S-1 filed · Nasdaq SPCX · trading expected June — the largest IPO in history, bundling Starlink, launch, and xAI. We dissect the business, financials, and technology end-to-end. (See the Latest Update box below for confirmed 2026.05 S-1 figures.)
Ahead of the roadshow, the offer price and trading date are now set. The figures below are cross-checked across multiple outlets and supersede the estimates in the body and the S-1 snapshot below.
Sources: CNBC, Reuters, Yahoo Finance, EchoStar IR, FCC, SpaceNews (June 2026 reporting, cross-verified). The fixed price and schedule may change with roadshow demand and market conditions.
Since this piece was first written, the SpaceX IPO has moved from speculation to reality. Below are the key facts confirmed or updated after the public S-1 filing. Where they differ, the figures in this box supersede the estimates in the body.
Sources: Bloomberg · CNBC · NBC News · Fortune · Via Satellite · SEC EDGAR · Space.com · TechCrunch · xAI official · DCD (Feb–Jun 2026). We do not assert an exact trading date or the inflated synthetic-market valuations (e.g., $2.4T).
The largest IPO ever, with 21 investment banks on the syndicate. Roughly 3× the size of Saudi Aramco's $29B (2019).
The largest valuation climb ever for a private company. Half of the 2026 5× jump, however, is the xAI merger.
Secondary share sales · ARK / Sacra / Bloomberg combined
Jul 2025 $400B → Dec 2025 $800B (tender, $421/sh) → Feb 2026 $1.25T (xAI acquisition) → Apr 2026 IPO target $1.75T → May 2026 S-1 filed → Jun 2026 priced at $135/sh ⇒ $1.77T (5-for-1 split, trading Jun 12).
① Starlink hits scale-profit. 9M+ subscribers, EBITDA margin 63%. ② Government backlog. NSSL Phase 3 $5.9B + Golden Dome $2B. ③ xAI merger. All-stock swap absorbed $250B of xAI value, lifting the combined entity to $1.25T. ④ Orbital data-center narrative. Goldman / Morgan Stanley assigned a vertical-integration premium for "AI infra + comms + launch" under one roof.
Starlink is the cash engine, Falcon owns launch, Starship is the future, xAI is the AI ticket, Starshield is national-security cash flow.
The single operator that controls ~65% of all active satellites globally. Crossed 10M subscribers in Feb 2026 and roughly 11.8M by April (150+ countries). Around 10,000–12,000 active birds — the largest satellite constellation in human history. 2025 Starlink-only EBITDA between $4.4B and $7.2B, depending on whether you cite Reuters' S-1 review (operating-income basis) or Sacra (full EBITDA basis).
The 2025 EchoStar acquisition pulled in $17B of AWS-4 / H-Block spectrum ($8.5B cash + $8.5B stock), unlocking the launch rights for next-gen satellites with 100× the capacity of first-gen D2C. Global commercial deals signed with T-Mobile, Optus, Telstra, Rogers, KDDI, and Deutsche Telekom.
82% of global commercial launch. 165–170 launches in 2025 (100% payload success). Up to 32 reuses per booster, with 40-flight certification in progress for 2026. Per-launch cost compressed from $400M → $62M, putting LEO at roughly $2,700/kg.
Competition has been absent for so long that the gap will keep widening. Blue Origin's New Glenn is now in commercial service but hasn't yet proven reuse at scale. ULA's Vulcan exists as the government's second-source.
Fully reusable super-heavy-lift. 150-tonne LEO payload, capable of deploying 60 Starlink V3 satellites per flight (20× a Falcon 9). Cumulative dev cost $15B+. Holds NASA's Human Landing System (HLS) contract for Artemis.
Only 5 flights in 2025 — 5× short of the 25-flight target. The V3 (Block 3) debut = Flight 12 launched on 2026.05.22; the upper stage deployed 22 Starlink simulators, but the booster catch failed and the FAA grounded the vehicle on 5/27 pending an investigation — bad timing right before the IPO roadshow. FAA caps stand at 25 flights/yr at Starbase Texas and 44/yr at LC-39A in Florida.
Velocity-vs-credibility pattern: operating programs (Falcon, Starlink) hit promises within ±10%, but new vehicles (Starship) slip 2–5 years. Falcon Heavy was 5 years late; Crew Dragon, 3. Starship is following the same arc.
Became a SpaceX subsidiary in Feb 2026 via an all-stock swap at $250B (xAI) / $1T (SpaceX). Tesla had already invested $2B in advance. Grok 4 is the current production model. Per Musk's roadmap on X, the team is concurrently training seven variants from 4.4 → 4.5 → Grok 5 (6T / 10T parameter editions). ※ xAI hasn't published exact parameter counts; numbers above are extrapolated from Musk's public posts.
Infra: Colossus 1 (200K GPUs) + Colossus 2 (Gigawatt-class), sites in Memphis and Southaven, Mississippi. Stood up 100K H100s in 122 days — a deployment speed Jensen Huang himself called "supernatural." Pathing to a 1M-GPU cluster.
Benchmarks: Grok 4 hit ARC-AGI-2 at 15.9% (GPT-5: 9.9%, Gemini 2.5 Pro: 21.6%) and Humanity's Last Exam at 25.4%. Grok 4 Heavy, with multi-agent parallel reasoning, reached 44.4%.
Caveat: the consolidation drove a brutal P&L hit. 2025 AI-segment operating loss: $6.4B. AI absorbed 61% of capex ($12.7B). Burn rate roughly $1B/month (Bloomberg). Most of the 12 founding members have left — Reuters reported 6 still inside as of Feb 2026, with FT later updating that only 2 of the original co-founders remain.
🆕 New revenue — Colossus 1 leased entirely to rival Anthropic
Anthropic is renting all of Colossus 1 (220K+ NVIDIA GPUs · 300MW) in Memphis. It pays $1.25B per month (discounted for the first two months during xAI's ramp-up). That annualizes to a ~$15B run-rate — about 80% of SpaceX's entire 2025 revenue ($18.7B), so it is not a line you can ignore. The deal was first disclosed in the S-1.
Contract length is disputed. SpaceX's S-1 states the term runs through May 2029 (over $40B total), but Musk clarified on X that it is effectively a flexible 180-day lease. What both agree on: either side can terminate on 90 days' notice. So the $40B-to-2029 headline assumes the full term, and with the 90-day exit plus Musk's 180-day framing, revenue beyond ~6 months is not committed.
Profit is not separately disclosed. Colossus 1 is already built, so incremental lease revenue likely carries a high gross margin — but the AI segment still burns ~$1B/month, so this deal does not by itself make the segment profitable; the net-income contribution is not disclosed. On renting compute to a rival, Musk said "no one set off my evil detector," while Anthropic is using it to lift Claude Pro/Max rate limits and expand Claude Code and API capacity.
Starlink's government / defense variant. Operates reconnaissance constellations for NRO and DoD (the Starshield program). Cumulative government contracts since 2008: $24B+ (CNBC / FedScout). April 2025 NSSL Phase 3 Lane 2 was $5.9B (28 launches through ~2029); the Pentagon's Golden Dome is $2B for 600 satellites.
Space Force has assigned 97% of its PLEO Starshield task orders to SpaceX (within a $13B ceiling). In April 2026 the Space Development Agency added another $178.5M order for missile-tracking-satellite launches.
Starlink delivers 70–80%. Launch revenue grows in absolute terms but shrinks as a share.
Sacra Equity Research / Reuters S-1 review
Quilty Space, Mar 2026 / spacexstock.com
| Segment | 2025 ($B) | 2026E ($B) | EBITDA Margin |
|---|---|---|---|
| SLStarlink | 11.4 | 16–20 | 60–63% |
| FLFalcon Launch | 4.1 | 5–6 | ~35% |
| GVGovt · Starshield | ~3.5 | 6–7 | ~40% |
| AIxAI / Grok | ~0.5 | 2–3 +lease* | deep loss |
| Total | 15.5–18.7 | 25–28 | ~40% |
※ Confirmed by the public S-1 (2026.05.20): 2025 total revenue $18.7B, net loss -$4.9B, adjusted EBITDA +$6.6B (Starlink $11.4B · Launch $4B · AI $3.2B). The earlier "pre-merger $15–16B profit" estimate is retired.
*New AI-segment lease revenue (not in the table): Anthropic rents all of Colossus 1 for $1.25B/month → a ~$15B annualized run-rate (~80% of SpaceX's 2025 total). But the S-1 states the term runs "through May 2029 / $40B+," while Musk called it a "flexible 180-day lease," and either side can exit on 90 days' notice → the full-term headline is large, but beyond ~6 months is uncommitted, so the 2026E total above conservatively excludes it. Net-income contribution undisclosed. (Sources: SpaceX S-1 · TechCrunch · xAI official · DCD)
SpaceX's moat isn't a single technology — it's the vertically integrated manufacturing speed and reuse economics.
Starship's heart. The first-ever mass-produced full-flow staged-combustion (FFSC) engine in human history. Raptor 3 upgrades thrust, reliability, and manufacturability simultaneously. Compared with Blue Origin BE-4 or ULA RL-10, the unit cost is 1/10 to 1/50, and the manufacturing tempo — a full booster + ship set every roughly 15 seconds at peak rhythm — is the real moat.
Falcon 9 boosters average 20+ reuses, with one booster having flown 32 times. LEO cost-per-kg is 1/4 to 1/8 of the satellite-industry average. If Starship V3 hits steady state, it can drop below $100/kg — the level where orbital data centers, space tourism, and point-to-point transport all tip into the realm of economic feasibility.
Today Falcon 9 deploys 28–29 V2 Mini satellites per launch. Starship will lift 60 V3 satellites per flight, each with ~100× the per-bird payload capacity of Gen 1. Combined with the EchoStar spectrum, D2C downlink will jump from 4 Mbps → 150 Mbps (test rollout late 2027, mass deployment 2028).
On a former Memphis Electrolux factory site, the team deployed 100K H100 GPUs in 122 days (industry average: 4 years). NVIDIA CEO Jensen Huang publicly called the pace "supernatural." Colossus 2 is the world's first Gigawatt-class training supercluster — 168 Tesla Megapacks plus proprietary liquid cooling.
Environmental flip side: 35 gas turbines providing 422 MW of interim power are putting xAI in a long-running fight with Memphis residents and environmental groups.
In Jan 2026 SpaceX filed with the FCC for a constellation of up to 1 million satellites. Solar power plus 24/7 sun in orbit lets you sidestep the power and cooling ceilings ground data centers face. ARK estimates that, below $100/kg, orbital compute is ~25% cheaper than terrestrial. Musk's stated goal: launching 100 GW of AI compute capacity per year.
This is the heart of SpaceX's merger-style valuation — a strategic premium that simple Sum-of-the-Parts can't capture.
Raised from the initial $1.75T target to $2T+. With the company still net-loss-making (-$4.9B), the debate is "growth narrative vs profitability."
Sorted by probability and impact on market cap.
| Risk | Likelihood | Impact |
|---|---|---|
| Starship V3 further delay 12+ months additional slip | High | Mid |
| xAI losses widen, talent leaves $1B+/mo burn continues | Mid-High | High |
| FCC additional spectrum denial April 2026 denial extends | Mid | Mid |
| Musk governance incident Political activity, X posts, etc. | Mid-High | High |
| Starlink ARPU drops further Faster emerging-market penetration | High | Low |
| Competitor entry Kuiper, AST SpaceMobile | Mid | Mid |
| Regulatory / antitrust (FCC, EU, SEC) Grok investigations spill into parent | Mid-High | High |
| Audit shock Post S-1 revenue / loss reset | Mid | High |
From IPO day to the next Mars window.
SpaceX's moat isn't a single technology — it's vertically integrated manufacturing speed. Launch → satellites → comms → AI → government contracts all close inside one company, with less external dependency than any other Big Tech. That's the case for a 95× sales multiple.
But two-thirds of the value is still unproven. Starship V3, on-orbit propellant transfer, full-speed D2C, the Artemis lunar landing, orbital data centers, Mars — none of it is verified at IPO time. The IPO price has already booked all of it.
Buy framework: ① Scale-in — 30% on IPO day, 30% after the first earnings, 40% once Starship V3 is stable. ② If direct ownership is too hot, get indirect exposure via XOVR (16.2% allocation), DXYZ (23%), RONB (14–22%), or direct EchoStar (SATS) shares. ③ Use Musk political / verbal headline volatility as buying windows.
Avoidance signals: ① First-quarter audit lands far from the consensus estimates. ② Further Grok governance trouble or talent flight. ③ Another double-digit drop in Starlink ARPU. ④ FCC or antitrust enforcement.
Bottom line: "It will go up — but expect 2× the volatility of NVDA". If you've followed any Musk company, you know the pattern. Only commit IPO money you can hold for 5–10 years; trade short-term volatility through the indirect-exposure ETFs.