Nvidia is trying to use its dominant position in the field of AI hardware to get a share of the cloud software field. The move also puts NVIDIA in competition with its customers, the traditional cloud providers that buy its AI chips, which has caused some customer backlash.
As a result of OpenAI’s release of ChatGPT last year, especially the large number of AI developers that have sprung up, demand for servers with NVIDIA’s AI chips has skyrocketed across all industries, and these cloud providers are quickly realizing that their own arithmetic is no longer sufficient to meet the market’s needs.
In this delicate moment, NVIDIA naturally also found business opportunities.
According to the information, Nvidia last year to Amazon web services (Amazon web services, AWS) and other cloud computing service providers put forward an unusual proposal, and these enterprises have been Nvidia AI chip’s biggest buyers.
Simply put, NVIDIA wants to rent customers these servers with NVIDIA chips. This would allow it to sublease its customers’ servers to AI software developers, which naturally include some of the world’s biggest cloud developers.
As NVIDIA’s leverage grew, for example, with the new generation of H100 chips at its disposal, which happened to be the most pressing need of these cloud providers at the time, some of them had to bow down.
Microsoft, Google and Oracle all agreed to NVIDIA’s proposal, but AWS did not because it feared NVIDIA’s entry would threaten its position in the market, according to a person familiar with the decision.
Referring to its latest quarterly earnings report, NVIDIA reported revenue of $13.51 billion for the current fiscal quarter, up 88 percent from the previous quarter and 101% from the same period a year earlier.
After a blowout in May, NVIDIA has become the first semiconductor company ever to reach a market capitalization of $1 trillion. After that, investors have been looking for evidence that its fiscal second quarter will remain strong.
In March, NVIDIA CEO Jen-Hsun Huang announced a plan to open up supercomputers used to develop AI technologies like ChatGPT to commercial customers, allowing virtually any business to use the powerful but costly equipment on a cloud rental basis.
Of course, NVIDIA rent is not cheap, including eight A100 or H100 flagship chip prices of 37,000 U.S. dollars per month (currently about RMB 270,000). However, the opening of such supercomputers for a wider range of commercial customers will accelerate the development of artificial intelligence. The explosion of AI technology over the past period has already pushed NVIDIA shares up 77 % this year, with a market capitalization about five times that of longtime rival Intel.
Looking at Amazon AWS, the service provider didn’t get NVIDIA’s assistance until July of this year, when it launched a new Amazon Elastic Compute Cloud (EC2) P5 instance based on NVIDIA’s H100 Tensor Core GPUs. The service is described as allowing users to easily scale generative AI, high-performance computing (HPC) and other applications through a browser.
At NVIDIA’s previous annual developer conference, Jen-Hsun Huang said the company is working with partners such as Oracle to provide cloud access to NVIDIA’s DGX supercomputer. This supercomputer integrates up to 32,000 NVIDIA chips. The service allows anyone with a Web browser to log into the supercomputer.
