When it comes to expanding gigantic AI data centers, tech companies are reaching financial limits. In order to cover the enormous demand for high-performance chips, industry giants are increasingly turning to alternative sources of funding. But the billions in loans pose dangers.
The construction of data center infrastructure worldwide is creating a race for capital. According to a report in the Wall Street Journal (WSJ), major players in the field of artificial intelligence are preparing a series of spectacular financing deals to cover the costs of the necessary computer hardware.
The chip company Broadcom has sought financing of over $50 billion in the past few weeks. This is intended to finance a tailor-made AI chip that the company is developing together with ChatGPT developer OpenAI.
The financial investors Apollo and Blackstone are considered possible lenders. The discussions are said to be at an early stage and the scope of the transaction could still change. The funding could include several gigawatts of OpenAI chip capacity, a person familiar with the discussions said. The deal is expected to close before the end of the year.
The cloud service provider Oracle is also said to be in talks with Apollo and Goldman Sachs to raise funds for a large chip purchase. In order to expand its AI plans, SpaceX also wants to raise money. As the Financial Times reports, Elon Musk's space company plans to raise $40 billion through loans and bonds to finance the purchase of Nvidia chips. Just like Broadcom, Oracle also wants to close the deal this year. According to the "WSJ" report, it is unclear how many chips Oracle wants to finance. Purchasing Nvidia chips for a 1 gigawatt data center would cost tens of billions of dollars.
Meanwhile, industry experts warn against such constructs: "There is a risk of long-term debt burden because private equity investors secure these projects using debt financing and leasing structures instead of equity," says the head of the asset manager MainSky Eckhard Schulte ntv.de. If the completion of data centers stalls, the tech companies would have to service these fixed interest and installment obligations from day one on their own, before the technology even generates productive sales.
Until now, cloud providers such as Amazon Web Services and Oracle have typically financed their computer hardware using their own cash flow. But companies took out hundreds of billions of dollars in bonds to massively expand their AI infrastructure. In doing so, they ultimately pushed the public debt market to its limits. According to experts, the fact that some of the big tech companies are turning to Wall Street investment firms shows that the extreme costs of expanding data centers worldwide are pushing both the corporations' cash flows and the classic bond market to their limits.
There's also a new group of chip buyers, including OpenAI and rival Anthropic, who don't have the financial means to purchase their own hardware. In the past, leading AI laboratories have therefore rented the majority of their computing capacity from cloud providers. They now prefer to own more of their infrastructure themselves. Not only to reduce costs, but also to reduce their dependence on other companies.
This trend towards more in-house hardware further exacerbates the industry's already enormous financial needs and has noticeable consequences for the entire market environment: "If the tech industry's gigantic capital needs compete directly with states' borrowing, this inevitably drives up interest rates on the bond market," says Schulte. For highly indebted corporations like Oracle and capital-intensive pioneers like SpaceX, this environment means exploding credit costs of over 7.5 percent. “If AI revenue is delayed, this can immediately lead to severe cash flow crises and acute rating crashes,” warns Schulte.
