SpaceX Reportedly Wants to Borrow $40 Billion for Nvidia Chips Despite a $100 Billion Cash Pile

SpaceX is reportedly seeking $10 billion in bank loans and $30 billion in investment-grade bonds to buy Nvidia chips, potentially lifting debt and finance leases from $39.5 billion to about $80 billion. The company reported $629 million in Q2 interest expense, while estimated interest on the proposed $40 billion borrowing at the June bonds’ 5.855% weighted-average rate would be about $2.3 billion annually. Q2 AI operating losses of $1.3 billion contrast with $1.7 billion of operating income from its connectivity segment; shares were down about 2% Wednesday.

Analysis

Funding the AI build raises the hurdle for equity

The key change is not simply more chip demand; it is a tighter link between Starlink’s cash generation and the return on SpaceX’s AI investment. If AI capacity ramps faster than monetization, connectivity profits may service debt incurred for a business still consuming cash. That makes the downside more asymmetric: capex delays or weaker AI utilization can pressure equity even if Starlink remains operationally healthy.

The financing could still be less dilutive than issuing shares, and the reported 2027 close leaves time for operating results to develop. But the size and timing create execution risk: debt is raised ahead of proof that the added compute earns adequate returns, while the EchoStar spectrum transaction may add another funding claim. The deal is only reported, and recourse, covenants, rating treatment, pricing and drawdown schedule are not specified; those terms matter more than the headline amount.

NVDA could gain near-term demand visibility, but a large customer’s ability to fund purchases is not equivalent to durable end-user demand. Watch for delivery timing or order changes if financing slips. ECHO may benefit from the contemplated debt payoff, but that is contingent on the spectrum transaction closing. APO’s role could create fee income, though the economics are unknown and not a standalone thesis.

The contrarian point: avoiding dilution is shareholder-friendly only if returns on the AI build exceed the all-in cost of debt. The market may be over-weighting financing availability and under-weighting the risk that Starlink subsidizes AI longer than expected. Near term, the reported financing is not yet a confirmed catalyst; over 1–3 months, terms and credit-market reception matter. Over 6–18 months, AI-segment profitability and capex productivity are decisive.

Key Decisions for Investors

• Keep SPCX underweight / avoid adding on financing headlines alone. Consider a defined-risk bearish position only after confirming the debt is committed and reviewing price action; the thesis is falsified by sustained AI-segment operating profitability alongside capex discipline.
• Before taking a relative long-NVDA/short-SPCX position, verify that the financing supports actual chip deliveries and that NVDA demand is not materially dependent on this transaction. Treat it as a watch item, not a clean pair trade, until order visibility is clearer.
• Track the debt term sheet and rating implications: recourse, collateral, covenants, coupon, maturity and whether the reported investment-grade structure is maintained. A materially higher cost or restrictive terms would worsen equity risk; favorable terms and no rating pressure would reduce it.
• Monitor quarterly AI operating results against the ongoing compute build, and Starlink/connectivity operating income and cash generation. If AI losses narrow and connectivity cash generation remains resilient, the cross-subsidy concern weakens; worsening losses or falling connectivity contribution would strengthen the bearish thesis.
• Do not trade ECHO or APO solely on this report. Reassess ECHO if the spectrum closing or debt-payoff terms change, and APO only when its role and economics are disclosed.