Oracle can’t keep up with demand for its powerful artificial intelligence data center infrastructure.
Oracle (ORCL 0.82%) was founded in 1977, and it has participated in almost every technology revolution since. It initially rose to prominence on the back of its database management software, but it then helped its business customers prepare for the dawn of the internet, cloud computing, and now, artificial intelligence (AI).
Oracle operates some of the best data center infrastructure in the industry for developing AI, and demand is significantly outstripping supply. The company will report its financial results for its fiscal 2025 first quarter (which ended Nov. 30) in early December, and there’s one number every investor needs to watch.
Image source: Getty Images.
A leader in AI data center infrastructure
To make AI chatbots and software applications “smarter,” they need to run on bigger large language models (LLMs). Industry-leading LLMs — like those developed by OpenAI and Anthropic — now have trillions of parameters, which makes them highly knowledgeable on a variety of topics. Training those LLMs requires a lot of computing power, and so does deploying them for customers to use, which is known as the AI inference stage.
Both training and inference happen inside data centers, which are filled with graphics processing units (GPUs) from suppliers like Nvidia. Tech giants like Oracle build, manage, and rent that infrastructure to developers for a fee, which is usually charged by the minute. Therefore, those developers will flock to data center providers with the fastest data centers, because that usually means lower costs.
The Oracle Cloud Infrastructure (OCI) Superclusters allow developers to scale up to 65,536 Nvidia GPUs (and soon, more than 131,000 of the latest Blackwell GPUs), which is more than any other data center provider. Plus, the company’s random direct memory access (RDMA) networking technology moves data from one point to another much faster than traditional Ethernet networks, which contributes to lower costs.
Automation is another key factor. No matter its size, every Oracle data center is identical in terms of functionality. That means the company can operate them all by using software, with no human staff required. That obviously keeps costs down, but it also enables Oracle to build and launch new data centers extremely quickly.
Oracle is widely recognized as one of the most cost-efficient data center operators in the industry, which is why it’s attracting top AI start-ups like OpenAI, Cohere, and Elon Musk’s xAI. In fact, the company can’t keep up with demand — it had 162 data centers either live or under construction as of its fiscal 2025 first quarter, but it plans to eventually have between 1,000 and 2,000.
The key number every investor should watch
The revenue Oracle generates by renting its data centers is accounted for under its OCI segment. That part of its business brought in a record $2.2 billion in revenue during the fiscal 2025 first quarter, which was a 45% increase from the year-ago period.
That growth rate accelerated from the previous quarter, and the company has often said OCI revenue could grow by at least 50% consistently.
As I mentioned earlier, Oracle can’t keep up with demand for its data centers. It’s building new ones as quickly as it can, but a lack of supply is the primary reason OCI revenue hasn’t been increasing by 50% or more during the past few quarters.
That’s reflected in the company’s remaining performance obligations (RPOs), which is sort of like an order backlog. RPOs came in at a record $99 billion during Q1, which was a whopping 53% increase from the year-ago period. Oracle said it signed 42 new contracts for GPU computing capacity during Q1 alone, worth $3 billion.
OCI revenue (and its growth rate) is the key number investors should watch when Oracle reports its latest results this month.
Oracle has a bright future in AI, but its stock isn’t cheap
Oracle stock is up by 72% in 2024 so far, and it’s trading near a record high. Based on its trailing-12-month earnings per share of $3.88 and its stock price of $181.41 as of this writing, it trades at a price-to-earnings (P/E) ratio of about 47.
That’s a 48% premium to the Nasdaq-100 index, which trades at a P/E of about 32, implying Oracle stock is quite expensive relative to its big-tech peers.
Therefore, investors with a time horizon of one year or less should probably avoid buying this stock. However, the picture looks a bit different for investors who are willing to hold the stock for the next five years or more.
The automated nature of Oracle’s data centers should drive its gross profit margin wider at scale. Therefore, if the company expands its data center footprint by 10-fold as I mentioned earlier, its earnings growth could accelerate during the next few years.
That alone could lead to stellar long-term returns for investors even if they pay a premium for the stock today. However, this scenario depends on continued strength in the OCI segment, so it’s important to monitor Oracle’s results not just in December, but in every quarter to come.
