Vertical Semiconductor, an MIT spinout, has raised $11 million to develop and commercialize gallium nitride (GaN)-based chips that deliver electricity to AI data centers with greater efficiency, the company said Wednesday.
Backed by Playground Global, the startup is targeting one of AI’s biggest challenges—energy waste during voltage conversion. AI data centers consume vast amounts of power, much of which is lost as heat when transforming high voltages from the grid into the low voltages required by processors.
Vertical’s chips use a stacked vertical transistor design, unlike conventional GaN chips with horizontal layouts, resulting in smaller, cooler, and more power-efficient components.
The company plans to ship prototypes this year and begin commercial production in 2026. The technology, developed at MIT by Professor Tomas Palacios and researcher Joshua Perozek, marks a major step toward improving the sustainability of large-scale AI operations.
CEO Cynthia Liao said the company aims to compete with established players such as Renesas, Infineon, and Power Integrations, all of which are collaborating with Nvidia to rework AI power infrastructure. “We offer a next-generation solution that’s a step-wise transformation, not just small gains,” she said.
