This post was originally published on this site
Executives at a Silicon Valley computer and server maker sent a message late Tuesday when they said the company needs more artificial-intelligence chips from Nvidia Corp.
The commentary confirmed that there are supply constraints for Nvidia’s popular graphics chips, which suggests that even as companies are excited about AI applications, they may have trouble immediately getting the hardware they need.
Shares of both Nvidia
NVDA,
and Super Micro Computer Inc.
SMCI,
— two companies that have surged on the AI boom — dropped Wednesday, with Supermicro tumbling nearly 23%, while Nvidia lost just over 3%. Their shares had surged about 323% and 206%, respectively, through Tuesday’s close, and have racked up sizable gains on the year.
Analysts were concerned about Supermicro’s lower-than-expected outlook for its September quarter, which suggests the company’s revenue-growth rate could be as low as in the single digits. Supermicro forecast a revenue range of $1.9 billion to $2.1 billion, indicating a year-over-year growth range of 5.5% to 22%.
That comes after the company reported June-quarter revenue growth of 34%, amounting to $2.18 billion.
Also read: AI has given a boost to a lesser-known Silicon Valley computer maker
Supermicro Chief Executive Charles Liang told analysts on the company’s call that the company is seeing supply constraints for the graphics chips coming from Nvidia that are designed for AI applications. One analyst asked about Nvidia’s past July-quarter projections for a doubling of its data-center revenue and whether there is a lag time between when Nvidia recognizes revenue and when equipment makers like Supermicro recognize revenue.
“We believe their capacity [is] growing and that’s why we talk to them every day, asking for more,” Liang said. “So hopefully we can gather more support from them and hopefully their capacity grows more quickly. So that’s all I can say now.”
Chief Financial Officer David Weigand also noted that Supermicro saw 52% of its revenue in servers coming from server racks designed specifically for AI-based applications, which was a jump from about 30% in the previous quarter.
Aaron Rakers, a Wells Fargo analyst who follows Nvidia but not Supermicro, wrote up a brief note highlighting the supply constraints that the computer maker was seeing. He noted that it implied a three-times sequential revenue increase in Nvidia’s quarter, for which the company will post results on Aug. 23.
While Supermicro is seeing very strong demand for its AI-designed servers, the company also said that overall data-center demand is otherwise declining. When asked about the company’s non-AI business and what the spending was like for non-AI-related servers, Liang said Supermicro’s other sales were about flat, while the overall industry was declining.
Investors will now be focused on Nvidia’s report later this month. The company forecast a mammoth $11 billion in quarterly revenue, but whether supply issues are going to affect results going forward is now a question on everyone’s mind.
