Supermicro posts record $60B orders as SpaceX leads 'neocloud' shift

Supermicro, a server hardware manufacturer based in the United States, reported preliminary fourth-quarter results indicating a record US$60 billion（約9.6兆円） in customer orders. According to the company's disclosure, a significant portion of this order book is tied to SpaceX, Elon Musk's aerospace and space-exploration company, and to other organizations building custom AI infrastructure systems. The company's bet on rack-scale AI infrastructure — specialized, modular server configurations designed to support large-scale artificial intelligence workloads at scale — is now translating into both volume and profitability. This marks a shift for Supermicro from merely accumulating orders to demonstrating that its strategy is generating actual profit growth. The emergence of what industry observers call "neocloud" buyers reflects a broader trend in how organizations approach AI. Rather than relying exclusively on hyperscalers (large cloud providers such as Amazon Web Services, Google Cloud, or Microsoft Azure), companies like SpaceX are opting to design, build, and operate their own AI infrastructure customized to their specific needs. This approach allows these organizations greater control over their technology, lower latency for their applications, and potentially lower per-unit costs at scale. However, Supermicro's preliminary results also came alongside softer revenue figures, suggesting that while the order backlog is substantial, conversion of those orders into recognized revenue may take time. The full fourth-quarter revenue disclosure will provide clarity on the pace at which Supermicro can fulfill this order book and convert it into profitability.

Supermicro's record US$60 billion（約9.6兆円） order backlog reflects a fundamental shift in how organizations are approaching artificial intelligence infrastructure. Historically, companies relied on hyperscalers (Amazon, Google, Microsoft) to provide cloud-based AI services. The emergence of "neocloud" buyers — led by SpaceX — indicates that some of the largest and most sophisticated organizations now prefer to design and operate custom AI systems optimized for their specific workloads. This shift favors specialized hardware makers like Supermicro, which sell rack-scale AI infrastructure (modular, high-density server configurations tailored for large-scale AI training and inference) rather than competing directly with cloud giants on breadth of service. The concentration of orders around a few major players like SpaceX, combined with the scale of the backlog, suggests that this trend is not marginal but represents a material redirection of capital away from traditional cloud toward custom infrastructure — a transition that benefits Supermicro's business model and profitability, not just its order volume.