If you had asked me five years ago what could possibly threaten tech giants with massive cash flows and impenetrable economic moats, I would have struggled to come up with an answer. Today, the answer is clear: Debt-fueled overinvestment in AI infrastructure.
There's a gargantuan amount of planned AI data center capacity set to come online over the next few years. Microsoft, Amazon, Meta Platforms, and Alphabet have greatly ramped up their capital spending to capture AI computing demand, and governments around the world are hurling vast sums of capital into AI investments. It's the fear of missing out on an unprecedented scale.
Of all the tech giants jumping headfirst into the AI inferno, Oracle (ORCL 1.47%) is putting itself at the most risk if the AI boom goes sideways.
Image source: Getty Images.
Oracle already has loads of debt
Oracle has spent billions of dollars buying back its own stock over the past decade, a strategy largely funded by debt. Oracle had minimal debt in the early 2010s, but the company's debt load has exploded since then. As of Aug. 31, Oracle's total debt reached $91 billion, and the company had another $14 billion in operating lease liabilities on its balance sheet. In the most recent quarter alone, Oracle paid $923 million in interest payments, which ate up more than 20% of its operating income.
Oracle's AI infrastructure deals, most notably a reported $300 billion agreement with OpenAI, are going to require a massive amount of new debt on top of what Oracle already owes. The company issued $18 billion in bonds in September, including some that mature in 40 years. Analyst Dave Novosel at Gimme Credit called the long maturities "unusual" for a technology company, since the tech industry changes so rapidly.
Already, Oracle's free cash flow, defined as operating cash flow minus capital expenditures, has plunged into negative territory. As the company builds AI data centers to fulfill its deals with OpenAI and other customers, more debt will be needed to get those projects across the finish line.
One question to ask is whether these investments are worth it under the best-case scenario. Oracle has projected adjusted gross margins between 30% and 40% for its AI infrastructure business, which is well below the 40% to 60% range for cloud computing and even further below the sky-high margins typically associated with software, Oracle's core business. Oracle is investing heavily in a business that doesn't appear all that lucrative.
NYSE: ORCL
Key Data Points
What happens if OpenAI doesn't pay?
What's becoming clear from the quality of frontier large-language models coming from Alphabet, Anthropic, and others is that OpenAI doesn't seem to have a competitive advantage or any meaningful lead in the AI race. For OpenAI to require the massive AI computing capacity it has agreed to purchase from Oracle, the company must significantly increase the number of users and overall usage of its platform. That becomes a trickier task when there are plenty of comparable alternatives.
OpenAI will also need to raise a mountain of new capital, with one analyst estimating the company's funding needs through 2030 at $207 billion. A potential IPO down the road could raise a lot of cash, but there are two main risks. First, the lack of a competitive advantage, and second, the potential for the AI boom to go bust. If sentiment around AI shifts enough in the negative direction, OpenAI may be forced to scale back its ambitions.
For Oracle, any roadblock for OpenAI represents a potential disaster. Oracle is taking on mountains of debt to build AI data centers that are somewhat questionable in terms of profitability, even if everything goes according to plan. If OpenAI faces funding challenges, Oracle could be saddled with an oversupply of expensive AI infrastructure. As Oracle continues to pile on more debt to fund its AI infrastructure, the risks intensify.
Oracle's AI data center building spree is starting to look like an all-in, bet-the-company kind of endeavor. It might work out, but investors should be prepared for a potential catastrophe.
