SpaceX’s talks to borrow $40 billion to buy Nvidia chips lay bare a dangerous loop at the heart of the AI boom, with the industry’s biggest supplier bankrolling the very customers who buy its products and global investors left carrying the risk, warns the CEO of deVere Group, one of the world’s largest independent financial advisory organisations.
The comments from Nigel Green come as Elon Musk’s rocket and AI conglomerate reportedly seeks around $10 billion in bank loans and $30 billion in investment-grade debt, in what would rank among the biggest debt financings of the entire AI buildout, while a separate Wall Street syndicate gathers a further $60 billion of AI chip financing.
He says: “Follow the money here and it comes straight back to where it started. Nvidia owns a sizeable stake in SpaceX.
“SpaceX has committed to building exclusively on Nvidia’s chips. And Nvidia is helping to assemble the financing pools that lend its customers the cash to pay for them.
“Every dollar in that loop gets counted as growth somewhere. Once you strip out the circular flows, the true level of independent demand becomes far harder to measure.”
Filings show Nvidia holds just under 123 million SpaceX shares, worth close to $21 billion at the end of June.
Since 2024 the chipmaker has pledged more than $100 billion to AI companies, and it’s pulling together a Wall Street consortium to provide over $500 billion of financing for its own customers, partially guaranteeing loans backed by the value of its chips.
SpaceX, meanwhile, lost $4.28 billion in the first quarter of this year on revenue of $4.69 billion.
“When a supplier guarantees the loans its customers use to buy its products, and those loans are secured against the same products, everything rests on chip values holding up,” notes the deVere CEO.
“If the next generation of hardware makes today’s chips obsolete faster than expected, the collateral, the earnings and the equity stakes all lose value at the same time.”
Big tech is on course to spend roughly $700 billion on capital projects this year, with around three quarters going into AI infrastructure.
Much of that hardware is being depreciated over five to six years, while critics argue its genuine economic life is closer to two or three. One estimate puts the resulting understatement of costs at around $176 billion between 2026 and 2028.
“Stretching depreciation flatters today’s profits by pushing the costs into future years,” explains Nigel Green.
“Add revenue recycled between suppliers and buyers, and a meaningful slice of the earnings being celebrated rests on accounting and financing choices as much as on end demand.”
The funding itself is changing shape, with analysts projecting around $300 billion of investment-grade bonds for AI data centres this year alone.
The deVere CEO says: “The AI buildout started on cash. It’s increasingly running on credit, and credit changes the risk profile entirely.
“A share price can fall and recover. Debt has to be repaid on schedule, whether the revenues show up or not. And this debt is landing in the bond funds and pension pots of savers right around the world.”
Nigel Green concludes: “Circular money makes a boom look stronger on the way up and fall harder on the way down.
“The AI giants may well grow into every dollar being lent to them. If they don’t, the losses won’t stay inside Silicon Valley and they’ll be shared by investors far beyond the companies at the centre of the loop.”
