Despite being valued at just over $3 billion, Nvidia is nervous
The GPU maker’s rapid growth over the past two years, one of the world’s most valuable companies, has been on shaky ground: While its chips are at the center of the AI revolution, the superscalers Nvidia relies on are the most important. Customers are also your biggest competitors. Amazon Web Services, Microsoft, and Google are all developing their own AI chips in an effort to save money and find a unique advantage.
At the same time, those same cloud companies stand to make more money off the GPUs Nvidia sells to end customers by leasing them: For every dollar a cloud provider spends on a GPU, the chip designer said, it gets $5 back. . Dollars over four years In theory, Nvidia could cut out the middleman and simply offer GPUs through its own cloud service and keep all the profits. But beyond regulatory concerns, the risk could wipe out its biggest customers and cost it tens of billions of dollars to get up and running. Instead, the company is looking to create a third option: a cloud within the cloud. Last year, Nvidia introduced DGX Cloud, a service offered on top of other companies’ cloud platforms. Cloud providers rent Nvidia servers and use them as a cloud that Nvidia can market and sell to companies looking for supercomputers with large GPUs. Google, Microsoft, and Oracle agreed to the deal early on, but AWS held out until December, when it finally shut down. The hyperscalers have not said anything about the exact deal, and none of them are advertising the service on their websites. Instead, DGX Cloud can be seen as a way for Nvidia to use the disappointing demand for its GPUs as a way to make room in the cloud and position itself in the market to capture some services revenue and develop direct relationships with the market. “I wouldn’t call it a Trojan horse at all,” Alexis Beverlin, director of DGX Cloud and former vice president of Meta Infrastructure, tells DCD.
“It’s a deep partnership,” he says repeatedly in the interview. “What we do is we work extensively with cloud service providers [CSPs].”
Nvidia announced in its May 2024 earnings report that it has committed to spending at least $9 billion on cloud computing services over the next few years, up from $4.5 billion in January. That figure includes DGX Cloud, showing the rapid expansion of the commitment. “DGX Cloud gives all CSPs access to the latest Nvidia technology,” Bjorlin said. “If we step back and think about the end-user experience, we have all the capabilities.”
He argues that this is essential for the future development of artificial intelligence. A model can change. Changing the mix of specialists puts different stresses on the network and changes the performance of workloads.
DGX Cloud is really trying to understand the broader set of areas where AI is evolving, to make sure that we are designing for that, so that ultimately Nvidia GPUs are the end point for each of these workloads as AI.”
The Balance Between Cloud and Nvidia Technology
This is necessary for the future development of artificial intelligence, he argues. “A model can change. The shift to a mix of experts puts different stresses on the network and changes the performance of workloads.
“DGX Cloud is really trying to understand the broader set of areas where AI is going to evolve, to make sure that we’re designing for that, so that ultimately Nvidia GPUs are the ultimate landing point for each of these AI workloads.”
