Quanta develops GB200 chip AI server, expected to mass produce in September
Yang Qiling, Senior Vice General Manager and General Manager of Yunda, a contract manufacturer Quanta, confirmed on April 23 that servers equipped with the current "strongest AI chip" NVIDIA GB200 are expected to be mass-produced in September. The four major cloud service providers (CSPs) have high customer demand, and it is expected that Quanta's server shipments will maintain double-digit growth this year.
Yang Qiling stated that the GB200 server is currently under development, and based on the current chip progress, the system side is expected to be mass-produced and shipped as early as September. Large CSP manufacturers are quite optimistic about the new NVIDIA Blackwell architecture, with high demand and initial orders.
At present, there is continuous training of large language models around the world, and the demand for AI servers and chips continues to grow. Quanta predicts that the revenue share of AI servers will jump from 20% last year to over 50% this year.
In response to the growing demand from customers, Quanta is intensifying its production expansion this year, and new production capacity has been gradually opened in overseas factories. Yang Qiling revealed that Quanta's Fremont plant in California and Nashville plant in Tennessee will both contribute new production capacity this year, and the German assembly plant will also start production.
Previously, it was reported that after the release of Nvidia GB200 chips, Quanta successively won the GB200 server OEM orders from Google, Amazon AWS, and Meta, the three major cloud service providers (CSPs). Among them, Google's orders entered testing as early as July to August. It is reported that the average selling price of each cabinet system for GB200 servers is as high as 2 million to 3 million US dollars (approximately 14.45 million to 21.68 million RMB). The GB200 integrates two B200 GPUs with one Grace CPU, and each Nvidia DGX GB200 server system contains 36 Nvidia GB200 supercohips.
At present, there is continuous training of large language models around the world, and the demand for AI servers and chips continues to grow. Quanta predicts that the revenue share of AI servers will jump from 20% last year to over 50% this year.
In response to the growing demand from customers, Quanta is intensifying its production expansion this year, and new production capacity has been gradually opened in overseas factories. Yang Qiling revealed that Quanta's Fremont plant in California and Nashville plant in Tennessee will both contribute new production capacity this year, and the German assembly plant will also start production.
Previously, it was reported that after the release of Nvidia GB200 chips, Quanta successively won the GB200 server OEM orders from Google, Amazon AWS, and Meta, the three major cloud service providers (CSPs). Among them, Google's orders entered testing as early as July to August. It is reported that the average selling price of each cabinet system for GB200 servers is as high as 2 million to 3 million US dollars (approximately 14.45 million to 21.68 million RMB). The GB200 integrates two B200 GPUs with one Grace CPU, and each Nvidia DGX GB200 server system contains 36 Nvidia GB200 supercohips.