Nvidia, Oracle, and Now SpaceX Are Borrowing Billions Amid the AI Boom. Is This a Warning Sign?

AI-related borrowing is accelerating, with Nvidia's $25 billion bond sale drawing more than $85 billion in orders, Oracle planning up to $50 billion of fundraising, and Amazon having borrowed over $80 billion since the start of 2025. The article argues that Nvidia, Amazon, and Alphabet can comfortably service debt from profits, while Oracle looks more stretched with about $100 billion of debt and roughly $24 billion of negative free cash flow in fiscal 2026. SpaceX is the biggest concern, as it may seek at least $20 billion in bonds despite losing nearly $5 billion in 2025 and $4.28 billion in Q1 2026.

Analysis

The market is effectively pricing AI capex like a utility buildout: lenders are willing to fund the balance sheets with the cleanest cash generators because the near-term credit story is stronger than the near-term equity story. That creates a second-order winner set in the bond complex first, then in the semiconductor and cloud supply chain as cheap debt extends the runway for hyperscaler spending and delays any capex air pocket. In other words, the credit market is not signaling caution yet; it is subsidizing the pace of the AI arms race.

The real signal is the widening quality gap among borrowers. For cash machines, debt simply increases optionality and likely pushes more orders toward GPUs, networking, and datacenter buildouts over the next 6-18 months. For leveraged builders with negative free cash flow, however, debt converts AI optimism into a refinancing problem: if utilization, monetization, or contract durability slips even modestly, equity holders become the residual shock absorber while bondholders start demanding tighter terms, higher coupons, or covenants.

The contrarian miss is that this is less about default risk today and more about duration risk in AI returns. The market is assuming that incremental infrastructure dollars will earn returns quickly enough to justify aggressive leverage, but the first failure mode is usually not bankruptcy — it is margin compression, lower ROIC, and de-rating of the most capital-intensive names. That argues for staying long the cash-rich platform winners while fading the most levered AI financing stories where refinancing is doing more work than operating cash flow.

Over the next 3-6 months, watch for any slowdown in bond oversubscription, widening spreads in lower-rated AI-linked issuers, or a miss on cloud capex efficiency metrics; those would be the first signs the financing window is tightening. Until then, the better trade is to own the names that can self-fund AI expansion and avoid those that need perpetual market access to keep the story intact.

Key Decisions for Investors

• Long NVDA / short ORCL as a 3-6 month pair: NVDA benefits from cash-funded optionality and balance-sheet flexibility, while ORCL is more exposed to capex intensity and refinancing sensitivity; target a multiple re-rating gap if credit spreads begin to discriminate between self-funded and levered AI spend.
• Add AMZN and GOOGL on pullbacks over the next 1-3 months: both have the operating cash flow to absorb AI infrastructure spend without stressing the balance sheet, and they are better positioned to capture any incremental AI demand if financing becomes less accessible for smaller players.
• Avoid owning leveraged private-market AI financing stories outright; if exposure is required, prefer deep out-of-the-money call structures rather than common equity to limit downside if refinancing markets tighten over the next 6-12 months.
• For bond portfolios, favor higher-quality AI-related issuers and avoid lower-rated new issues in the space until spreads cheapen meaningfully; the current primary market is still pricing stability, but the first crack will show up in secondary performance before any equity weakness.
• Monitor ORCL credit default swap spreads and new-issue concession as a tactical risk signal; if either widens sharply, trim any pro-growth AI basket and rotate toward cash-generative semis and hyperscalers.