Why Is SpaceX Borrowing $40 Billion for Nvidia AI Chips Despite Having $100 Billion in Cash?

In mid-2026, SpaceX ended its first half with roughly $100 billion in cash after raising about $86 billion in its June IPO. Yet reports soon emerged that the company was seeking another $40 billion in loans and bonds, mainly to buy Nvidia AI chips.

The numbers appear contradictory: why borrow when cash reserves look so strong? The reason is that SpaceX is spending so fast that it could quickly erode even this large cushion if it wants to keep other parts of the business liquid and flexible.

This article will delve into the drivers behind the financing, the chips it will fund, the advantages of debt over cash, and the risks involved.

SpaceX became a public company in June 2026 with a record-setting IPO that brought in roughly $85.7 billion to $86 billion after the underwriters exercised their overallotment option. Shortly afterward, the company disclosed cash and equivalents of about $100 billion to $101 billion. At the same time, it already carried around $39 billion to $40 billion in debt and finance leases, much of it from a $25 billion investment-grade bond sale completed less than two weeks after the IPO. That earlier bond deal largely refinanced a bridge loan tied to the February 2026 absorption of Elon Musk’s xAI into SpaceX.

The new financing talks, first reported by the Financial Times and later confirmed in outline by other outlets citing people familiar with the discussions, center on a package of about $10 billion in bank loans and $30 billion in investment-grade debt. Apollo Global Management is expected to lead the effort, with firms such as PIMCO among those said to be looking at participation. If completed, the transaction is targeted for 2027. The stated purpose is to fund a large order of Nvidia AI chips.

Capital spending numbers show why the cash pile is under pressure. In the second quarter of 2026 alone, SpaceX recorded $18.4 billion in capital expenditures, roughly $15.8 billion of which went to AI infrastructure. In the first half of the year, total capital spending reached about $28.5 billion, while operating cash generation was much lower, around $3.5 billion. 

Management has signaled that the pace of the AI build-out will remain elevated. Musk has publicly discussed expanding computing capacity from roughly 1.4 gigawatts to nearly 10 gigawatts by the end of 2027. At that scale, sustained quarterly infrastructure spending in the high teens of billions of dollars would quickly consume large portions of the existing cash reserves if no external financing were arranged.

Other commitments add to the picture. SpaceX has indicated plans involving EchoStar spectrum that could require several billion dollars to clear related debt, with closing expected in late 2027. Aerospace projects and Starlink expansion continue to demand heavy investment. In that environment, treating the $100 billion as a permanent rainy-day fund rather than a spending account becomes a deliberate choice.

The Shift After Folding in xAI

The February 2026 decision to fold xAI into SpaceX fundamentally changed the company’s capital profile. What had been a rocket-and-satellite business with growing Starlink revenue suddenly included a large, fast-growing AI division that requires continuous purchases of high-end GPUs. Before the combination, capital spending was already heavy because of launch vehicles and satellite deployment.

After the integration, the AI side added a new and steeper demand curve for specialized hardware. SpaceX has said it intends to build its AI clusters exclusively on Nvidia hardware, citing performance reasons. That choice locks the company into a steady stream of large orders rather than spreading purchases across multiple suppliers.

In September 2026, Musk said the Colossus 2 facility was already running hundreds of thousands of Nvidia chips, with plans to more than double that count by year-end in some configurations. Those statements moved the discussion from general ambition to concrete near-term targets. The facility’s existing scale already puts SpaceX among the largest private AI compute operators, and the planned expansion keeps pressure on for additional deliveries through 2026 and into 2027.

Turning Capacity Into Revenue

This is not abstract planning. Reports indicate that SpaceX is already generating revenue by leasing compute capacity to other AI companies, including multi-hundred-million-dollar monthly contracts. Customers pay for access to the GPUs rather than building their own clusters from scratch. To keep those customers and attract more, the company needs to stay ahead on capacity. If utilization rises and waiting lists form, competitors can step in. 

Each additional gigawatt of AI compute can require tens of billions of dollars in hardware, power, cooling, and related infrastructure. Scaling from roughly 1.4 gigawatts to the multi-gigawatt targets Musk has outlined, nearly 10 gigawatts by the end of 2027, implies capital outlays that dwarf even the large cash balance built from the IPO.

The math is straightforward. Sustaining quarterly infrastructure spending in the mid-to-high teens of billions of dollars quickly consumes cash if the company pays for every purchase outright. The AI division’s growth therefore creates a structural need for external financing even when the overall balance sheet looks strong.

Timing the Debt to Match the Build-Out

The timing of the proposed debt also matters. A financing that closes in 2027 would arrive as the heaviest bills for the next phase of chip deliveries and data-center build-out come due. Using debt for hardware purchases lets existing cash cover near-term operations, aerospace programs, and any spectrum-related payments without forcing the company to liquidate marketable securities or reduce its buffer. 

In practical terms, the $100 billion cash position functions more like a strategic reserve than a spending account. Drawing it down too quickly for one investment category would leave less flexibility for rockets, Starlink constellation growth, or unexpected opportunities.

The Intertwined Supplier Relationship

Market observers have noted that Nvidia holds a sizable stake in SpaceX, valued at around $21 billion at recent prices. That ownership doesn't change the economic need for chips, but it shows how intertwined supplier-customer relationships have become in the current AI infrastructure cycle. 

Nvidia benefits from large, predictable orders; SpaceX benefits from preferred access and deep technical alignment. The arrangement works as long as the compute capacity generates returns, whether through internal model development or leased services to outside customers

Taken together, the xAI integration, the exclusive Nvidia strategy, the revenue already coming from leased capacity, and the multi-gigawatt expansion targets explain why SpaceX is looking beyond its cash pile. The reported $40 billion package is less a sign of shortage and more a tool to match the pace of hardware purchases to business growth while protecting liquidity for the rest of the company.

Several practical reasons favor debt over simply writing checks from the $100 billion pile. The choice reflects careful capital allocation rather than any shortage of funds. By using borrowed money for the large Nvidia order, SpaceX can protect its existing liquidity while still moving forward with the AI build-out at the pace it has set.

Liquidity Preservation

Capital spending at the current rate could exceed $70 billion annually when AI infrastructure is included. If SpaceX paid for the next large Nvidia order entirely in cash, its buffer for other businesses and contingencies would shrink quickly. Borrowing keeps that buffer intact. 

The company has signaled a desire to maintain a minimum cash level and has access to investment-grade markets, so the cost of debt remains relatively manageable compared with the opportunity cost of depleting liquid assets. In practice, this means the $100 billion continues to serve as a strategic reserve rather than being drawn down for a single category of spending. That reserve can cover day-to-day operations, unexpected costs, or shorter-term needs without forcing the sale of marketable securities at potentially inconvenient times.

Access to Institutional Capital

SpaceX carries a BBB-range investment-grade rating (S&P BBB, Moody’s Baa1, Fitch BBB+). That rating opens the door to insurance companies, pension funds, and other large fixed-income buyers who cannot or will not buy high-yield debt. The structure of roughly $10 billion in bank loans plus $30 billion in investment-grade bonds is designed to tap those pools at scale. 

Earlier in 2026, the company raised $25 billion in bonds with strong initial demand, showing that the market was willing to lend at those levels. Reaching the same investor base again lets SpaceX raise a large sum in one coordinated transaction instead of relying solely on internal cash or smaller, more expensive funding sources.

Matching the Useful Life of the Assets

High-end GPUs depreciate over several years. Spreading the cost over longer-dated bonds can better align financing with the period when the chips generate revenue through model training, inference services, or leased capacity. Paying cash upfront concentrates the cash outflow in a short window.

By contrast, debt spreads the financial impact across the same multi-year window in which the hardware is expected to produce returns. This matching reduces the risk of a sudden, large reduction in available cash at a moment when other parts of the business also need funding.

Flexibility for Other Capital Projects

Rockets, Starship development, Starlink constellation growth, and potential spectrum acquisitions all require cash. Keeping the IPO proceeds available for those uses while financing the AI hardware separately is a form of capital allocation discipline.

Analysts have pointed out that the reported $40 billion package would roughly double the existing debt load, yet the company’s equity market value remains in the multi-trillion-dollar range, giving it substantial overall capacity. The separation of funding sources therefore lets management pursue growth in both the aerospace and AI businesses without forcing trade-offs that would slow one side to accelerate the other.

In short, the decision is less about needing money to stay solvent and more about choosing the lowest-friction way to fund a very large, discrete hardware order while protecting the balance sheet for the rest of the business.

The plan is not without friction. After the $25 billion bond sale in June, secondary trading showed some selling pressure as investors digested the combination of rising debt and heavy spending. Credit-default-swap spreads on SpaceX debt widened after the latest reports, and the company’s bonds weakened in the secondary market. Those moves signal that lenders are watching the pace of AI capital expenditure carefully.

Valuation adds another layer of pressure. SpaceX has traded at high multiples of forward sales, around 50 times, meaning investors are already pricing in substantial growth. If the new chips do not translate into revenue and cash flow quickly enough, the multiple could compress even if absolute sales rise. Capital spending equal to a large percentage of revenue (figures around 180 percent of sales have been reported in recent periods) raises the bar for returns on those investments.

There are also structural risks. Nvidia chips lose value over time as newer generations appear. Long-maturity bonds used to finance them could outlast the collateral's economic life if the GPUs themselves are pledged. Some market commentary has highlighted the circular nature of current AI financing, in which suppliers help arrange capital for customers who then buy more of the suppliers’ products. While that dynamic is not unique to SpaceX, it increases the importance of independent demand for the compute capacity being built.

On the positive side, short interest in the stock has remained relatively low, and many hedge funds held positions after the second quarter. Lenders appear willing to engage in discussions, suggesting that the investment-grade rating and the company’s overall scale still command attention. Management has historically been willing to adjust spending if capital-market conditions tighten, which provides some downside protection for credit investors.

Investors and observers will watch several markers in the coming quarters: the final terms and size of any completed financing, the actual delivery and utilization rates of the new chips, the trajectory of free cash flow, and any updates on spectrum or other large non-AI capital needs. The difference between a successful scale-up and a prolonged cash burn will depend heavily on how quickly the AI infrastructure begins generating durable revenue.

SpaceX’s reported plan to borrow roughly $40 billion for Nvidia AI chips, despite sitting on about $100 billion in cash, is best understood as a deliberate capital-structure choice rather than a sign of distress. The company is scaling AI compute at a rate that would rapidly draw down even a large cash reserve if paid for entirely out of pocket. By using a mix of bank loans and investment-grade bonds, SpaceX can keep its existing liquidity available for rockets, satellites, spectrum, and day-to-day operations while still securing the hardware it believes is necessary to compete in AI.

The move fits a broader pattern in which the largest technology and infrastructure builders are increasingly turning to debt markets to fund multi-year capital programs. Whether the chips deliver the growth current equity valuations assume remains an open question, and the bond market’s earlier reaction shows lenders aren't writing blank checks. Still, access to investment-grade capital, a multi-trillion-dollar equity valuation, and clear demand for AI capacity give SpaceX room to execute the strategy.

Readers will want to track the final shape of any financing package, the pace of capacity additions, and how that capacity converts into revenue. Those data points will ultimately decide whether the decision to borrow looks prescient or overly aggressive.

Why doesn’t SpaceX just spend its $100 billion cash on the chips?

Spending the cash would leave less flexibility for aerospace programs, Starlink growth, spectrum-related payments, and unexpected needs. Debt lets the company keep a large liquid buffer while still funding the hardware.

What is the expected structure of the $40 billion package?

Reports describe roughly $10 billion in bank loans and $30 billion in investment-grade bonds, with Apollo Global Management expected to lead and a target close in 2027.

How much debt does SpaceX already have?

At the end of June 2026, the company carried approximately $39 billion to $40 billion in debt and finance leases, much of it from the $25 billion bond issue completed after the IPO.

What credit rating does SpaceX hold?

It carries investment-grade ratings in the BBB range (S&P BBB, Moody’s Baa1, Fitch BBB+), which allows broader institutional participation in its debt.

How large is the planned AI compute expansion?

Public comments have pointed to growth from roughly 1.4 gigawatts of capacity toward nearly 10 gigawatts by the end of 2027.

Is Nvidia involved beyond supplying the chips?

Nvidia holds a significant equity stake in SpaceX, valued at around $21 billion in recent periods, and SpaceX has stated a preference for Nvidia hardware in its AI clusters.

Has the bond market reacted positively so far?

The earlier $25 billion issuance saw strong initial demand, but secondary trading later showed some pressure as investors focused on spending pace and rising leverage. Spreads widened after the latest reports.

What other large cash needs does SpaceX face?

Beyond AI infrastructure, the company has pointed to potential spectrum-related payments in the range of several billion dollars and ongoing heavy investment in launch vehicles and satellite networks.

The reported financing, if completed, will be one of the larger single debt packages tied to AI hardware. Its success will depend less on today's cash pile and more on how effectively the new chips generate returns in the years ahead.

Disclaimer: This article is for informational and educational purposes only. It does not constitute financial, investment, or legal advice. Cryptocurrency investments carry significant risk and high volatility. Always conduct your own research (DYOR) and consult a qualified financial advisor before making any investment decisions. Past performance is not indicative of future results.