A slowdown in Nvidia’s business doesn’t seem to be expected anytime soon.
Larry Ellison owns 42% of Oracle (ORCL 1.53%)is a $465 billion technology giant that is building some of the most powerful data centers for artificial intelligence (AI) development.
NVIDIA (NVDA 2.94%) Nvidia supplies Oracle and most other tech companies with data-center chips called graphics processing units, or GPUs. The company has seen a spectacular surge in revenue over the past year, and demand for GPUs continues to outpace supply. But some investors are beginning to question whether Oracle and its peers can afford to pour billions more into the chip giant to fuel their AI ambitions.
Concerns that AI momentum is starting to fade are the main reason Nvidia shares are down 14.5% from their all-time high, but the market may have missed comments made by Ellison at Oracle’s financial analyst conference this month that hinted at even better news for Nvidia investors.
Oracle falls far short of AI infrastructure goals
Oracle’s data centers are unique in that they are automated. Regardless of size, each data center is operationally identical and can be built quickly because they don’t require human workers. Additionally, Oracle’s RDMA (Random Direct Memory Access) GPU networking technology allows data to flow from one point to another faster than traditional Ethernet networks.
Because most AI developers pay by the minute for computing power, Oracle’s data centers offer significant cost savings over competing infrastructure, which has led to surging demand from major AI startups such as OpenAI, Cohere and xAI. Oracle had 85 data centers operational as of the first quarter of fiscal 2025 (ended Aug. 31) with 77 more in the works, but Ellison believes it could operate 2,000 data centers in the long term.
Next year, Oracle plans to offer a cluster of 131,072 GPUs, a big jump from the company’s largest cluster today, which has about 32,000 GPUs. But there’s one other difference: The new cluster will use Nvidia’s latest Blackwell chips, which can run AI inference 30 times faster than the flagship H100 that Oracle uses today. In theory, this will enable developers to build the largest AI models ever made.
This will be a big win for NVIDIA, which reported data center revenue of $26.3 billion in its fiscal second quarter (ended July 28), driven mostly by GPU sales, up 154% from the same period a year ago. This growth rate slowed compared to the previous quarter as the numbers got too big, but NVIDIA customers are showing no signs of pulling away.
In fact, Oracle is spending $6.9 billion on data center infrastructure in fiscal 2024 and plans to double that amount in fiscal 2025. But it gets even better.
Ellison’s latest comments are great news for Nvidia
During the analyst conference, Ellison told the audience about the time he hosted Tesla CEO (and xAI founder) Elon Musk and Nvidia CEO Jensen Huang for dinner at Nobu in Palo Alto, where Ellison recalled Musk and Huang pleading for more GPUs.
Please take our money…take more. You are not taking enough…take more of our money. Please.
–Comments made by Ellison and Musk to Jensen Huang over dinner, according to Ellison.
Oracle Cloud Infrastructure (OCI) generated $2.2 billion in revenue in the first quarter (mainly from renting datacenter capacity to customers), up 46% year over year. However, Oracle ended the quarter with a record $99 billion in remaining performance obligations (RPOs), up a staggering 53%. The company said it signed 42 new contracts for $3 billion worth of GPU capacity in the first quarter, which led to a growing backlog.
Oracle can’t serve all its AI developers or convert their RPOs into revenue unless it brings more data centers online, so Ellison is begging Huang for more GPUs.
Tesla is in a similar situation: The company is vying for dominance in self-driving software and is looking to bring a cluster of 50,000 GPUs online by the end of the year to further train its AI models. Tesla plans to spend $10 billion on the infrastructure, but will need more capacity over time.
Image credit: Nvidia.
Now might be a great time to buy Nvidia stock
Oracle and Tesla aren’t the only companies investing heavily in data centers: Microsoft is spending $55.7 billion on AI infrastructure-related capital expenditures (capex) in fiscal year 2024 (ending June 30), and plans to spend even more in fiscal year 2025. Similarly, Amazon’s capex spending is expected to exceed $60 billion this year.
With trailing-12-month earnings of $2.20 a share, Nvidia’s stock is trading at a price-to-earnings (P/E) ratio of 52.7, compared with the 30.9 P/E ratio of the Nasdaq-100 Technology Index, which includes many of Nvidia’s biggest tech companies.
However, NVIDIA’s fiscal 2026 begins at the end of January 2025, and Wall Street expects the company to earn $4.02 per share for the year. This gives the company a forward P/E ratio of 28.8. In other words, investors willing to hold onto NVIDIA shares for at least the next year and a half could be getting a bargain at current prices — assuming Wall Street’s forecasts are accurate.
Nvidia’s business will eventually slow, as the current level of AI spending will be very difficult to sustain in the long term. Moreover, competition is slowly coming into the GPU space, which could erode some of the company’s market share over the next few years.
But based on the facts we have now, Nvidia stock is likely worth buying at current prices, and the planned spending on AI from some of the company’s largest customers suggests no slowdown is in sight anytime soon.
John Mackey, former CEO of Amazon subsidiary Whole Foods Market, serves on The Motley Fool’s board of directors. Anthony DiPizio has no position in any of the stocks mentioned. The Motley Fool owns shares of and recommends Amazon, Microsoft, Nvidia, Oracle, and Tesla. The Motley Fool recommends long January 2026 $395 calls on Microsoft and short January 2026 $405 calls on Microsoft. The Motley Fool has a disclosure policy.
