Some infrastructure diagrams are more than just technical schematics—they tell a story about what’s possible when cutting-edge computing power meets a mission that truly matters.
This one comes from a global medical research and treatment organization working to change the future of cancer detection and treatment. They’re using what is, frankly, one of the more impressive deployments I’ve seen to date—a rack-scale AI supercomputer designed for one purpose: accelerating life-saving insights.
The heart of the system is a liquid-cooled Supermicro platform running 64 NVIDIA HGX B200 GPUs. That’s over 1 million CUDA cores and more than 11 terabytes of high-bandwidth memory in a single AI system. Connected via a terabit-class, multi-100Gbps network to nine Supermicro-based all-flash servers, each running the Qumulo Data Platform, this setup sustains over 50 GB/sec reads and 42 GB/sec writes—at sub-millisecond latency—across a namespace of 25 to 30 billion files, all durably written and protected across the cluster.
Let that sink in.
Fifty gigabytes per second, across tens of billions of data points, all being used to feed advanced AI models—designed to spot the early signals of cancer metastasis. That means earlier diagnoses. More precise treatment planning. Improved quality of life. And hope—for millions.
This cluster starts with 1.25 petabytes of usable TLC-based NVMe storage but can scale 3x in-rack before tiering seamlessly to a hybrid TLC/QLC-based flash layer, keeping performance high and cost efficient. Every element of this architecture—GPU density, bandwidth, scale, file system throughput—was purpose-built for this workload.
But the technology isn’t the most impressive part.
What matters more than specs is purpose. And this infrastructure has one: enabling a new level of perceptive AI for medical science. One that sees patterns no human ever could. One that gives clinicians the data-driven confidence to intervene sooner. One that may save lives, at scale.
At Qumulo, this is why we build. Not just to handle billions of files. Not just to deliver terabytes per second. But to support the people who are driving forward the edge of what’s possible in medicine, science, and the human experience.
To the team building this system and the clinicians depending on it—thank you. You inspire us every day to keep building amazing technology.
