Vertical Semiconductor just pulled off an $11M seed round, and if you’re not paying attention, you’re missing a seismic shift under the silicon surface. This MIT spin-out from Cambridge isn’t chasing the next shiny AI model, it’s rebuilding the power grid inside the machine. The team, led by CEO & Co-Founder Cynthia Liao, CTO & Co-Founder Dr. Joshua Perozek, and Chief Advisor & Co-Founder Professor Tomás Palacios, isn’t tweaking around the edges. They’re tackling the part of AI no one likes to talk about: the ugly power bottleneck that’s frying efficiency and budgets inside data centers.
Here’s the setup. AI data centers are growing hungrier than a pack of GPUs at 3 a.m., power demand is projected to jump 165% by 2030. The heat, the energy loss, the inefficiency, it’s all stacking up. Vertical Semiconductor’s answer is a vertical GaN (gallium nitride) power transistor that cuts energy loss by up to 30% and slashes the power system footprint in half. That’s not marketing fluff. We’re talking real, measurable gains, built on a decade of research from MIT’s Palacios Group, where Perozek literally wrote the thesis on vertical GaN devices.
Playground Global led the $11M round, with JIMCO Technology Ventures, milemark-capital, Shin-Etsu Chemical, and Kastor Capital joining the charge. When a firm like Playground Global, known for spotting deep-tech moonshots before the world catches on, steps in, something’s up. As Venture Partner Matt Hershenson put it, the Vertical team “cracked a challenge that’s stymied the industry for years.” Translation: this isn’t just another chip startup, it’s the key to making AI’s power obsession sustainable.
The company’s roots trace back to MIT’s i-Teams course, where Liao, Perozek, and Palacios saw how vertical GaN could flip data center power from a liability into a performance multiplier. From there, they hustled through MIT competitions, early NSF I-Corp funding, and rebranded from Vertical Horizons to Vertical Semiconductor. The focus never wavered: bring this lab-born breakthrough to life.
Now, with 8-inch wafer demos running on standard CMOS lines, Vertical is moving from prototype to product. Early sampling starts late 2025, and commercial solutions are slated for 2026. The ambition is simple but lethal, powering AI data centers without bleeding watts like a leaky faucet. Their vertical architecture pushes current through the wafer, not across it, enabling higher voltage, faster switching, and cooler performance.
If they deliver, we’re looking at a future where AI infrastructure runs cleaner, faster, and far more efficiently. Less energy waste. Fewer carbon tons. More computational power per kilowatt. It’s the kind of innovation that reminds you the real frontier of AI isn’t just in the models, it’s in the electrons feeding them.
