Why We Invested: Mueon
September 2025
Deepthi Madhava
At Oregon Venture Fund, we love backing big ideas from bold teams. Sometimes, those ideas look less like software and more like…cubes.
As generative AI adoption accelerates, today’s data centers are straining under the weight of soaring performance demands, rising energy consumption, and mounting costs. By 2030, they are expected to devour twice as much electricity as they do today, roughly equivalent to all the power consumed by Japan. Traditional rack-based designs are power-hungry, prone to hotspots, and difficult to scale efficiently, making the need for a new approach clear.
That’s where Mueon comes in. They’re tackling one of the toughest challenges in AI infrastructure with a novel solution: compute cubes. Think of them as Legos for the data center.
Mueon’s proprietary system architecture vertically stacks compute, memory, power, and cooling inside compact modular units, called CUBELETSTM. Mueon CUBELET’s can incorporate processors from any primary architecture family – eg, GPU, X86, and ARM. While 3D packaging and system integration of semiconductors isn’t new, existing approaches have been constrained by thermal management issues, yield, and costs associated with today’s standard 19” NVL72 computing rack, which hasn’t changed much since AT&T introduced the format back in 1922. Mueon’s technology addresses these hurdles, unlocking up to 10x higher performance per watt, with half the latency and cost of a comparable server today.
Today’s data centers weren’t built for the scale and speed that AI demands. Some of the biggest challenges include:
Too much wiring – Layers of switches, cables, and links add cost, complexity, and energy use.
Slow memory access – Compute and memory are far apart, creating bottlenecks and wasting power moving data around.
Inefficient power use – Delivering and regulating energy at massive scale leads to waste and excess heat.
Heat problems – Cooling racks full of GPUs is expensive and limits how dense systems can be.
Hard to scale – Adding capacity means more racks, switches, and cooling infrastructure - making growth costly and space-hungry.
Mueon’s cubes provide a smarter, scalable building block for the next generation of AI and data center systems.
We’re excited about Mueon for a few big reasons:
Dream team – Former Intel fellows and industry experts, including the co-inventor of Intel’s Foveros 3D packaging - with deep expertise in CPU/GPU architecture, memory, and power design, and decades of experience scaling with Moore’s Law.
Early partners – Industry, government, and academia are already collaborating with the team on prototypes. Nearly every major player in the data center ecosystem is involved as either an investor or development partner, or both.
Big vision – A modular architecture with the potential to redefine how compute is built at every level.
If Mueon succeeds, they will not just be another chip startup, they will define a new category in advanced packaging.
We’re thrilled to back Wilfred Gomes, Jack Hwang, and the entire Mueon team in pursuing their bold vision, alongside Intel Capital, Geodesic, and other investors.
The future of AI infrastructure? We’re betting it’s cube-shaped!