How the ASUS Ascent GX10 Is Transforming AI in Higher Education

The need and demand for AI and machine learning programs is booming across the globe, with one 2025 study indicating that the number of Master’s programs in AI has grown almost 170% in U.S. universities in the last three years alone. Undergraduate programs have more than doubled in the past year.

Until now there hasn’t been a hardware solution both affordable enough and sufficiently powerful enough to provide faculty and students with reliable access to the tooling required to learn, develop, and deploy relevant projects and models with billions of parameters without resorting to accessing and renting hardware through the cloud. This presents its own challenges: unpredictable costs, competition for access, higher latency, and myriad compliance risks.

The ASUS Ascent GX10 is designed to solve these problems, and more. It’s affordable, powerful, scalable, and small. A personal AI supercomputer the size of a mini PC, powered by NVIDIA’s groundbreaking GB10 superchip, that brings petaflop-class performance directly to your desktops. With 128GB of unified system memory, faculty and students can comfortably fine-tune models up to 200B parameters. Optionally, connect two GX10s to handle larger models like Llama 3.1, with 405 billion parameters.

ASUS recently traveled to the University of Delaware to personally deliver their first Ascent GX10 unit. There we spoke to Sunita Chandrasekaran, Computer and Information Sciences Associate Professor, and Director of the First State AI Institute. She explained how this hardware fills a critical gap for their research and how it will transform their program. You can watch the full video below:

"We have faculty from coastal science, sport analytics, plant and soil sciences, art conservation—very different workloads," she explains. "Having a system like the ASUS Ascent GX10 means we can throw these datasets onto a powerful local machine and explore predictive models we simply couldn’t run on a laptop." She continues, "There is so much to do… When you have something [like this], the opportunities are endless. We cannot wait to see what kind of cool science these units can open up for us.”

Critically for your university, this keeps costs lower and more predictable. According to publicly-published rates and pricing, for faculty-led machine learning and AI programs, cloud access to capable AI development hardware and GPUs can easily exceed $1,000 per week. Those prices may increase without notice depending on demand, with additional infrastructure costs that must be considered.

Considering many programs average 250-400 lab hours weekly, a deployment of GX10s could feasibly pay for themselves in a single academic term, all while ensuring complete data sovereignty and regulation compliance. You’re in control of your program and your resources – including sensitive training data.

Every Ascent GX10 comes preinstalled with NVIDIA’s DGX OS, with access to the full NVIDIA AI software stack used in larger, datacenter-grade deployments. This means that for a fraction of the cost, your program can offer students and developers tooling and support for powerful frameworks and platforms like NVIDIA NeMo for model customization, Cosmos for physical AI, Metropolis for vision AI application development, Holoscan for AI sensor processing, and NVIDIA Isaac for robotics development.

The GX10’s utilization of the NVIDIA AI software stack means that your university or lab has the option to seamlessly scale their hardware stack up to more powerful workstation solutions like the ASUS ExpertCenter Pro, or even the ASUS AI Pod, without introducing any friction or bottlenecks to their curriculum or development. It’s a transformational foundation for any university program that doesn’t want to be artificially locked into a specific class of hardware or compute.

With the ASUS Ascent GX10, universities can offer hands-on coursework that mirrors real-world AI development, including large-scale model fine-tuning, multimodal workflows, robotics intelligence, and agent-based systems, without being constrained by shared cluster queues or cloud spending limits. All of this, while reducing energy consumption by up to 60% compared to legacy servers often deployed by existing programs.

To help your program evaluate your needs before investing, ASUS has set up a Virtual Lab program. This is a remote, on-demand environment with access to a single or dual GX10 instance for up to 72 hours. This offers a risk-free way to validate performance and test your actual AI workloads and projects on actual hardware, utilizing the same NVIDIA AI software stack you’d have access to with a physical unit. You can submit an application to the Virtual Lab here.

You can learn more about the Ascent GX10 here. Let’s work together to empower the next generation of innovators and creators.