Supermicro is collaborating with SpaceXAI to construct a massive AI data center designed for gigawatt-scale capacity. The project utilizes Supermicro's modular DCBBS infrastructure and DLC-2 liquid cooling technology to manage dense GPU racks.
SpaceXAI, formed after SpaceX acquired xAI in February 2026, aims to reach over 2 GW of power by the end of 2026 and nearly 10 GW by late 2027. CEO Charles Liang noted that this initiative is expected to be the company's fastest deployment to date.
In September 2026, Supermicro began delivering Nvidia Vera Rubin NVL72 systems to SpaceXAI. These units integrate 72 GPUs per rack to support intensive training for models like Grok, with installations scaling from 5 megawatts to full gigawatt capacity.
Supermicro reported over $60 billion in new quarterly orders for fiscal Q4 2026, driven by demand from hyperscale clients. Consequently, the company increased its gross margin guidance to a range of 15% to 17%.
