Super Micro: The AI Fever Continues (With No End in Sight)

Those who predicted an imminent AI bubble burst got a cold shower today. Super Micro Computer (SMCI) shares jumped after the close of trading, and this is not just another bout of market volatility. The company published its sales forecast for the current quarter, and it makes one thing clear: data centers around the world want even more capacity, more chips and, most importantly, more server racks. If Nvidia is the brain of the modern AI revolution, Super Micro is its skeleton and circulatory system. Without its racks with advanced cooling, the most powerful GPUs would turn into very expensive space heaters.

To understand why this jump matters, it helps to recall the context of recent months. Super Micro has had a rough stretch. Earlier reports raised questions among regulators, and reports from short seller Hindenburg Research added fuel to the fire, accusing the company of accounting manipulation. Investors began to wonder whether the market was overheated and whether hardware purchases were slowing. They are not. The forecast figures show that demand for AI server solutions remains abnormally high. Companies on the scale of Meta*, Google and Microsoft keep building giant clusters, and Super Micro remains one of the main beneficiaries of this construction project of the century.

The main growth driver is the shift to liquid cooling. Modern Nvidia chips, such as the H100 and the upcoming Blackwell, give off a monstrous amount of heat. Traditional fans can no longer cope, and this is where Super Micro struck gold. Its direct liquid cooling (DLC) systems let operators pack more compute into less space without risking a meltdown of the server rack. That is exactly what cloud providers need as they fight for every square meter in their data centers. The fact that the company looks confidently ahead means the industry's capacity shortage has still not been overcome.

It is interesting to watch how Super Micro manages to hold its lead against tough competition from giants such as Dell and Hewlett Packard Enterprise. The secret is speed. While traditional corporations are still agreeing on specifications, SMCI engineers are already assembling finished solutions. They get first access to new Nvidia chips thanks to the close ties between Jensen Huang and Charles Liang. This tandem effectively dictates the rules of the infrastructure market. If Super Micro says sales will grow, then Nvidia is shipping chips in the necessary volumes, and software companies are ready to pay any price for them.

Of course, risks remain. Dependence on supply chains and political instability can hit production at any moment. In addition, the questions about the transparency of SMCI's accounting have not gone away, and investors will watch every move by management. Still, the market's current optimism shows that the hunger for compute outweighs any corporate scandal. We are in a phase where the infrastructure layer of the AI market still dominates the application layer. While everyone looks for a "killer app" built on GPT, data center owners simply keep buying "shovels" from Super Micro.

The takeaway: AI infrastructure is a marathon, not a sprint. Super Micro's forecast confirms that the big players do not plan to cut hardware spending in the coming years. Get ready for new records in power consumption and data center construction.

*Meta is designated an extremist organization and banned in Russia.