SpaceX is preparing to raise about $40 billion to fund a massive Nvidia chip purchase, and the size of the deal says a lot about where the AI infrastructure race is heading.
According to the report, the financing would be split between roughly $10 billion in bank loans and $30 billion in investment-grade debt, with Apollo Global Management expected to lead the deal. The transaction is expected to close in 2027.
On the surface, this looks like a huge chip order.
But for investors, there is a bigger story.
SpaceX is increasingly positioning itself not just as a rocket and satellite company, but as a major AI computing infrastructure player.
And Nvidia is sitting right in the middle of that strategy.
Why does SpaceX need so many Nvidia chips?
SpaceX’s AI ambitions are expanding rapidly.
The company is building large-scale computing infrastructure to support AI workloads, while also planning something far more ambitious: AI data centers in orbit.
Musk has said SpaceX plans to build its AI infrastructure exclusively around Nvidia technology, including its Vera Rubin architecture.
That gives Nvidia something extremely valuable: another enormous customer with plans to keep buying advanced AI hardware at scale.
For SpaceX, the logic is equally straightforward.
More GPUs mean more computing capacity.
More computing capacity means SpaceX can potentially sell that capacity to companies that desperately need access to AI infrastructure.
That creates a potential business model beyond simply spending billions on technology.
SpaceX is trying to turn compute into a business
This is perhaps the most important part of the story.
SpaceX isn’t only buying chips for its own AI ambitions.
It is increasingly looking at computing capacity as something it can rent out to other companies.
The company already has major agreements involving AI infrastructure.
According to the material, Google is expected to pay SpaceX around $920 million per month from October 2026 through June 2029 for access to Nvidia GPUs, CPUs, memory and related equipment.
Anthropic has also agreed to pay approximately $1.25 billion per month through May 2029 for access to computing capacity from SpaceX’s Colossus infrastructure.
That changes the economics of the chip spending.
Instead of simply asking, “Why is SpaceX spending so much on GPUs?”
The bigger question becomes:
How much revenue can SpaceX generate from those GPUs once they are operational?
The numbers are getting enormous
SpaceX’s computing ambitions are already operating at a massive scale.
The material says Colossus 2 could reach roughly 1.21 million processors by the end of 2026, while the broader Memphis infrastructure could reach around 1.44 million GPUs.
The company is also targeting more than 2 gigawatts of computing power by the end of 2026, with its 2027 target approaching 10 gigawatts.
That is an extraordinary amount of infrastructure.
And it explains why the company needs extraordinary amounts of capital.
The $40 billion financing would essentially help SpaceX accelerate this buildout without relying entirely on its existing cash flows.
But borrowing $40 billion is not without risk
This is where investors need to look beyond the headline.
SpaceX already has a significant capital spending requirement.
The company’s bonds sold off after its earlier debt issuance, reflecting investor concerns around rising leverage and heavy capital expenditure.
Its 2056 bonds were trading at around 85 cents on the dollar, with yields substantially above US Treasuries.
That tells us something important.
Investors are willing to finance SpaceX, but they are also demanding compensation for the risks involved.
The company has ambitious plans across rockets, satellites, Starlink and AI infrastructure.
Now it is adding another enormous investment program on top.
The AI opportunity is huge, but so is the capital requirement.
Why Apollo’s involvement matters
Apollo Global Management is expected to lead the financing.
That is significant because Apollo has built a major credit business around lending to large companies and structuring enormous financing transactions.
It has also been active in financing the broader semiconductor and AI infrastructure boom.
Apollo previously led a $35 billion financing deal involving Broadcom chips, showing how private capital is increasingly becoming part of the AI hardware funding machine.
This is an important trend for investors.
The AI infrastructure boom is no longer being funded only by technology companies and traditional banks.
Private credit, insurers, pension funds and other institutional investors are increasingly becoming part of the financing chain.
Nvidia gets another major demand driver
For Nvidia investors, the SpaceX deal is obviously significant.
A $40 billion financing package designed around Nvidia chips represents another enormous pool of potential demand.
And SpaceX is not approaching Nvidia as a casual customer.
Musk has publicly said the company plans to build exclusively on Nvidia because he believes its Vera Rubin architecture is the best option.
That kind of commitment matters.
Nvidia’s biggest question over the long term is not whether companies want AI chips today.
It is whether demand can remain strong as AI infrastructure spending moves into the next phase.
Deals like this provide another piece of evidence that the appetite for computing power remains extremely strong.
But there is another side to the Nvidia story
There is a concentration risk investors should not ignore.
SpaceX is becoming one of Nvidia’s major customers while also building its own AI infrastructure and potentially generating revenue by renting Nvidia-powered computing capacity to other companies.
That creates a powerful ecosystem.
But it also means the two businesses are becoming increasingly connected.
If SpaceX’s AI infrastructure takes longer to generate returns, the impact could spread across the ecosystem.
SpaceX could face pressure from its capital spending and debt commitments.
Nvidia could face weaker-than-expected demand from one of its major customers.
The bigger the AI infrastructure bet becomes, the more important execution becomes.
And then there is the orbital AI dream
The most futuristic part of the story is also one of the most interesting.
SpaceX plans to eventually put AI computing infrastructure into orbit.
The company has indicated that its first AI satellites could launch in the fourth quarter of 2027, with the broader orbital computing network expected to scale from there.
The idea is simple but ambitious.
Instead of building every data center on Earth, SpaceX wants to put computing infrastructure in space.
That could potentially offer advantages around energy, cooling and physical infrastructure.
But getting there requires successful launches, reliable hardware, enormous capital investment and a business model capable of generating returns.
The first real test comes when the satellites actually begin operating.
Starlink could be the financial foundation
There is one reason SpaceX can take on bets this large that many other companies cannot.
Starlink.
The satellite internet business has become an important revenue and earnings engine for SpaceX.
According to the material, Starlink generated around $11.39 billion in revenue and approximately $4.5 billion in segment operating income in 2025.
That cash-generating business gives SpaceX a financial base from which to pursue much more speculative and capital-intensive opportunities.
The strategy increasingly looks like this:
Starlink generates cash → SpaceX invests in infrastructure → AI capacity is rented to customers → computing revenue helps fund the next stage of expansion.
If that cycle works, SpaceX could create an entirely new growth engine.
The bigger AI infrastructure race
This isn’t just about SpaceX.
The deal shows how much capital the AI industry may require over the next several years.
AI companies need GPUs.
GPUs need data centers.
Data centers need power, cooling, networking and financing.
And all of that infrastructure needs to be built before the revenue from AI applications fully catches up.
That is why Wall Street is increasingly focused on the AI infrastructure economy, rather than just individual AI models.
The companies supplying the infrastructure could benefit from years of spending.
But the companies financing that infrastructure are also taking on increasingly large bets.
What should investors watch next?
For Nvidia investors, the key question is whether massive chip demand continues translating into sustainable revenue growth.
For SpaceX investors, the bigger question is whether the company’s AI infrastructure can turn enormous capital expenditure into equally enormous recurring revenue.
A few things will matter:
- How quickly SpaceX brings new Nvidia capacity online
- Whether Google, Anthropic and other customers continue expanding their commitments
- How much debt SpaceX ultimately carries
- Whether AI compute leasing becomes a major source of revenue
- Whether Starship can support the company’s orbital AI ambitions
- Whether orbital data centers move from concept to commercial reality
- Whether Nvidia maintains its dominance as competition from alternative AI chips grows
The real bet isn’t on chips. It’s on demand.
A $40 billion financing package sounds enormous.
But the real question isn’t simply whether SpaceX can buy $40 billion worth of Nvidia hardware.
It is whether there will be enough customers willing to pay for the computing power that hardware creates.
Right now, SpaceX appears to believe the answer is yes.
Google wants capacity.
Anthropic wants capacity.
AI companies around the world are competing for compute.
And Nvidia continues to sit at the center of that supply chain.
SpaceX is betting that this shortage of computing power will last long enough to justify one of the biggest infrastructure investments in the AI era.
For investors, that makes this much more than a chip purchase.
It is a giant bet on the future economics of AI compute.