Web Analytics

Vertical Semiconductor Raises $11M Seed to Revolutionize Power for the AI Era

Vertical Semiconductor, a spin-out from MIT, has secured $11 million in its seed round to commercialize a new class of vertical GaN (gallium nitride) power transistors designed to radically improve power delivery efficiency in AI data centers.

The round was led by Playground Global, with participation from JIMCO Technology Ventures, milemark•capital, and Shin-Etsu Chemical. The funding will accelerate Vertical’s transition from research prototype to scalable production  -  an inflection point few deep-tech startups ever reach.


Breaking the Power Bottleneck in AI

AI’s exponential growth is creating an invisible infrastructure crisis: power inefficiency. Converting electricity from bulk voltage to the fine levels required by processors is incredibly wasteful, producing heat, instability, and massive energy loss. Traditional transistor architectures are struggling to keep up with the density and speed of modern compute systems.

Vertical Semiconductor’s breakthrough lies in stacking transistors vertically rather than spreading them laterally. This 3D structure shortens current paths, improves thermal management, and enables far greater density  -  meaning more power delivered in less space with less energy lost as heat.

The result: up to 30% higher energy efficiency and a 50% smaller power footprint for data center racks. That’s not just a hardware improvement  -  it’s a systems-level leap that could redefine how AI infrastructure scales.


From Lab to Market: The Founder’s Tightrope

Founded out of MIT’s Palacios Group, Vertical Semiconductor was born in a lab, not a boardroom  -  yet it’s now operating in one of the world’s most capital-intensive sectors. Co-founders Dr. Joshua Perozek (CTO) and Professor Tomás Palacios (co-founder and scientific advisor) brought years of semiconductor research into the commercial sphere. Cynthia Liao, MIT Sloan alum and CEO, leads the company’s transition from theory to product.

It’s here where most hardware startups stall  -  and where founders can learn something critical: deep-tech innovation doesn’t just demand superior technology; it demands strategic proof of manufacturability.

This is where the real founder insight lies. In capital-intensive startups, traction isn’t measured by users  -  it’s measured by validation milestones. The founders of Vertical understood that before scaling, they needed to prove their technology could fit existing manufacturing flows. Instead of chasing immediate revenue, they focused on demonstrating compatibility with standard CMOS fabrication, leveraging 8-inch wafers to show that their design could scale inside existing fabs without billion-dollar overhauls.

That single strategic decision did what marketing never could  -  it turned a scientific project into an investable business.

And that’s the ultra-value takeaway for founders in any high-complexity space: investors don’t fund inventions  -  they fund de-risked pathways to commercialization. The smartest founders aren’t always those who invent the future; they’re the ones who make it manufacturable, repeatable, and compatible with today’s reality.


Building for the Next Generation of AI Power

With its newly raised capital, Vertical plans to expand R&D, scale wafer production, and begin early sampling of packaged devices by late 2025. The roadmap aims for a fully integrated power delivery solution by 2026, built to support the extreme loads of next-generation AI processors and hyperscale data centers.

What makes this round particularly meaningful is its investor mix. Playground Global  -  an early backer of deep-tech leaders like Relativity Space and Nervana Systems  -  brings both credibility and a network steeped in hardware-AI convergence. JIMCO Technology Ventures and Shin-Etsu Chemical add manufacturing and materials expertise, ensuring Vertical isn’t just scaling prototypes, but scaling processes.

milemark•capital, known for its focus on infrastructure innovation, rounds out a syndicate clearly aligned on more than capital  -  they’re aligned on industrial transformation.


Why It Matters for the Future of AI Infrastructure

AI models are growing 10x faster than hardware efficiency improvements. Every watt saved at the transistor level compounds into megawatts saved across global server farms. Vertical’s vertical GaN architecture directly attacks that imbalance, promising not just cost savings, but climate relevance.

Its approach could cut energy waste across data centers by double digits, potentially saving billions in electricity and cooling costs annually. For hyperscalers like Google or AWS, those margins translate directly into competitiveness  -  not to mention sustainability compliance.

The real disruption, however, isn’t in power conversion alone  -  it’s in bringing the semiconductor world closer to software speed. By making its technology manufacturable within existing fabrication pipelines, Vertical shortens the distance between research and deployment  -  a rare feat in deep tech.

If the company succeeds in scaling production and maintaining yield integrity, it could define a new class of high-density, AI-optimized power electronics  -  a foundational layer for the next generation of compute infrastructure.


What’s Next for Vertical Semiconductor

Vertical Semiconductor’s immediate focus is to validate large-scale reliability testing, strengthen its engineering team, and deepen partnerships with major chipmakers. The company expects early pilot deployments in 2026, targeting AI server power modules and data center OEMs.

Long term, the team envisions their vertical GaN structure enabling a family of power devices  -  from EV inverters to aerospace systems  -  wherever efficiency and size are mission-critical.

As Cynthia Liao recently shared in an MIT interview, “Our goal isn’t just to make better transistors. It’s to make AI infrastructure itself more sustainable.”

For the semiconductor world, that goal  -  backed by $11 million in new capital and the weight of world-class investors  -  signals that the next leap in AI efficiency won’t come from more chips. It’ll come from smarter powers.



Related Articles