December 19, 2024 – Recent speculation regarding delays in the shipment of NVIDIA’s GB200 AI servers has sparked widespread discussion, but CEO Jensen Huang has provided a reassuring update to the market. He confirmed that the GB200 is currently in full production and progressing smoothly, effectively dispelling any lingering doubts.
Huang further emphasized the impressive capabilities of NVIDIA’s latest Blackwell platform servers, which achieve a remarkable combination of high performance and energy efficiency during inference. He noted that the servers’ performance has been boosted by an astonishing 30 times, highlighting NVIDIA’s prowess in technological innovation and drawing more attention to the Blackwell platform.
“An increasing number of companies are racing to build AI capabilities, and a three-month gap can potentially change the game,” Huang stated, expressing confidence in the “superb performance” of the Blackwell platform servers. His words not only stabilized the market but also provided a sense of assurance to GB200 AI server manufacturing partners such as Foxconn, Quanta, Wistron, and Inventec.
It’s worth mentioning that Foxconn, Quanta, and other initial manufacturing partners responsible for shipping the GB200 AI servers had previously revealed plans to achieve small-volume shipments by the end of this year and ramp up production in the first quarter of next year. Now, with Huang’s reassuring update, these partners are more optimistic about the shipment prospects of the GB200 AI servers. They have confirmed that they will proceed as planned and are expected to generate substantial revenue growth for their respective companies.
In an interview, Huang also disclosed that the shipment progress of the Blackwell platform’s AI chips is equally smooth and reiterated their exceptional performance. He pointed out that the process of preparing data for training models and conducting model training, which previously took months to complete, can now be reduced to just one-third or one-fourth of the original time using the Blackwell platform. This significant improvement will undoubtedly enhance the efficiency of AI model training, providing a more solid foundation for the widespread application of AI technology.
