Super Micro Computer Inc. (SMCI-Q), which joins the S&P 500 on Monday, enjoys a rare advantage among server makers that are trying to tap the generative AI boom – close ties with Nvidia (NVDA-Q) that help it launch products faster than rivals Dell (DELL-N) and Hewlett Packard Enterprise (HPE-N).
The company has historically been among the first to receive artificial-intelligence chips from Nvidia and Advanced Micro Devices as it helps them check server prototypes, giving it a head start over rivals, analysts and industry experts said.
That has helped turn the company into a key supplier of servers essential for generative AI apps and sent Super Micro’s shares up 289 per cent so far this year.
Some analysts estimate Super Micro, whose California headquarters are less than 16 kilometres away from those of Nvidia and AMD, can within just “a few weeks” manufacture, assemble, test and ship a server rack, if components are available.
“Super Micro has developed a model that is very, very quick to market. They usually have the widest portfolio of products when a new product comes out from Nvidia or AMD or Intel,” said Hans Mosesmann, an analyst at Rosenblatt Securities.
Super Micro’s revenue more than doubled in the last three months of 2023 and analysts polled by LSEG expect triple-digit percentage growth to continue till at least the September quarter of 2024.
Bank of America analysts estimate the company’s share of the AI server market to grow to about 17 per cent in 2026, from 10 per cent in 2023.
That optimism is reflected in Super Micro’s growing market value, which now stands at US$60-billion, compared with some US$5-billion before the launch of ChatGPT in November, 2022.
The AI demand has also lifted shares of Dell by 40 per cent this year, including a gain about 32 per cent earlier this month after an upbeat annual forecast. HP Enterprise shares are down about 1 per cent in 2024.
Both Dell and HP Enterprise did not respond to requests for their go-to-market times for AI servers.
But the massive rally also means that Super Micro is now trading at 40 times expected earnings, far higher than its PE of 15 last December, according to LSEG data.
The expensive PE valuation increases the risk of sell-off if Super Micro fails to meet future earnings expectations.
LIQUID COOLING
Super Micro is also riding on a rush of demand for liquid cooling tech from data centre operators deploying much hotter and power-hungry AI chips, analysts and industry experts said.
They added that the heat produced by chips used in AI servers has more than doubled and is set to double again in the next two to three years.
Super Micro’s tech, developed in-house, involves putting cold liquid over a plate placed on top of a chip and cuts power consumption by about 40 per cent compared with air cooling.
By 2030, liquid cooling is projected to become the dominant method for cooling servers, said Ben Smith, product head at immersion cooling technology provider Green Revolution Cooling.
HP Enterprise and Dell, however, expect the adoption of liquid cooling to be much lower.
Analysts also cautioned it could be tough for Super Micro to maintain its advantage in AI servers as larger rivals shift focus to the market, adding that the company has been offering its products at cheaper rates to gain market share.
Be smart with your money. Get the latest investing insights delivered right to your inbox three times a week, with the Globe Investor newsletter. Sign up today.