The company that’s winning because it makes the best chip in the business is now bolstering that innovation by jumping into cloud computing. Nvidia Corp. launched DGX Cloud, a virtual artificial intelligence supercomputer, putting its brand on all things AI as businesses around the world have rushed to capitalize on AI’s momentum since ChatGPT’s launch exactly one year ago.
Its DGX Cloud product, which chief executive Jensen Huang describes as the “iPhone moment for AI,” was released in March. It offers customers the ability to access Nvidia’s powerful AI computing in monthly rental subscriptions starting at $36,999, as opposed to paying $200,000 to buy a physical server.
As more businesses shift to virtual servers, DGX Cloud is Nvidia’s way of giving customers a second option to fulfill their computing needs, while charging a premium for it. On the roadmap to building AI products, Nvidia’s DGX is the vehicle that will carry most, if not all products from start to end.
“DGX Cloud is the most exciting development in the company,” said Jim Kelleher, director of research and senior analyst at Argus Research Corp. “They’re offering AI delivered as a service, as it cannot be built outside.”
There are a few benefits to offering AI supercomputing as a service. First is the speed by which customers can access the product – demand for Nvidia’s A100 and H100 graphics units has skyrocketed over the last year, but customers have to wait months at a time to receive the units, potentially disrupting their computing process. DGX Cloud, on the other hand, can be accessed either with physical servers or through a web browser at a monthly cost, which customers like because it cuts costs for building, setting up, and maintaining these servers.
“It’s cheaper for enterprises to rent from cloud computing companies than to build their data centers in-house,” said Brian Collelo, director of technology equity research at Morningstar Research Services LLC.
Signing up more customers – even at a monthly subscription cost – may also help Nvidia continue to increase its revenues at a blistering clip, which investors and analysts have now come to expect. Nvidia’s total revenue in the third quarter of fiscal 2024, which began in March, was more than $18 billion, up more than tripled from a year before. The company’s data center revenue, which includes DGX Cloud, nearly quadrupled from the third quarter a year ago, to more than $14 billion.
Wall Street has lofty expectations for Nvidia revenues going forward. Analysts forecast the company’s revenue will go up to $92 billion in fiscal year 2025 and $108 billion in 2026.
The company has insisted it is not competing with traditional cloud giants like Microsoft and Amazon. Instead, it says, it intends to focus on managing AI workloads within the cloud, collaborating with existing cloud providers. A simple analogy for DGX Cloud as a product would be a Starbucks kiosk inside a Barnes & Noble bookstore. Customers get more options to streamline their AI work by either buying physical servers or simply logging into Nvidia’s AI platform through a web browser.
In September, The Information, a prominent online tech publication, reported that Nvidia planned to lease servers within cloud providers’ data centers, powered by Nvidia GPUs, and in turn rent them to customers with AI-specific needs.
“I think the other cloud companies see Nvidia as a niche provider with a focus on AI for a portion of workload,” said Collelo.
Nvidia’s stock is the top-performing stock on the S&P 500 and tech-heavy NASDAQ. The company’s share price has more than doubled since the start of this year, closing at $455 a share Wednesday, while the S&P 500 gained close to 20% in the same period.
Cloud providers like Microsoft and Amazon fit Nvidia’s graphics processing units (GPUs) — computer chip that runs simultaneous computer calculations through multiple pathways — inside their data centers to accelerate computing power. This power can then be used to train large language models like that of ChatGPT, a major example of public cloud computing. DGX Cloud, while a small contributor to Nvidia’s software revenue, would help the company foster closer relationships with its existing customers.
The company has added more specialty cloud partners to appeal to more customers. CloudWerx, a cloud consulting firm in Santa Clara, works with clients who prefer to use Google Cloud over other cloud providers. DGX Cloud, which is yet to run on Google Cloud, is still good news to them.
“In the very near future, I will be able to go to Google Cloud, spin up a DGX box, and get all the stuff you do on the Nvidia Cloud, because that’s the same product,” said Sidhant Gupta, chief technology officer at CloudWerx, a cloud consulting firm in Santa Clara, Calif.
Competitors are not far behind, but analysts say it will be a while before they catch up with Nvidia. Microsoft, one of Nvidia’s biggest customers, announced in November that it was speedily working on its own AI specialty chip to reduce its reliance on Nvidia. Amazon and Google are both reportedly working on building in-house chips that could potentially compete with Nvidia. Tech companies overseas are also prepared to make their chips if the U.S. continues to ban the export of certain chips to certain countries. Meanwhile, Advanced Micro Devices (AMD), Nvidia’s direct competitor in the U.S., announced a new AI chip Wednesday, adding to Nvidia’s list of worries.
Nvidia spokespeople declined to comment for this article.
Nvidia’s decision to enter cloud computing is a winning strategy if everything falls into place. Whether users adopt DGX directly or virtually, they will add to Nvidia’s bottom line. Wall Street analysts predict the company will bring in close to $59 billion in total revenue this year. Over time, cloud providers may come up with their in-house chips or go to Nvidia competitors like AMD and Intel. But until then, Nvidia can continue to charge a premium for its chips, making the company extremely profitable as an investment.