On the evening of May 23rd, Beijing time, Jensen Huang, CEO of NVIDIA, said today that the new AI models that support video generation will further stimulate the demand for NVIDIA's AI chips.
Before that, AI chatbots represented by ChatGPT have already triggered a shortage of NVIDIA's AI chips. But NVIDIA now expects that the new AI models that can create videos and conduct human-like voice interactions will further stimulate the growth of NVIDIA's AI chip orders.
Last week, OpenAI just released a new flagship generative AI model, GPT-4o, which supports text, voice, and video generation capabilities. It has an impressive human-like conversation ability. At the same time, competitors such as Google have also shown similar new products.
Jensen Huang told the media today: There is a lot of information in life that must be based on video and physics. So this is the next important thing. 3D video, and a whole bunch of other things, these systems will be very large.
Training and running advanced AI systems require more computing power, which boosts the demand for NVIDIA's Grace Hopper chips, such as the H200. This chip was initially used in OpenAI's GPT-4o, a multimodal AI model that can conduct realistic voice conversations and interact across text and images.
In addition, other customers of NVIDIA, including Google DeepMind and Meta, have also released AI image or video generation platforms.
Earlier today, NVIDIA released its first-quarter financial report, with revenue increasing by 262% year-on-year and net profit increasing by 628% year-on-year. In addition, the performance guidance for the next quarter is also much higher than expected.
Derren Nathan, the head of equity analysis at the investment and online trading platform Hargreaves Lansdown, said: The demand for NVIDIA chips is broad-based, and large language models need to increasingly support multiple modes, not only being able to understand videos, but also being able to understand text, voice, 2D and 3D images.
In addition, the video AI models used in the automotive industry are also becoming a major driving force for the demand for NVIDIA chips. Colette Kress, CFO of NVIDIA, said at the earnings call today that Tesla has expanded its processor cluster used for AI training to about 35,000 H100 chips to achieve autonomous driving.
Kress also said that it is expected that the automotive industry will become the largest vertical enterprise in NVIDIA's data center business this year.
And Nathan said: Video generation is definitely one of the powerful and already proven application cases of AI, and not limited to single content production.