SpaceX: 10 GW Could Change The Entire Story

SpaceX ended Q2 with 1.4 GW of compute capacity, and management believes it could approach 10 GW during 2027. Q2 AI revenue was $2.6 billion, while new hosting agreements contributed $1.6 billion and AI capex reached $15.8 billion. Starlink generated $11.39 billion of 2025 revenue and $7.17 billion in adjusted EBITDA, providing a profitable foundation for the expansion.

Analysis

The key issue is conversion, not the headline capacity target. A 10 GW ambition makes firm power, grid interconnection, cooling and networking potential deployment bottlenecks; GPU availability alone will not determine the ramp. That shifts prospective value toward power and data-center infrastructure providers, but only if SpaceX’s buildout produces incremental, bankable orders rather than internal plans. Do not assume a direct revenue benefit to any supplier before order evidence.

For SPCX, Starlink’s operating cash generation may provide strategic funding flexibility, but EBITDA is not free cash flow: the article gives no consolidated liquidity, financing, or competing-capex picture. The reported AI capex therefore raises the hurdle for proving attractive returns. Hosting agreements also need diligence: verify whether the cited contribution is recognized revenue, committed backlog, or another measure, and assess customer concentration and cancellation terms.

Near term, growth-oriented sentiment may support SPCX, but the 2027 target is a multi-quarter execution story. Over 1–3 months, focus on delivered capacity, utilization, recognized AI revenue, and funding. Over 6–18 months, power access and capital intensity become the structural tests. The contrarian risk is that markets value planned gigawatts as productive capacity before utilization and returns are demonstrated; the upside alternative is that existing Starlink economics let SpaceX finance infrastructure competitors cannot match. The thesis weakens if deployment slips, AI revenue fails to convert to cash, or capex rises without corresponding contracted demand.

Key Decisions for Investors

• Do not chase SPCX solely on the capacity ambition. Treat it as a conditional growth thesis until filings or earnings disclosures reconcile AI revenue, hosting-agreement economics, capex, liquidity, and utilization.
• Set a 1–3 month catalyst watch for quarterly evidence of capacity actually energized and monetized, plus the definition and durability of hosting commitments. Upgrade only if deployment and recognized revenue progress together; downgrade on slippage or rising spend without demand conversion.
• For a second-order expression, monitor grid, power-generation, electrical-equipment, and data-center infrastructure names for confirmed incremental orders; avoid assuming beneficiaries from the capacity target alone. A broad infrastructure exposure is preferable to a single-supplier bet until SpaceX identifies procurement and timing.
• Falsifiers: material delay to the 2027 buildout, falling utilization or weak cash conversion despite reported AI revenue, hosting commitments that are cancellable or not backlog, or evidence that consolidated funding needs constrain Starlink investment.