Oracle (ORCL 0.47%) is signing mega-deals to deliver AI infrastructure to hundreds of clients. The company's remaining performance obligations, a metric that measures the value of signed customer contracts that haven't yet been recognized as revenue, reached $523 billion at the end of the most recent quarter.
While Oracle has more than 700 AI customers, OpenAI accounts for most of that backlog. A reported $300 billion cloud infrastructure deal with the AI start-up is responsible for much of the RPO growth Oracle has experienced this year. Fulfilling that contract will require a massive build-out of AI data centers.
Image source: Getty Images.
The obvious question investors should be asking is this: What happens if OpenAI can't pay? The company addressed this issue in its second-quarter earnings call in December, but its answer leaves a lot to be desired.
An adaptable AI cloud lowers risk today
When an analyst asked Oracle management how quickly the company could convert a data center from one customer to another in the event a large customer was unable to pay its bill, Oracle co-CEO Clay Magouyrk laid out some impressive information about Oracle's cloud infrastructure.
When Oracle needs to move capacity from one customer to another, it only takes a few hours at most. This quick turnaround is the result of architectural decisions made when Oracle started building out its cloud infrastructure business. In fact, Oracle is constantly shifting capacity around to meet customer needs.
"So, you know, we have lots of customers that might sign up for a few thousand of one type of GPU, then they'll come back and say, well, actually, you know, I'd like to get even more capacity somewhere else. Will you take this back? And we do that all day every day. We're constantly moving customers around," said Magouyrk during the earnings call.
The ability to reassign AI computing capacity quickly means customer concentration risk today is minimal. If a customer doesn't pay, Oracle can shift that capacity to other customers with ease. In the current environment, where demand for AI infrastructure is greatly outstripping supply, there's no shortage of customers willing to take on that excess capacity.
From a financial perspective, this quick switching enables Oracle to maintain high utilization rates. That's critically important for Oracle's gross margin. The depreciation of GPUs and other server equipment is incurred regardless, so any meaningful downtime would hurt the bottom line.
NYSE: ORCL
Key Data Points
It won't matter if the industry overbuilds
While Oracle's ability to shift AI infrastructure capacity between customers reduces customer concentration risk right now, it won't matter in an oversupply scenario. If the AI industry overbuilds AI infrastructure over the next few years, a large customer not paying would be a disaster for Oracle, as other customers would be unable to absorb that capacity.
While there's no risk of an oversupply in the near term, the situation could change quickly. Earlier this year, Microsoft CEO Satya Nadella discussed the potential for an overbuild of AI capacity, saying that Microsoft would lease some capacity rather than building it to reduce risk. Goldman Sachs predicted that data center occupancy rates will peak in late 2026 before moderating as growth slows and data center capacity comes online.
The fundamental problem is that a slew of large tech companies are building massive amounts of capacity independently of each other based on optimistic estimates regarding future demand for AI. AI is a transformative technology, but no one can predict anything about the AI market two or three years out. If the demand doesn't materialize as expected, there's going to be an abundance of AI computing capacity in a few years.
For a company like Oracle, which uses debt to finance AI data centers and is heavily dependent on OpenAI securing the funds necessary to fulfill its massive commitments, an oversupply scenario would be an unmitigated disaster. Customer concentration risk may not be significant today, but in a few years, it could pose an enormous risk for Oracle.
While Oracle attempted to soothe concerns about its heavy dependence on OpenAI, investors should still be very concerned about what the worst-case scenario looks like.
