GB200 is about to debut: The proportion of ODM orders and shipment time are expo

    The most powerful AI chip on the planet is very attractive to server manufacturers.  The AI ​​craze continues to burn.As the market demand for computing power continues to increase,NVIDIA's latest GB200 super AI chip has also joined the battle.There are a total of 9 large manufacturers in the first batch of GB200 supply chain,of which 8 are Taiwanese manufacturers and the remaining one is the US server manufacturer Supermicro.Recently,industry insiders revealed that among the above 9 ODM manufacturers,Hon Hai,Quanta,Wistron,and Supermicro have a higher proportion of orders,while other ODMs have a smaller proportion.In March this year,NVIDIA officially released the most powerful chip GB200 and announced that the first batch of cooperative foundries equipped with the most powerful AI chip include: Foxconn under Hon Hai,Inventec,Pegatron,Quanta's Yunda,Wistron's Wiwynn,ASUS,Aeon,Gigabyte,and Supermicro.It is reported that the GB200 chip is mainly equipped with two B200 GPUs and Grace CPU chips,with extremely large computing performance and has a very high discussion in the market.Recently,it coincides with the peak season of shareholders' meetings,and major manufacturers have also announced their server production lines equipped with GB200 chips and the expected delivery time.  Foxconn delivery time: second half of the year Foxconn has successfully entered the first batch of cooperation list of Nvidia's most powerful chip GB200,and the latest super AI server DGX GB200 is also expected to start mass production in the second half of the year,and the GB200 series products will be shipped in the form of Rack.The current rough estimate of the order volume is as high as 50,000 units,and the estimated revenue contribution can reach NT$2.5 trillion to NT$3.0 trillion.Therefore,Foxconn can be said to be the big winner of the platform generation change,holding three new products of DGX GB200 system cabinets (respectively: DGX NVL72,NVL32,HGX B200).The supply chain has revealed that among the three products launched by Foxconn this time,the DGX NVL72 stands out as the most expensive single unit and the most powerful system.It is equipped with 72 B200 and 36 Grace CPUs (equivalent to 18 servers),along with 9 switches,boasting ultra-high performance.Although the entire cabinet is designed by Nvidia and cannot be modified,Foxconn has taken over the production of the cabinet entirely.Looking ahead,with the assistance of two major advantages - liquid cooling technology and the electric vehicle revolution,Foxconn is expected to enter a golden takeoff period by 2025.

Wistron's Shipment Time: Fourth Quarter

Regarding the shipment dynamics of the AI server products equipped with the GB200 chip under Wistron,Chairman Hong Lining admitted on the 24th that the company is still working hard,but will definitely keep up with their progress,hoping to show results in the fourth quarter at the earliest.In the future,the company will also follow the progress step by step and will never fall behind.In addition,Hong Lining also invited everyone to visit the company's booth at the Nangang Exhibition Hall on June 4th,where the ultra-strong AI server GB200 NVL72 will be displayed in person.

As a close partner of Nvidia,Wistron is not only one of the manufacturers that have taken an early position in the supply chain,but also one of the first companies in the global market to be equipped with the latest super chips.Looking ahead,Chairman Hong Lining emphasized that the revenue of the company's AI servers has maintained a quarter-on-quarter increase,and it is expected that by the end of this year (the end of the fourth quarter of 2024),the proportion in the overall revenue will exceed 50%.

Inventec's Shipment Time: End of the Year,Early Next Year

Inventec's CFO,You Jinbao,said that it is still too early to talk about shipment performance.The GB200 product line under the company is not yet ready,and the entire product is not yet prepared,let alone the actual shipment dynamics.In fact,the product is still in the development process,and predicting future shipment performance is still too early.He emphasized that the B100 product line will be shipped at the earliest by the end of the year,and it is estimated that the GB200 will not be mass-produced and shipped until the end of the year to the beginning of next year.

The proportion of servers under Inventec is about 40-45%.In recent years,the company has gradually expanded its business from the brand OEM market to the first-line customers of cloud CSP,and the proportion of CSP business in the department has increased to 30%.The company is also optimistic that as the market shortage of materials has gradually eased,future order momentum will show a leap in growth,which will also drive the shipment to increase several times a year.The proportion of server performance is also expected to increase to more than 10%,and it is directly stated that the growth in server operations will become the main growth momentum of the group this year,and is expected to achieve double-digit growth.

Quanta's Shipment Time: SeptemberIn response to the shipment dynamics of GB200,Yang Qilin,Senior Vice President of Quanta and General Manager of WND,recently pointed out that the GB200 product is expected to enter mass production and shipment in September.Looking forward to future development,he revealed that the order range of his server orders has covered the four major CSP companies,and the total order volume is not small.This also makes the demand for AI servers this year more optimistic than expected.He estimates that the overall server shipment performance this year is expected to improve quarter by quarter,which will also contribute to the annual server revenue,striving for double-digit growth.

Yang Qilin also said that WND has implemented NVIDIA MGX architecture in its own QuantaGrid and QuantaEdge product lines,as well as pioneering progress such as GB200 NVL72 and Blackwell,in conjunction with BlueField-3 DPU,NVIDIA Quantum-X800,and Spectrum-800,we are ready to provide a series of products with unprecedented performance,efficiency,and security.

Super Micro's shipment time: Next year

Super Micro,a major supplier of servers and storage,has taken a big bite of NVIDIA GB200 business opportunities,estimating that it will ship more than 10,000 AI servers equipped with GB200 next year,accounting for as much as 25% of NVIDIA's overall GB200 cabinet share.It is reported that in the supply chain of Super Micro AI servers,Shuanghong is mainly responsible for supplying water cooling plates,Nideco Super is entered the heat dissipation supply chain through its parent company Nidec,Huatai is responsible for some OEM business,and U-Group supplies the server's LONG-DIMM.

Recently,Super Micro has started to increase the pressure on the supply chain to ensure smooth mass production and shipment in the second half of the year.Looking forward,Super Micro President Liang Jian Hou pointed out that he is optimistic about the development of the entire liquid cooling market,and in the future,we will continue to assist customers in introducing AI into data centers through the latest technologies (including overall liquid cooling solutions),while helping the supply chain to promote the latest GPUs and accelerators to the market.