The top secret of the king of AI, Nvidia Jensen Huang, is hidden in the Inventec exhibition area. Photo by Chen Liwen
With the blessing of Nvidia Jensen Huang, the godfather of AI, the number of people at the COMPUTEX exhibition has skyrocketed by 8 to more than 85,000 people, and when the spotlight in the exhibition hall is chasing the GB200 NVLink 72 AI cabinet of up to 5~6 million US dollars, no one has noticed that Huang has left a real "Top Secret" in the Inventec exhibition area! Inventec's P4000 is an MGX 4U rack server, which is not as powerful as GB200, but it is more in line with the needs of general enterprises and small and medium-sized cloud service providers (CSPs). In other words, NVIDIA's high-end and mainstream products have been prepared simultaneously.
The supply chain also revealed that Huang's whirlwind and product advancement speed are also amazing, the next-generation Rubin architecture is already in trial production, and the GB200 is ready to be assembled and shipped in Taiwan, which is significantly ahead of the previous plan for shipments at the end of the year.
Inventec showed off its latest NVIDIA GB200 NVLink AI cabinet at this year's COMPUTEX, and Huang also left his signature on the cabinet, but the real highlight was the P4000 that was left to the sidelines.
This MGX 4U Jaguar-style server, which was humorously signed by Huang as a "Top Secret", can be equipped with 16 B200 for NVLink connection, with a water-cooled NVLink switch, although the computing power is not as good as the latest architecture of GB200, but compared with HGX computing power design can be more than doubled, the price is more friendly than GB200, and more in line with the needs of the highest specifications of general enterprises and small and medium-sized CSPs.
NVIDIA's product coverage strategy and speed of advancement are equally amazing! Huang was asked at a recent global press conference that if there are only two AI companies left in the world, one is Nvidia, who will the other be? Who is your enemy? Huang's answer showed a unique strategic style, emphasizing that Nvidia is a market maker, not a market predator, and that innovation has been integrated into the corporate culture and will continue to create new markets in the future.
In fact, it can also be seen from the supply and production side that Nvidia has shortened the product from the latest Blackwell architecture to the next-generation Rubin architecture platform, which is equivalent to shortening the product from 2 years to 1.5 years. According to industry analysts, Huang's strategy has always been simple and crude, that is, to make the opponent desperately unable to catch up, and the most terrifying thing is that it is not just a slogan, the next generation of Rubin is in trial production, and the GB200 has been ready to be assembled and shipped in Taiwan, and the production speed is all ahead of schedule.