NVIDIAPowering Intel’s historic rise is its data center business, which grew a staggering 427% in the most recent quarter as companies ramped up development of the company’s artificial intelligence processors.
Now, Nvidia is hinting to investors that customers who spend billions of dollars on its chips will also be able to benefit from AI. This concern is swirling internally because there is only so much cash that infrastructure can spend before customers start seeing some profit.
If Nvidia’s chips can provide a strong and sustainable return on investment, it suggests there may be room for the AI boom to continue as it moves past the early stages of development and companies plan long-term projects. doing.
The most important customers for Nvidia’s graphics processing units are major cloud providers. Amazon web services, microsoft Azul, Google cloud and Oracle Cloud. The company said cloud accounted for “mid-40%” of its $22.56 billion in data center revenue in the April quarter.
A new crop of GPU-focused data center startups are also emerging that buy Nvidia’s GPUs, install them in server racks, load them into data centers, connect them to the Internet, and rent them out to customers by the hour.
For example, GPU cloud CoreWeave currently charges $4.25 per hour to rent an Nvidia H100. Large amounts of this kind of server time are essential to train large language models such as OpenAI’s GPT, and many AI developers will eventually have access to Nvidia hardware.
After Nvidia reported better-than-expected earnings on Wednesday, Chief Financial Officer Colette Kress told investors that cloud providers are seeing an “immediate and powerful return” on their investments. She said that for every $1 they spend on Nvidia hardware, cloud providers can rent it for $5 over the next four years.
Kress also pointed to the company’s HDX H200 product, which combines eight GPUs to provide faster access, saying newer NVIDIA hardware will provide an even better return on investment. Meta Instead of direct access to cloud computers, it uses the Llama AI model.
“This means that for every $1 spent on an NVIDIA HDX H200 server at current prices, API providers offering Llama 3 tokens could generate $7 in revenue over four years,” Kress said.
Part of the calculation includes whether the tip runs 24 hours a day or less frequently and how the tip will be utilized.
Nvidia CEO Jensen Huang told analysts on an earnings conference call that OpenAI, Google, Anthropik and up to 20,000 generative AI startups are lining up for all the GPUs cloud providers can put online.
“All the work that’s been done so far [cloud service providers] “Customers are putting a lot of pressure on us to deliver the system and get it up and running as quickly as possible,” Huang said.
Although Meta is not a cloud provider, it has expressed an intention to spend billions of dollars on 350,000 Nvidia chips, Huang said. Facebook’s parent company, Meta, will need to monetize its investment through its advertising business or by incorporating chatbots within its current apps.
Meta’s servers are an example of “critical infrastructure for AI production,” or “what we call an AI factory,” Huang said.
Nvidia also surprised analysts by giving an aggressive timeline for its next-generation GPUs, called Blackwell, to be available in data centers in the fourth quarter of the fiscal year. Those comments were tempered. Fear of slowing down Businesses are waiting for the latest technology.
Initial customers for the new chips include Amazon, Google, Meta, Microsoft, OpenAI and Oracle. Teslaand Elon Musk’s xAI, Huang said.
Nvidia shares rose 6% in after-hours trading, surpassing $1,000 for the first time. In addition to the earnings release, Nvidia announced a 10-for-1 stock split after the company’s stock price rose 25-fold over the past five years.
