Quanta AI server shipments exceed optimistic expectations for double-digit growth this year | Industry Hotspots | Industry

Quanta (2382), a major OEM, is receiving orders for its AI servers. Yang Qiling, senior deputy general manager and general manager of Yunda, revealed yesterday (23rd) that the momentum of server shipments this year will reach “more optimistic double-digit growth.” , raised the outlook for server shipments this year, and confirmed for the first time that it has received an order for the GB200 server, “the most powerful AI chip on earth” from NVIDIA. It will be mass-produced in September, and shipment momentum will increase quarter by quarter.

According to industry expectations, Quanta is more optimistic about the momentum of server shipments this year than originally expected, mainly driven by strong order demand from the four major cloud service providers (CSPs) Microsoft, Google, Amazon AWS and Meta.

Yang Qiling pointed out that thanks to the continuous launch of new large language models (LLM), client demand for AI training servers has increased. This year, the proportion of AI servers in overall server revenue will jump from more than 20% last year to more than 50%. , becoming the main product line of the server business.

Yang Qiling said that servers equipped with Huida GB200 are under development. Judging from the current chip progress, the system is expected to be mass-produced and shipped as soon as September. Large CSP manufacturers are very optimistic about the new Blackwell architecture, and the demand is large. In the early stage, The volume of orders released is also not small.

Yang Qiling is optimistic that overall server shipments will increase quarter by quarter this year, maintaining the forecast of double-digit growth in server shipments for the whole year. With the acceleration of GPU iteration, the growth of server shipments this year is expected to reach a “more optimistic” Double-digit percentage.”

Regarding the MGX server, which specializes in enterprise customers, Yang Qiling pointed out that the current customer demand is also very good, but the procurement speed may not be as fast as the GB200 server from CSP factory. It is expected to grow slowly, but it will develop steadily.

As customer demand grows, Quanta’s production expansion plan for the past two years is also in full swing. The amount of capital expenditure this year has increased from 8.8 billion yuan last year, and is estimated to increase to 10 billion yuan, an increase of about 13.6%.

Yang Qiling said that new production capacity at the Fremont plant in California and the Nashville plant in Tennessee will be opened this year, and the German assembly plant in Europe will also be put into production.

ITRI ​​held the 2024 VLSI TSA International Symposium yesterday, during which the Pan Wenyuan Foundation awarded the 18th 2024 ERSO Award. Yang Qiling was one of the four winners. He led Werder to use the advantages of GAI to successfully transform and improve efficiency and productivity. It also integrates QCT (Quanta Cloud Technology) products into next-generation telecommunications to assist the State Grid Center to create world-class AI cloud services.

Further reading

Quanta’s Guangming card positioning robot