Nvidia has reported record revenue of $13.51bn for its second quarter of 2024, an increase of 101% from a year ago, driven by artificial intelligence (AI) accelerated computing applications.
The company posted a 171% increase in datacentre revenue from the previous year, which CFO Collette Kress said has been led by cloud service providers and large consumer internet companies.
She said that strong demand for the Nvidia HGX platform based on the company’s Hopper and Ampere graphics processor (GPU) architectures was primarily driven by the development of large language models and generative AI. The company’s datacentre compute business grew 195% from a year ago, driven by demand for the Hopper-based HGX platform, she added.
In a transcript of the earning call, posted on Seeking Alpha, Kress said that there has been “tremendous demand for Nvidia accelerated computing and AI platforms”.
Referring to the Biden administration’s plans to limit export of high tech products to China, she said: “Over the long term, restrictions prohibiting the sale of our datacentre GPUs to China, if implemented, will result in a permanent loss and opportunity for the US industry to compete and lead in one of the world’s largest markets.”
When asked about whether there was enough applications to propel the demand for accelerated computing that Nvidia has benefited from, co-founder Jensen Huang described the opportunity to distil large language AI models into smaller models.
“[When] you create these large language models and derive from them smaller versions of the models, this is essentially a teacher-student model. It’s a process called distillation. These smaller models might have excellent capabilities on a particular skill, but they don’t generalise as well,” said Huang.
Looking at enterprise applications that can make use of advances in accelerated computing and AI, Huang discussed the company’s ongoing collaboration with VMware, which improves IT manageability. He said that for enterprises to train and deploy AI systems, they need have management systems, an operating system, security and software-defined datacentre infrastructure, which is VMware’s domain.
“We’ve been working several years with VMware to make it possible for it to support not just the virtualisation of CPUs, but a virtualisation of GPUs, as well as the distributed computing capabilities of GPUs, supporting Nvidia’s BlueField for high-performance networking.”
With several hundred thousand VMware customers around the world, he said the VMware Private AI Foundation makes it possible for enterprise to deploy AI in their own datacentres.
According to Huang, in combination with HP, Dell, and Lenovo’s new server offerings based on the Nvidia L40S architecture, any enterprise can build “a state-of-the-art AI datacentre” and work with their own generative AI models and applications.