Finance & Banking | 2 min read

SpaceX Sells $32 Billion in Bonds to Finance Nvidia AI Chip Purchases — Setting a New Template

SpaceX priced $32 billion in bonds through an off-balance-sheet SPV to fund $47 billion in Nvidia Rubin GPU purchases — a structure Goldman Sachs says could become the AI industry financing template.

Hector Herrera
Hector Herrera
A financial trading floor featuring chips, vehicle, related to SpaceX Sells $32 Billion in Bonds to Finance a chip manufact
Why this matters SpaceX priced $32 billion in bonds through an off-balance-sheet SPV to fund $47 billion in Nvidia Rubin GPU purchases — a structure Goldman Sachs says could become the AI industry financing template.

SpaceX has found a way to finance $47 billion in Nvidia GPU purchases without putting a dollar of debt on its own balance sheet — and the structure may become the standard playbook for how AI companies pay for the chips they need.

Goldman Sachs and Apollo arranged a $32 billion bond offering through a special-purpose vehicle called Saturn Funding Trust, with Class A3 notes priced at an 8.26% yield — roughly two percentage points above comparable investment-grade corporate bonds. The SPV (a legally separate entity created solely to hold assets and issue debt) buys the Nvidia Rubin GPUs, owns them outright, and leases or deploys them back to SpaceX's Starlink and AI operations. SpaceX gets the computing power; the debt sits in Saturn Funding Trust, not on SpaceX's books.

Why This Structure Exists

Large GPU purchases create an unusual financing problem. The chips are expensive — Nvidia's Rubin architecture units run well above $100,000 per unit at scale — but they depreciate quickly as newer generations arrive, and they don't serve as clean collateral the way real estate or aircraft do. Traditional corporate bonds work for stable, cash-generating businesses; they're less natural for a company buying assets that will need to be replaced in three to five years.

The off-balance-sheet SPV model solves multiple problems at once:

  • Debt isolation. SpaceX's corporate credit rating and borrowing capacity remain unaffected.
  • Asset-backed structure. Bond investors hold a claim on the GPUs themselves, not just SpaceX's overall creditworthiness.
  • Tax efficiency. Depending on jurisdiction, the SPV can capture depreciation benefits that might not be fully available to SpaceX directly.

The 8.26% yield reflects the premium investors demand for taking on that asset-depreciation risk — the bonds pay more than investment-grade paper precisely because Rubin GPUs are worth less each year.

The Template Argument

Market observers quoted in Bloomberg's coverage describe this as a potential template for AI infrastructure financing broadly. Hyperscalers — Amazon Web Services, Microsoft Azure, Google Cloud — have so far absorbed GPU costs through operating budgets or standard corporate debt. But as AI chip procurement scales into the hundreds of billions of dollars annually, the balance-sheet constraints become real even for trillion-dollar companies.

An SPV model lets any large buyer of AI hardware:

  1. Separate chip financing from core business operations
  2. Attract a different class of institutional investor (asset-backed buyers vs. corporate bond buyers)
  3. Scale purchases beyond what traditional borrowing capacity would allow

What to Watch

Whether Saturn Funding Trust performs as designed depends on two things: the ongoing revenue Nvidia GPUs generate (Starlink's AI processing workloads are substantial but not unlimited) and whether Nvidia's current SEC accounting investigation affects GPU valuations and demand enough to pressure the collateral assumptions baked into the bond structure. If the deal performs well, expect competitors to replicate the Saturn model within 12 to 18 months.

Key Takeaways

  • ✓ Saturn Funding Trust
  • ✓ off-balance-sheet SPV model
  • ✓ Asset-backed structure.
  • ✓ potential template for AI infrastructure financing

Did this help you understand AI better?

Your feedback helps us write more useful content.

Hector Herrera

Written by

Hector Herrera

Hector Herrera is an AI systems architect in Houston and founder of Hex AI Systems. He designs and runs AI systems in production and writes daily about how AI is reshaping business, government and everyday life. 20+ years building for the web. Houston, TX.

More from Hector →

Get tomorrow's AI briefing

Join readers who start their day with NexChron. Free, daily, no spam.

More from NexChron