Supermicro says making AI hardware is tricky and low-margin
Briefly

Supermicro says making AI hardware is tricky and low-margin
""due to last-minute configuration upgrade ... with expanded volume.""
""This shift [was] largely caused by the complexity of these new GPU racks, which requires intricate integration, testing and validation, making them more time consuming to source and build,""
""So we had to make sure we ship exactly the best quality, the most reliable system to [the] customer. And that's why we spend a lot of time to burn in our solution. And that's why we build up such a huge capacity,""
Supermicro reported $5 billion in quarterly revenue, down $800 million year-over-year and well below its $6–$7 billion forecast. The company could not count about $1.5 billion of revenue after a customer requested configuration upgrades that delayed recognition. Gross margin fell to 9.5 percent, substantially below peers, driven by a strategic large design win that carried higher costs and lower margins. The margin pressure reflected the effort to ramp a mega-scale optimized rack platform based on Nvidia's GB300, requiring intricate integration, extensive testing and burn-in, and significant capacity build-out. Analysts questioned the recurrence of low-margin projects.
Read at Theregister
Unable to calculate read time
[
|
]