NvidiaApple's historic rise is being driven by its data center business, which grew a whopping 427% last quarter as companies continue to snap up its AI processors.
Now, Nvidia is signaling to investors that customers who spend billions of dollars on its chips will be able to make money from AI, too. It's a concern hanging over the company because there's only so much cash customers can spend on infrastructure before they need to see some profits.
If Nvidia's chips can provide a strong, sustainable return on investment, it suggests that the AI boom may have room to continue as it moves past the early stages of development, and as companies plan longer-term projects.
Nvidia's top customers for its GPUs are major cloud providers — Amazon web services, Microsoft I visit, Google cloud and inspiration clouds. The company said it accounted for “mid-40%” of Nvidia's $22.56 billion in data center sales in the April quarter.
There's also a new crop of GPU data center startups that buy GPUs from Nvidia, install them on server racks, load them into data centers, connect them to the Internet, and then rent them to customers by the hour.
For example, CoreWeave, a GPU cloud, currently offers a rate of $4.25 per hour to rent an Nvidia H100. This kind of server time is necessary in large quantities to train a large language model like OpenAI's GPT, which determines how many AI developers end up accessing Nvidia hardware.
Following Nvidia's better-than-expected earnings report on Wednesday, Chief Financial Officer Colette Kress told investors that cloud providers are seeing an “immediate and strong return” on investment. If a cloud provider spends $1 on Nvidia hardware, it can lease it for $5 over the next four years, she said.
Chris also said that Nvidia's newer hardware will have a stronger return on investment, citing the company's HDX H200 product, which combines 8 GPUs, providing access to… dead Llama AI model, rather than initial access to a cloud computer.
“This means that 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,” Chris said.
Part of the calculation includes how the chips are used, whether they are running 24 hours a day or less frequently.
OpenAI, Google, Anthropic and as many as 20,000 AI startups are lining up to get every GPU that cloud providers can put online, Nvidia CEO Jensen Huang told analysts on an earnings call.
“All the work that's being done at all the (cloud providers) is taking up every GPU that's out there,” Huang said. “Customers are putting a lot of pressure on us to deliver and maintain systems as quickly as possible.”
Meta has announced its intention to spend billions on 350,000 Nvidia chips, even though the company is not a cloud services provider, Huang said. Meta, Facebook's parent company, will likely have to monetize its investments through its advertising business or by embedding the chatbot within its existing apps.
The Meta server cluster is an example of “core AI production infrastructure,” or “what we refer to as AI factories,” Huang said.
Nvidia also surprised analysts by offering an aggressive timeline for its next-generation GPU, called Blackwell, which will be available in data centers in the fiscal fourth quarter. These comments allayed fears of a slowdown as companies wait for the latest technology.
Early customers for the new chips include Amazon, Google, Meta, Microsoft, OpenAI, and Oracle. TeslaAnd Elon Musk's XAI, Huang said.
Nvidia shares jumped 6% in extended trading, surpassing $1,000 for the first time. In addition to announcing earnings, Nvidia announced a 10-for-1 stock split after the company's stock price rose 25-fold over the past five years.
