The NVIDIA DGX Spark desktop supercomputer is revolutionizing AI research and education at universities worldwide by delivering data-center-grade performance in a compact, accessible form. Designed for local deployment, each unit leverages the NVIDIA GB10 superchip and the DGX operating system to run AI models of up to 200 billion parameters, enabling researchers and students to work with large language models and complex AI systems without relying on the cloud. At the University of Wisconsin-Madison’s IceCube Neutrino Observatory in Antarctica, DGX Spark powers AI analysis of neutrino data in one of the most extreme environments on Earth. With no local hardware suppliers, limited power, and harsh conditions, the system’s reliability and low footprint make it ideal for remote scientific work. Researchers use it to detect signals from cosmic events like supernovas and dark matter, expanding the reach of astronomy beyond traditional light-based observations. At NYU, the ICARE project uses DGX Spark to evaluate AI-generated radiology reports in real time. By running large language models locally, the system ensures patient data stays on-site while enabling continuous, interpretable feedback on AI performance. This allows researchers to rapidly refine tools for clinical use, improving accuracy and trust in AI-assisted diagnostics. Harvard’s Kempner Institute employs DGX Spark to study the genetic roots of epilepsy. Scientists analyze thousands of brain neuron mutations, using the system to simulate protein structures and predict how genetic changes affect neural function. The desktop supercomputer serves as a bridge between initial experiments and large-scale cluster computing, streamlining the research workflow. Arizona State University has deployed multiple DGX Spark units campus-wide, supporting AI initiatives in memory care, transportation safety, and sustainable energy. One team led by Professor Yezhou “YZ” Yang uses the system to develop AI-powered robotic dogs for search-and-rescue missions and assistive tools for visually impaired individuals. Mississippi State University integrates DGX Spark into its computer science curriculum, giving students hands-on experience with cutting-edge AI hardware. The system has inspired student-led outreach, including an unboxing video that highlights its role in building the next generation of AI engineers. At the University of Delaware, the First State AI Institute uses DGX Spark to run large models across diverse fields—from sports analytics to coastal science—without the high costs of cloud computing. The ASUS Virtual Lab program allows institutions to test performance remotely before adopting the system. In Austria, the Institute of Science and Technology (ISTA) uses an HP ZGX Nano AI Station, based on DGX Spark technology, to train and fine-tune large language models on a desktop. Their open-source LLMQ software allows full model and data retention in 128GB of unified memory, eliminating the memory bottlenecks common on consumer GPUs. Stanford researchers use DGX Spark to prototype complete AI pipelines for biological agent workflows, testing and benchmarking models locally before scaling to larger clusters. Performance benchmarks show it matches cloud GPU instances, achieving around 80 tokens per second on a 120-billion-parameter model using MXFP4 precision. DGX Spark is also central to Treehacks, Stanford’s global student hackathon, where participants will work directly with the system from February 13–15. The event highlights how accessible, powerful AI hardware is empowering students to innovate at scale. With its blend of performance, portability, and privacy, DGX Spark is transforming how universities conduct AI research, teach next-generation skills, and solve real-world problems—from the South Pole to the classroom.