SpaceX Plans to Take on $40 Billion in New Debt. Here's Why Investors Should Get Used to This

SpaceX reported $15.8 billion of AI capex in the quarter ended June 30, up from $749 million a year earlier, and is reportedly in talks to raise $40 billion in debt to buy Nvidia GPUs. JPMorgan analyst Doug Anmuth projects capex of about $200 billion in each of 2027 and 2028; the article warns that debt financing could raise interest costs, while stock issuance could dilute shareholders. It also highlights SpaceX’s roughly $2.2 trillion market capitalization and lack of profitability as risks.

Analysis

The key distinction is between a potential GPU purchase and durable, profitable demand. For NVDA, a large prospective buyer can support backlog expectations, but the signal is only as strong as funded orders, delivery timing, and recognized revenue; debt discussions are not equivalent to committed spend. If SpaceX’s buildout proceeds, demand could also reach power, networking, and cooling suppliers, but those spillovers depend on facilities being built and energized—not merely chips being ordered.

For SPCX, the risk is financing structure and returns on capital, not capex in isolation. Debt would add fixed claims against uncertain AI monetization; equity financing would dilute holders. A further risk is opportunity cost: accelerated AI investment may compete with other capital needs, while an AI addressable-market estimate does not establish an achievable share or attractive unit economics. The cited $200 billion annual capex path is an analyst projection, not company guidance.

Near term, financing terms and order confirmation matter more than the long-run AI narrative. Over 1–3 months, verify debt proceeds, GPU delivery commitments, and any company update to capex or cash-flow expectations. Over 6–18 months, the thesis turns on utilization and monetization relative to invested capital. The contrarian angle is that NVDA demand may be overread from a single prospective customer, while SPCX downside may be underappreciated if investors capitalize AI ambition before returns are demonstrated. Falsifiers: canceled or delayed orders weaken the NVDA read-through; evidence of funded deployment and improving monetization weakens the SPCX risk case.

Key Decisions for Investors

• Treat the reported GPU demand as a watch item for NVDA, not yet a standalone long catalyst. Seek confirmation of funded orders, delivery schedules, and revenue conversion before increasing exposure.
• For SPCX, prefer a cautious or underweight stance until financing terms and expected returns on AI investment are clearer. Reassess if deployment is funded without materially worsening leverage or dilution, and if monetization evidence improves.
• A relative-value expression—short SPCX against long NVDA—could isolate financing and execution risk from AI infrastructure demand, but size it conservatively: NVDA is also exposed to order concentration and execution, and the spread can widen if SpaceX’s spending accelerates. Confirm borrow, liquidity, and valuation first.
• Track company guidance, financing disclosures, GPU order or delivery confirmation, and cash-flow/utilization evidence over the next 1–3 months; absent those data, avoid treating analyst capex projections as a base-case forecast.