Nvidia shows no signs of AI slowdown after over 400% increase in data center business

Nvidia’s historic run is fueled by its data center business, which rose by 427% in the last quarter as companies continue to purchase its AI chips.

Now, Nvidia is suggesting to investors that customers that spend billions of dollars on its chips will be able to profit from AI as well. It’s been a source of anxiety for the corporation because clients can only spend so much money on infrastructure before they need to show a profit.

If Nvidia’s chips can give a significant and durable return on investment, the AI boom may have more opportunity to grow as it progresses beyond the early stages of development and corporations prepare for longer-term initiatives.

Nvidia’s primary clients for its graphics processing units include major cloud providers, including Amazon Web Services, Microsoft Azure, Google Cloud, and Oracle Cloud. They accounted for “mid-40%” of Nvidia’s $22.56 billion in data center sales in the April quarter, the company stated.

There is also a fresh crop of dedicated GPU data center firms that purchase Nvidia GPUs, install them in server racks, load them into data centers, connect them to the internet, and then rent them out to consumers by the hour.

For example, CoreWeave, a GPU cloud, presently charges $4.25 per hour to rent an Nvidia H100. This type of server time is required in enormous quantities to train a large language model like OpenAI’s GPT, and that is how many AI developers end up using Nvidia hardware.

Following Nvidia’s better-than-expected earnings announcement on Wednesday, finance chief Colette Kress told investors that cloud providers were seeing a “immediate and strong return” on their investment. She claimed that if a cloud company invests $1 in Nvidia gear, it can rent it out for $5 over the next four years.

Kress suggests that Nvidia’s HDX H200 solution, which combines 8 GPUs and accesses Meta’s Llama AI model, has a higher return on investment compared to traditional cloud computing.

“That means for every $1 spent on NVIDIA HDX H200 servers at current prices, an API provider serving Llama 3 tokens can generate $7 in revenue over four years,” Kress stated.

Part of the computation includes how the chips are used, such as whether they run continuously or infrequently.

During the earnings call, Nvidia CEO Jensen Huang told investors that OpenAI, Google, Anthropic, and up to 20,000 generative AI startups are vying for every GPU that cloud providers can bring online.

“All of the work that’s being done at all the [cloud service providers] are consuming every GPU that’s out there,” Huang stated. “Customers are putting a lot of pressure on us to deliver the systems and stand it up as quickly as possible.”

Huang stated that Meta has expressed its desire to spend billions of dollars on 350,000 Nvidia chips, despite the fact that the company is not a cloud service provider. Facebook parent Meta will most likely have to recoup its investment through advertising or by incorporating a chatbot into its existing programs.

Huang described Meta’s server cluster as “essential infrastructure for AI production,” also known as “what we refer to as AI factories.”

Nvidia also stunned investors by announcing an aggressive timeframe for its next-generation GPU, known as Blackwell, which will be available in data centers in the fiscal fourth quarter. These comments allayed concerns about a delay while businesses waited for the latest technology.

According to Huang, the initial users for the new processors are Amazon, Google, Meta, Microsoft, OpenAI, Oracle, Tesla, and Elon Musk’s xAI.

Nvidia stock rose 6% in extended trading, topping $1,000 for the first time. In addition to reporting earnings, Nvidia announced a 10-for-1 stock split, following a 25-fold increase in the company’s share price during the previous five years.