Super Micro Computer Inc, a long-time Nvidia Corp partner and server maker, now derives more than half of its revenue from AI-related server sales, and that’s a potential risk for investors to consider.
On Wednesday, Supermicro SMCI, +5.35% reported fiscal first-quarter revenue of $2.12 billion, just below the high end of its previous guidance, amid tight supply constraints. It also raised its full-year guidance by $500 million to between $10 billion and $11 billion, from a previous range of $9.5 billion to $10.5 billion. The company’s shares jumped 3.7% in after-hours trading, and are up more than 200% so far this year, fuelled by investor excitement over its role in selling AI-enabled servers.
Supermicro, which has worked closely with Nvidia NVDA, +3.79% to offer servers with its in-demand GPUs for AI, also said it is working closely with both Intel Corp. INTC, +2.16% and Advanced Micro Devices Inc. AMD, +9.69% to offer their AI-focused graphics processors, such as Intel’s Gaudi and AMD’s MI300 families.
“We expect many of these products to be widely adopted and expand our share in the accelerated computing market,” Supermicro Chief Executive Charles Liang told analysts in a conference call on Wednesday.
In the first quarter, Supermicro said more than 50% of its revenue came from AI-related servers, with one large data centre customer accounting for more than 25% of its revenue. That was a big jump from February, when Supermicro said more than 20% of its revenue for the full year of 2022 would come from AI platforms. In the fourth quarter, the company said AI accounted for nearly half of its revenue.
This increase in its business coming from AI is a positive, but it could also be a negative if demand ever slows after a huge increase in enterprise spending on AI.
Matt Bryson, an analyst at Wedbush Securities, raised the issue in August after Supermicro’s fourth quarter results. He rated Supermicro a market underperformer, as it was trading at more than 20 times fiscal 2025 estimates at the time. He also cited other concerns, such as “our uncertainty about the sustainability of some of the dynamics benefiting SMCI today (e.g., the longevity of the surge in emerging AI cloud demand)” and the potential for margin compression as component availability improves.
Supermicro offers many additional services to its customers, including server customisation and liquid cooling solutions, which have traditionally been associated with supercomputers. Liang said he expects 20% of its data centre customers to move to liquid cooling, especially as power consumption and thermal challenges increase due to the huge power requirements of AI applications.
But as the company’s revenues become more dependent on the AI boom, it is potentially setting itself up for investor disappointment if that boom gradually slows or, dare we say, suddenly crashes.
This is not just a question for Supermicro, but it can also be asked about a number of companies currently benefiting from the big AI spending spree, most notably Nvidia. In Nvidia’s case, in addition to its now massive datacenter business, it generates some other revenue from gaming and, to a lesser extent, automotive and visualisation. Supermicro also sells standard servers to the IT industry that are not AI-enabled.
Still, as both Supermicro and Nvidia derive more and more revenue from the AI boom, it is a potential double-edged sword for investors.
