FY2026 revenue $1.81B (+20.6%), diluted EPS $0.97 vs $(0.11); R&D ~10% of revenue, vertically integrated InP fab→datacom optics; equity $11.0B. |

Deal-Breakers (Your Hard Preferences)

No hard preferences declared for this segment.

How to Evaluate Any Company in this Niche (Checklist)

  • [ ] Check growth signals: pull DC/AI-infra revenue YoY % and segment mix from the latest 10-Q/S-1 itself — flag every figure tagged "reported, not filing-confirmed" (e.g. SpaceX ~150MW+ turbine power).
  • [ ] Check comp data: triangulate Colossus-class economics against CoreWeave (the only clean public neocloud comp) on $/GPU-hour and contracted backlog/RPO.
  • [ ] Check learning signals: trace the L1-L5 read explicitly — power (MW + turbine permits), optical (Coherent capex/inventory), networking (Astera Spectrum-X/UALink), GPU (NVIDIA Blackwell/Rubin ramp).
  • [ ] Check stability signals: quantify customer concentration — top-5 % of revenue (Astera ~90%; NVIDIA DC ~40-50% in 2-3 buyers) and single-tenant exposure (Anthropic = sole external Colossus customer).
  • [ ] Check geopolitical signals: read China revenue % and export-control risk-factor language (Astera 29% China; NVIDIA H20/China-spec charge).
  • [ ] Check culture/governance: ask whether any culture claim is filing-confirmed vs AI prior — several names here scored null; do not model attrition off job-mix proxies.

Reverse-Hype Watch

  • NVIDIA customer-concentration: ~40-50% of data-center revenue in a handful of hyperscaler/neocloud buyers; the Colossus signal is a single-customer illustration — a capex pause among 2-3 buyers hits revenue sharply.
  • NVIDIA rest-and-vest: ~2-3% attrition vs ~13% industry is partly equity-driven golden-handcuff lock-in, not pure satisfaction.
  • SpaceX: xAI's ~1M-GPU ambition is "reported; explicitly NOT confirmed" while Colossus has only one external customer (Anthropic, 2026-05-06) on a multi-provider sourcing posture — capacity claim unbacked by customer breadth.
  • Anthropic up-round-with-burn: $30B Azure + up to 1GW NVIDIA + up to 1M Google TPU and $183B post-money are sized far above a <$5B run-rate; not profitable.
  • Coherent: capex $547M (+77%) and inventory +48% (to $2.13B) ahead of AI-optics orders while 180d business signals are empty and operating cash fell to $10.1M from $503.3M.

Under-reported for this segment: coverage fixates on the $1.75-2T headline valuation and launch cadence, but the analytically decisive variable — the physical-layer ownership split of Colossus (SpaceX vs xAI vs colo: power-purchase contracts, gas-turbine permitting, land/grid interconnect) — is the least-sourced item in the public record. Until the public S-1 confirms who books the power and the GPU lease, the merchant-neocloud thesis stays unfalsified, and Anthropic-as-sole-tenant means the comp set is one data point, not a market.