Estimated reading time: 6 minutes · Last updated:
Bloomberg reported that Space Exploration Technologies (NASDAQ: SPCX) is seeking about $40 billion in financing to buy AI processors from Nvidia (NASDAQ: NVDA). The package market participants have discussed would pair $10 billion of bank loans with $30 billion of investment-grade debt and — if completed — is not expected to close until 2027. SpaceX's rapid push into compute leasing, large monthly customer contracts and heavy second-quarter capex underpin the financing need. Some of the figures and customer details appearing here were first reported by The Motley Fool; the specific numbers were assembled from company disclosures and regulatory filings.
Key takeaways
- Reported financing: Bloomberg reported SpaceX is seeking $40 billion in financing to buy Nvidia AI chips, including $10 billion of bank loans and $30 billion of investment-grade debt.
- SpaceX scale and cost: SpaceX expects cumulative compute capacity near 10 gigawatts by the end of 2027; it spent $18.4 billion on capex in the second quarter and reported a negative $25 billion cash-flow result through the first six months of the year.
- Nvidia exposure: Nvidia reported $89 billion of data center revenue in the second quarter; hyperscalers accounted for $48.7 billion of that segment.
Table of contents
Why $40 billion would matter for Nvidia
A financing that large — if used to buy Nvidia hardware — would deepen an already tight commercial relationship between one of the largest AI infrastructure buyers and the company that supplies much of the hardware. Nvidia posted $89 billion of data center revenue in the second quarter, up 117% year over year, and the hyperscalers contributed $48.7 billion of that segment. That concentration shows how much of Nvidia’s data-center growth has come from a relatively small set of big customers.
By publicly committing to Nvidia’s Vera Rubin architecture, SpaceX aligns its expansion with Nvidia’s product roadmap. Elon Musk said SpaceX has decided to build "exclusively" on Nvidia because he believes the Vera Rubin architecture is the best AI computer. If SpaceX scales as it plans, additional unit purchases by SpaceX would add to the same demand pool that hyperscalers already represent.
Market-level activity also underscores the scale: hyperscalers issued roughly $194 billion of bonds through early July to finance infrastructure, and Goldman Sachs Asset Management has suggested those companies could issue another $420 billion of debt in 2027 to support build-outs. A SpaceX purchase financed by debt would therefore fit a broader trend of large tech firms tapping bond markets to fund AI compute.
How big SpaceX’s compute and cash needs already are
SpaceX is building a compute-leasing business at scale. Reported contracts indicate Anthropic pays about $1.25 billion per month for roughly 325,000 Nvidia GPUs, Google contracted near $920 million per month for about 110,000 GPUs, and Reflection AI is tied to a $150 million-per-month agreement that references Nvidia’s GB300 chips inside SpaceX’s Colossus 2 data center. SpaceX also disclosed $6.7 billion of additional cloud-services revenue in the opening weeks of the third quarter, and an unnamed hosting customer is expected to generate about $1.1 billion of monthly revenue beginning in December.
Those customer commitments help explain why SpaceX’s capital needs are so large. The company recorded $18.4 billion of capital expenditures in the second quarter and produced a negative $25 billion cash-flow result through the first six months of the year while funding Starlink, Starship and AI data-center investments. Scaling toward the stated target of roughly 10 gigawatts by the end of 2027 will require more hardware purchases, expanded power capacity and additional data-center build-out.
Bloomberg’s report that Apollo Global Management is rumored to be leading a financing package that would include $10 billion of bank loans and $30 billion of investment-grade debt explains one plausible structure for how SpaceX might afford a multi‑billion-dollar hardware purchase without fully draining operating liquidity. The transaction, however, remained unconfirmed by SpaceX, Nvidia or Apollo and was not expected to close until 2027.
Investor implications without giving advice
For Nvidia shareholders, the apparent logic is straightforward: a large, debt‑funded buyer that commits to a single vendor increases medium‑term demand for that vendor’s chips. Nvidia’s second-quarter data center revenue and the scale of hyperscaler contracts show that a large share of its growth comes from major infrastructure buyers.
That said, nothing in the available reporting guarantees a financing package will close or that every purchase will flow to Nvidia. Supply constraints, architecture shifts, contractual churn and the outcome of any competitive procurement process can alter the eventual hardware mix. Investors should treat the $40 billion figure as a reported financing target, not a completed sale.
To make the scale concrete: hyperscaler-related revenue of $48.7 billion accounted for a substantial portion of the $89 billion data center quarter; using those two inputs implies hyperscalers represented roughly 54.7% of Nvidia’s data center revenue in Q2 (derived from $48.7 billion divided by $89 billion). That share helps explain why a major new buyer would be meaningful to Nvidia’s top line.
| Customer | Monthly payment | Approximate Nvidia GPUs |
|---|---|---|
| Anthropic | $1.25 billion | 325,000 |
| $920 million | 110,000 | |
| Reflection AI | $150 million | GB300-access (Colossus 2) |
| Unnamed hosting customer | $1.1 billion (beginning December) | Not specified |
Cases for and against a positive impact on Nvidia
The case for
- Sustained hyperscaler and large-customer demand would keep Nvidia pricing power and utilization high, supporting continued growth in data center revenue.
- A long-term vendor commitment from a deep-pocket buyer such as SpaceX would create repeat orders and scale benefits for Nvidia’s Vera Rubin architecture.
The case against
- The financing has not closed; if it fails, anticipated Nvidia orders tied to the package may not materialize.
- Supply constraints, competitor product advances or a shift away from Nvidia’s architecture by any major buyer could reduce the expected uplift to Nvidia’s revenue.
What to be careful about
- The $40 billion package is reported and unconfirmed; terms may change or the deal may not close in 2027.
- SpaceX’s simultaneous investments in Starlink and Starship could compete for cash and delay or reduce planned hardware purchases.
- A concentration of demand among a few buyers exposes Nvidia to order volatility if even one large contract changes.
- Debt-funded expansion increases leverage on SpaceX’s balance sheet and could make timing and scale of purchases conditional on market conditions.
The bottom line
The reported $40 billion financing request frames SpaceX as a potential very large buyer of Nvidia hardware and fits a broader pattern of debt-funded AI infrastructure expansion among major tech players. Bloomberg’s scoop and the customer and cash figures disclosed elsewhere show why such a purchase would matter to Nvidia’s data-center business. That said, the package was reported and unconfirmed; timing and terms remain subject to negotiation. Investors should treat the numbers as reported inputs to a conditional scenario rather than as a guaranteed revenue stream for Nvidia.
What to watch
- Watch for confirmation that a financing package has closed or advanced in 2027, including any statements from SpaceX, Nvidia or Apollo Global Management; the reporting set expectations for a 2027 close.
- Watch SpaceX’s progress toward cumulative compute capacity of about 10 gigawatts by the end of 2027; the company has made that target public.
- Watch for the unnamed hosting customer to begin the monthly revenue stream the company disclosed as beginning in December; reporting cited a $1.1 billion-per-month expectation.
Frequently asked questions
What did Bloomberg report about SpaceX’s financing?
Bloomberg reported Space Exploration Technologies is seeking about $40 billion in financing to purchase Nvidia AI chips, with a rumored structure of $10 billion of bank loans plus $30 billion of investment-grade debt and a likely close in 2027.
How big are the customer contracts that run through SpaceX’s data centers?
Published figures show Anthropic paying $1.25 billion per month for access to roughly 325,000 Nvidia GPUs, Google paying $920 million per month for about 110,000 GPUs, and a Reflection AI arrangement tied to $150 million per month in GB300 access.
How exposed is Nvidia to large infrastructure buyers?
Nvidia reported $89 billion of data center revenue in Q2; hyperscalers contributed $48.7 billion of that total, which implies roughly 54.7% of the data center quarter was driven by hyperscaler-related sales.
Why would SpaceX borrow rather than pay from cash?
SpaceX recorded $18.4 billion of capex in the second quarter and reported a negative $25 billion cash-flow result through the first six months, and the reported borrowing trend among hyperscalers — about $194 billion issued through early July — indicates firms are using debt to accelerate AI infrastructure build-outs.
Related reading
This article is information, not financial advice. Anyone acting on it should do their own checks.
