xAI — from narrative to largest segment.
xAI was added through an all-stock merger that closed February 2026; the combined entity was valued at $1.25 trillion at closing (xAI ~$250B, SpaceX standalone ~$1T at the time). SpaceX reports it as the AI segment — renamed Super Intelligence (SI) in the Q3 2026 earnings release — covering the Grok model line, the X platform (formerly Twitter), Cursor (acquired Aug 14, 2026) and early-stage orbital AI research. FY2025 segment revenue was $3,201M on a $(6,355)M operating loss. In Q3 2026, SI revenue reached $7,560M — about 52% of SpaceX's total — with $431M of operating income.
How three companies became one segment.
The segment is itself a roll-up. The corporate actions that matter:
| Date | Transaction | Implied value |
|---|---|---|
| Mar 2025 | xAI acquires X Corp (formerly Twitter) in an all-stock deal | Combined entity ~$110B at announcement |
| Feb 2026 | SpaceX absorbs xAI in an all-stock merger | Combined entity ~$1.25T at closing (SpaceX ~$1T + xAI ~$250B) |
| May 2026 | S-1 filed for combined entity | Target IPO valuation $1.77T |
| Aug 14, 2026 | SpaceX closes its acquisition of Cursor | Implied equity value $60B |
The accounting treatment under purchase accounting: xAI's pre-merger assets, including X Corp, were rolled onto the SpaceX balance sheet at the merger closing date at fair value. Goodwill and intangibles were recognized to the extent purchase price exceeded identifiable net assets. The S-1's pro forma footnote (Note 3) lays out the allocation; goodwill on the xAI segment is the largest single intangible on the consolidated balance sheet.
The Super Intelligence segment, quarter by quarter.
SpaceX called this segment AI in the S-1 and the Q2 2026 release, and renamed it Super Intelligence (SI) in the Q3 2026 release. It went from the smallest segment by revenue in FY2025 to the largest in Q3 2026:
| Period | Revenue | Op income (loss) | Segment capex |
|---|---|---|---|
| FY2025 | $3,201M (17.1% of total) | $(6,355)M | $12,727M |
| Q1 2026 | $818M | — | — |
| Q2 2026 | $2,561M | $(1,257)M | $15,828M of $18,369M group |
| Q3 2026 | $7,560M (~52% of total) | $431M* | $16,444M of $19,593M group |
Sources: S-1 (FY2025); SpaceX Q2 2026 8-K (Aug 4, 2026) and Q3 2026 8-K (Oct 20, 2026). Q1 2026 revenue is the comparative in the Q2 8-K; “—” means not shown in the figures summarized here. *Includes $1,217M of acquisition-related stock-based compensation, mostly Cursor.
Compute. SpaceX's reported AI nameplate compute has stepped up from 1.0 GW to 1.4 GW to 1.7 GW at Sept 30, 2026. On the Q3 call, SpaceX said it expects 2.5–3 GW by the end of Q4 2026, and guided Q4 group capex to $35–40B. SpaceX also rents AI compute to Google, Anthropic and others, according to CNBC.
Capex. SI dominates SpaceX's capital spending: $15,828M of $18,369M in Q2 2026 and $16,444M of $19,593M in Q3. For the consolidated picture, see the SpaceX earnings page.
Four businesses, one P&L line.
Grok & xAI models
The xAI model family — Grok and successor versions — is the core AI product. Revenue runs through three channels:
- API access. Direct developer access to Grok via the xAI API. Pricing competitive with Anthropic, OpenAI, Google.
- Integration into X Premium. Grok is the headline AI feature inside X Premium and X Premium+ subscriptions.
- Enterprise licensing. Early-stage. The S-1 references "enterprise pilots" without breaking out revenue.
X platform
The X platform (X Premium, X Premium+, X advertising) was a core revenue source for the segment at the time of the S-1. SpaceX doesn't break out X revenue from xAI model revenue within the segment's $3,201M of FY2025 revenue.
Cursor
SpaceX agreed to acquire Cursor for $60B and closed the deal on Aug 14, 2026 at an implied $60B equity value. The deal shows up in SI's Q3 results: the segment's $431M of operating income is after $1,217M of acquisition-related stock-based compensation, mostly related to Cursor. SpaceX does not break out Cursor's revenue in the figures summarized here.
Orbital AI (research)
The most speculative line in the prospectus. The S-1 describes a multi-year research program to host inference and training workloads on satellite-mounted compute, leveraging:
- Continuous solar power. LEO satellites can have near-continuous solar exposure with optimized orbits.
- Passive radiative cooling. Vacuum makes heat rejection a different physics problem than terrestrial data centers.
- Proximity to Earth-observation data sources. Inference at the data, rather than downlinking petabytes for ground inference.
No revenue line is attached. Investors should treat orbital AI as a 5–10 year option, not a 2026–2027 P&L driver. The Starship payload capacity to LEO is the gating constraint — orbital data centers require Starship to be in commercial service.
Three ways the merger creates value.
The S-1 outlines three integration narratives, in order of how concrete each one is today:
1. Network optimization (live)
Grok already runs production traffic routing and capacity allocation across the Starlink mesh. The S-1 confirms this in the technical description of Starlink network operations. Engineering output of the merger is most concrete here — and any benefit shows up indirectly in the Connectivity segment, which earned $4,423M of operating income on $11,387M of FY2025 revenue.
2. Distribution (active)
The X platform provides a captive marketing surface for Starlink and xAI products. The S-1 references co-marketing of Direct-to-Cell with T-Mobile US through X, and bundle pricing for X Premium + Starlink Roam subscribers. Distribution synergy is real but moderate — X has ~600M monthly active users; the marketing surface area is large but conversion is gradual.
3. Orbital AI data centers (research)
The most speculative integration. Discussed above. No revenue attached. Depends on Starship commercial service.
What can break the xAI story.
- Compute economics. Frontier model training costs continue to rise. xAI's Memphis-area Colossus data centers are among the largest single training installations, and the capex line is material and ongoing — SI capex was $16,444M in Q3 2026 alone, and SpaceX guided Q4 group capex to $35–40B.
- Colossus turbines and litigation. In July 2026, reports (TradingKey, IBTimes) said a lawsuit sought to shut down the gas turbines powering SpaceX's Colossus 2 data center, which those outlets said could imperil part of a reported $45B Anthropic compute contract; SpaceXChart found no court ruling or confirmed shutdown in that matter. Separately, in October 2026 a federal judge denied the DOJ's bid to dismiss an NAACP Clean Air Act citizen suit over turbines at the Memphis-area Colossus data centers, according to CNBC (Oct 20, 2026).
- Model competition. Anthropic, OpenAI, Google DeepMind, Meta, and the open-weight community all field competitive frontier models. xAI's differentiation rests partly on integration with X distribution and partly on model quality — neither is a settled moat.
- X platform revenue volatility. X advertising revenue has been volatile post-acquisition (2022 onwards). Brand-safety incidents, geographic exposure, and ad-format changes all move the line.
- Related-party valuation. The xAI / SpaceX merger was approved by both boards but Elon Musk holds significant ownership of both. The $250B xAI mark was set by a board-approved process but is not a public-market price.
- Orbital AI timeline. The orbital data center thesis is real but distant. Investors discounting that line to present value should apply meaningful execution discount.
- Regulatory. X faces content regulation in the EU (DSA), UK (OSA), India, Brazil, and elsewhere. xAI faces AI-specific regulation in the EU (AI Act) and emerging state-level frameworks in the U.S.
For broader risk framing see the risk factors page.