Mueon is developing three-dimensional modules that put compute, memory, interconnects, power delivery and cooling into a single building block for AI data centers. The Portland, Oregon, startup announced a $65 million Series A to move its Cubelets platform from development into broader system validation and production-readiness work.
The premise is physical: data consumes electricity when it moves among chips and other components spread through a server rack. Mueon says stacking those functions more tightly can shorten those paths and limit rack sprawl. It is targeting up to 10 times the power and cost efficiency in the same rack footprint, a goal rather than a demonstrated customer result.
To get there, the company is developing the silicon and memory architecture alongside chip packaging, power delivery, cooling and interconnects. It is also building orchestration software so the modular hardware can work through familiar AI frameworks and deployment workflows. That whole-system design is meant to support different accelerators instead of tying Cubelets to one chip.
That stack is not yet presented as a production system. Mueon said it will use the round to expand engineering and operations, validate the platform and prepare it for production while deepening customer engagements. The announcement names d-Matrix and Ampere as technology partners, but it does not identify paying customers or give a commercial launch date.
Maverick Silicon led the financing, with Intel Capital, Geodesic Alliances Fund, Oregon Venture Fund, M-Ventures, Wellington Capital, IQT and Defined participating. Mueon said it emerged from stealth with a $15.5 million seed round in 2025 and has now raised more than $80 million. Maverick Silicon’s Manish Muthal will join the board alongside Intel Capital’s Srini Ananth.