They say that the most difficult part of transportation planning is last-mile delivery. A network of warehouses and trucks can bring products within a mile of almost all customers, but logistical challenges and costs add up quickly in the process of delivering those goods to the right doors at the right time.

There’s a similar pattern in the AI space. Massive data center installations have empowered astonishing cloud-based AI services, but many researchers, developers, and data scientists need the power of an AI supercomputer to travel that last mile. They need machines that offer the convenience and space-saving design of a desktop PC but go well above and beyond the capabilities of consumer-grade hardware, especially when it comes to available GPU memory.

Enter a new class of AI desktop supercomputers, powered by ASUS and NVIDIA.

The upcoming ASUS AI supercomputer lineup, spearheaded by the ASUS ExpertCenter Pro ET900N G3 desktop PC and ASUS Ascent GX10 mini-PC, wield the latest NVIDIA Grace Blackwell superchips to deliver astounding performance in AI workflows. For those who need local, private supercomputing resources, but for whom a data center or rack server installation isn’t feasible, these systems provide a transformative opportunity to seize the capabilities of AI.

Scaling up memory to meet the parameter count of large AI models

A key piece of the puzzle for accelerating locally run AI workloads is available GPU memory. If a given model doesn’t fit into local memory, it may run very slowly, or it may not run at all. The 32GB of VRAM provided by the highest-end NVIDIA consumer-grade graphics card on the market, the NVIDIA GeForce RTX 5090, is sufficient for many smaller models. But scaling up your system’s VRAM to handle models with even more parameters isn’t necessarily a straightforward affair.

Multi-GPU systems are a feasible solution for some users, but others have been looking for a solution that’s designed specifically for the needs of AI workflows. By equipping the Ascent GX10 and ExpertCenter Pro ET900N G3 with large single pools of coherent system memory, we’re able to put astonishing quantities of memory at your fingertips. The Ascent GX10 wields four times as much GPU memory as a GeForce RTX 5090, while the ExpertCenter Pro ET900N G3 offers up to 784GB — over twice as much GPU memory as a workstation equipped with four NVIDIA RTX PRO™ 6000 GPUs.

AI supercomputer performance in a desktop PC form factor

Designed from the ground up for AI workflows, the ASUS ExpertCenter Pro ET900N G3 will be one of the first pioneers in a new class of computers, based on the NVIDIA DGX station.

This system is powered by the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip. Featuring an NVIDIA Blackwell Ultra GPU and an NVIDIA Grace CPU connected via the NVIDIA® NVLink®-C2C interconnect, this superchip provides a slice of data center performance in a desktop workstation. Even more so than today’s high-end desktop systems, the ExpertCenter Pro ET900N G3 ensures that businesses and researchers can develop and run large-scale AI training and inference workloads thanks to up to 784GB of large coherent memory.

It all runs on the NVIDIA AI Software Stack including NVIDIA DGX OS, a customized installation of Ubuntu Linux purpose-built for optimized performance in AI, machine learning, and analytics applications, with the ability to easily scale across multiple NVIDIA DGX Station systems.

The AI supercomputer in the palm of your hand: the ASUS Ascent GX10

The ASUS ExpertCenter Pro ET900N G3 is much easier to deploy than a solution based on rack servers, but there are situations where even a desktop-class form factor is still too large. The ASUS Ascent GX10 democratizes AI by putting petaflop-scale AI computing capabilities in a design that you can hold in the palm of your hand.

The ASUS NUC lineup demonstrates our proven expertise in offering complete PC experiences in ultracompact designs. No mere iterative step forward, the Ascent GX10 takes our experience in the mini-PC market and melds it with the groundbreaking performance of the NVIDIA GB10 Grace Blackwell Superchip. This superchip connects a Grace CPU with 20 Arm cores with a robust Blackwell GPU through NVIDIA® NVLink®-C2C technology. All told, it delivers up to 1,000 AI TOPS of processing power, with 128GB of coherent unified system memory allowing the system to handle 200 billion parameter AI models.

Need the Ascent GX10 to handle even larger models, such as Llama 3.1 with its 405 billion parameters? Integrated NVIDIA® ConnectX®-7 Network Technology allows you to harness the AI performance of two Ascent GX10 systems working together.

Part of a complete AI solution set

ASUS stands out from every other manufacturer on the market with the breadth of AI products that we’re able to offer. The ExpertCenter Pro ET900N G3 and Ascent GX10 slot into a complete lineup that meets the needs of AI enthusiasts at every level. For those looking to build their own AI PC out of consumer-grade components, for those who need AI performance built into their everyday laptop, for enterprises who need a single-rack AI server solution, even for those institutions looking to design, deploy, and operate a data center for AI applications, the ASUS product portfolio is ready.

Yet the ExpertCenter Pro ET900N G3 and Ascent GX10 are far more than mere additions to our AI product stack. The jump from AI PC to AI supercomputer is nothing less than revolutionary, and these systems give you this level of performance in a complete turnkey solution that fits on a desktop.

Aspects of these systems are still in development, but we’ll share more details as soon as we’re able. In the meantime, explore how ASUS can help your organization seize the capabilities of AI.