To meet the surging demand for AI compute resources, a leading European cloud service provider turned to ASUS for a next-generation infrastructure solution that could help it move faster into the AI market. Within weeks, the provider deployed the ASUS XA-GB721-E2, a rack-scale system ASUS AI POD based on the NVIDIA GB300 NVL72 architecture, achieving one of the fastest Grace Blackwell Ultra delivery outside major hyperscalers.

Project Background

As enterprises, developers increasingly turned to its platform for training and inference services, the provider sought to scale its compute capacity rapidly without compromising efficiency or reliability especially for cloud service providers. Traditional server expansion models were no longer sufficient. The company needed fully integrated rack-scale solutions capable of delivering massive GPU performance while maintaining predictable thermal and power efficiency.

That’s when ASUS stepped in with the XA-GB721-E2 — a high-density, 48U MGX™-compliant rack featuring 72 NVIDIA Blackwell Ultra GPUs and 36 Grace CPUs, interconnected through NVLink™ 5.0 and ConnectX-8 SuperNICs. Designed for liquid cooling and modular serviceability, the system provided exactly what the customer required: extreme compute density, and long-term sustainability.

Why ASUS

Three key strengths defined ASUS’s success in this project:

Dedicated team for a distinguished client

ASUS assembled a cross-functional task force that brought together sales, R&D project managers, and mechanical engineers. The team provided tailored, real-time support throughout the entire process—from system design and validation to shipment—ensuring smooth communication and precise coordination across every stage.

Early collaboration with NVIDIA for engineering excellence

ASUS maintained a deep, long-term partnership with NVIDIA and was involved from the early stages of GB300 NVL72 development. Through this collaboration, ASUS engineers gained early access to technical updates and design insights, allowing them to secure samples ahead of other vendors and complete validation and mass production more efficiently.

Global logistics expertise with reinforced protection

Each XA-GB721-E2 rack was handled with professional care, featuring reinforced packaging and transported using air-suspension trucks to minimize vibration and risk during shipment. Backed by global experience in rack-scale GPU logistics, ASUS ensured the systems arrived safely and ready for immediate installation—delivering reliability beyond performance.

Delivering speed and reliability at scale

Delivery began in September, with over hundred compute trays scheduled for delivery before October—an exceptionally short window for such a complex project. Thanks to pre-validated configurations and synchronized global coordination, ASUS successfully completed one of the earliest rack-scale GB300 shipments in the market, allowing the provider to offer new AI infrastructure services ahead of schedule.

Project Result

Within weeks, the system reached full operation, delivering more than 10 petaFLOPS of compute performance. This immediate boost in capacity enabled the customer to accelerate its cloud service rollout and solidify its position as one of the fastest movers in the next wave of AI infrastructure deployment.

The project exemplified ASUS’s customer-first philosophy: combining engineering expertise, collaborative execution, and dependable delivery. From early engagement with NVIDIA to on-site deployment support, ASUS provided a seamless experience that empowered the customer to achieve ambitious goals within record time.

As AI workloads continue to evolve, the ASUS XA-GB721-E2 with GB300 NVL72 stands as a proven foundation for scalable, sustainable AI infrastructure—demonstrating how close partnership and technical excellence can turn vision into performance.