When the chipmaker helps finance the customers who buy its chips, is that an AI boom, or a loop?
What happened
Elon Musk's SpaceX, which absorbed xAI before its record $86bn IPO in June, is looking to raise around $40bn to buy Nvidia chips, according to the FT and Bloomberg. The FT reports the package will be roughly $10bn of bank loans and $30bn of investment-grade debt, with Apollo leading the deal and selling it on to a broad base of investors, with Pimco among a small group of lenders in talks.
The arrangement is expected to close in 2027, although Bloomberg says the talks are still at an early stage, so nothing is final yet. If it goes ahead, it would rank among the biggest debt deals yet for the AI build-out. Musk has said SpaceX will build "exclusively on Nvidia".
Yesterday, I quoted Apollo's Huw van Steenis saying the big AI spenders need to "dip into every pond", public and private, to fund their capex. It turns out his own firm was already lining up one of the biggest ponds of all. This is a reminder that when someone tells you where the money is going, it's worth knowing which way it flows through their own business.
Follow the money
The detail that matters most is the rating. SpaceX is rated BBB, the second-lowest rung of investment grade, which means insurers and pension funds can buy its debt in size, something they are far more restricted from doing with junk. Apollo's insurance and annuities arm, Athene, typically takes a large slice of the deals Apollo leads. In June, Apollo also led a $35bn financing for chips made by Broadcom, at the time the largest private credit deal on record.
Then there is Nvidia itself. In August it teamed up with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR on financing platforms designed to mobilise more than $500bn of outside capital for its chips and wider AI infrastructure. That's a target to be raised over time, not a single pot of money.
But according to the FT, Nvidia could backstop up to 25% of the chip value. Separately, the FT reported that Nvidia has held early talks with insurers about covering loans to smaller AI cloud companies, where the chips themselves are pledged as collateral, and Jensen Huang has been pitching chips as an "investable asset class".
Put that together and the chain looks like this: savers' and pensioners' money flows into insurers and asset managers, who lend to the companies buying Nvidia's chips, with Nvidia partly standing behind some of the loans. Not all AI debt is sitting in a pension fund. But the plumbing is increasingly being built to reach the most conservative, long-term money in the system. Morgan Stanley estimates AI infrastructure will need around $1.5tn of external financing by 2028. That money has to come from somewhere.
The counterargument
Demand seems to be insatiable. Paramount's $52bn debt sale to fund its Warner Bros. Discovery deal drew well over $100bn of orders, even at yields of up to around 9%. That is hardly a market refusing to fund big, leveraged deals. On their live show, Ritholtz Wealth's Josh Brown and Michael Batnik argued that the bear case of investors refusing to turn up for AI bonds can be "put to bed", pointing to data showing older Nvidia chips are still renting at firm prices. That undercuts the simplest version of the bear case, that the collateral becomes obsolete almost overnight. And BBB is still investment grade: SpaceX is a real business, with real revenues, not a start-up with a slide deck.
My read
I don't necessarily see a bubble. But I do think the risk is shifting in a way most equity investors aren't watching. The same chips can be the product, the collateral and, in some deals, part of the guarantee, while the money backing them is increasingly coming from insurers and pension funds. If chip values were to fall faster than expected, the losses wouldn't necessarily stay with the shareholders who signed up for volatility. Some of it could land with policyholders who didn't.
The bond market is already hinting at this. The FT reports SpaceX sold $25bn of bonds less than two weeks after its IPO, and they sold off soon after. Its 2056 bonds have recently traded at around 85 cents on the dollar, at a spread over Treasuries that the FT says is "similar to the level of junk bonds". Bloomberg reported that the cost of insuring SpaceX's debt against default jumped as soon as news of this deal emerged. The rating says investment grade. The price says something closer to high yield. When the rating and the price disagree, I'd listen to the price.
The psychology here is a familiar one. With yields this high, large investors want to put big sums to work, and a BBB label gives them permission. But focusing on the rating, rather than the collateral and the cash flows behind it, may be how risk gets mispriced in the AI mania.
The Comms Read
Equity investors will buy a vision; Musk has proved that repeatedly. Credit/debt investors want to see cash flows, covenants and collateral. When a company moves from selling a story to equity investors to asking for $40bn from lenders whose upside is capped but whose downside isn't, the IR job changes completely. For any company tapping the debt markets to fund AI spending, the message is simple: the audience has changed, so the disclosure needs to change with it.
Nvidia's "investable asset class" framing deserves a read of its own. Reframing a fast-moving piece of hardware as an asset class is a clever way to widen the pool of capital. Whether it holds up depends on what those chips are worth, and the cash they generate, over the coming years. I hope their crystal ball is well polished…
What to watch
The pricing of the $30bn of bonds and who ends up holding them; whether Nvidia's insurer talks turn into actual deals; SpaceX's next set of disclosures; whether the next mega-IPOs follow the same playbook of listing and then going straight to the debt market; and the hyperscalers' results later this month, starting with Microsoft.
Also interesting...
Porsche: smaller, but richer? Porsche has told investors to stop expecting the old volumes to come back. At its investor day yesterday, new chief executive Michael Leiters, ex-McLaren and ex-Ferrari, set out a "value over volume" plan for a smaller, stronger business that breaks even on fewer than 200,000 cars a year, having been geared, he said, to build 350,000. By 2030, the average price of its 10,000 most expensive cars is to rise by about a fifth, from around €270k to €330k.
In the UK, the Telegraph calculates, a 911 Turbo S Cabriolet would go from about £209k to £251k. Development costs are being cut by up to a fifth and 40% of management roles are going. The urgency is in the margin: 18% in 2023, 1.1% last year. It's the Ferrari playbook: sell fewer, charge more. Investors weren't fully sold, though: the shares rose as much as 5%, then slipped into the red.
I wonder, if someone is looking at a top end second hand Porsche, but is worried about the depreciation. Is now the moment to pounce? With this new pricing structure set to land, the residuals should hold up well if newer models become vastly more expensive. Food for thought. Reminds me, I must write my list for Father Christmas!
Disclosure: I hold shares in Nvidia.
Time will tell. GLA & DYOR.
Disclaimer
This newsletter is published by James Lea / CorpCast for general information, educational and journalistic purposes only. Nothing in it constitutes financial, investment, legal or tax advice, or a personal recommendation, and it is not an offer, invitation or inducement to buy, sell or hold any security, cryptocurrency or other financial instrument.
Any views, opinions, forecasts or interpretations expressed are the author's own personal opinions at the time of publication, do not reflect the views of any business or entity the author is affiliated with, and may change without notice. No account is taken of any reader's personal circumstances, objectives or financial situation, and nothing here should be relied upon as the basis for any investment decision.
The author may from time to time hold long or short positions in securities, cryptocurrencies or other assets referenced in this newsletter, and may transact in them at any time without further notice. This is a standing disclosure covering all issues and is not repeated or itemised per asset.
Investing and trading involve substantial risk of loss, including the potential loss of your entire investment. Past performance is not a reliable indicator of future results. Figures, data and third-party information are believed reliable but are not guaranteed to be accurate, complete or up to date.
James Lea / CorpCast is not authorised or regulated by the Financial Conduct Authority. Always conduct your own independent research and consult a suitably qualified, FCA-regulated financial adviser before making any investment decision.