The pitch is straightforward, almost suspiciously so: cut data center energy waste by 30 percent. In an industry where efficiency gains are typically measured in single-digit percentages, that claim would normally invite skepticism. But Claros, a Virginia startup that emerged from the shadows of a defense-focused incubator, has convinced some of Silicon Valley's most discerning investors that it might actually deliver.
The company announced a $30 million seed round on March 19, an oversubscribed deal co-led by General Catalyst and Red Cell Partners. Systemiq Capital, Aero X Ventures, Trenches Capital, and a handful of other backers also joined. The valuation? Claros isn't saying.
What the company will say is that its technology—a combination of chip-level voltage regulators and a reimagined power distribution system—represents a fundamental rethinking of how electricity flows through AI infrastructure. Whether that's visionary or merely clever repackaging of existing concepts will become clearer as engineering samples reach customers later this year.
A Problem Hiding in Plain Sight
Data centers have a power problem, one that's become impossible to ignore as artificial intelligence workloads devour electricity at an unprecedented clip. The International Energy Agency projects global data center consumption could more than double between 2024 and 2030, reaching somewhere near 945 terawatt-hours. In the United States alone, these facilities are expected to account for roughly half of all new electricity demand through the decade's end, according to reporting by Axios in February.
That surge has utilities scrambling, environmentalists alarmed, and data center operators searching for anything that might bend the cost curve. Enter Claros.
The company's approach targets inefficiencies most people never think about—the energy lost as electricity converts from AC to DC, steps down from high voltage to low, and travels across circuit boards before reaching processors. These conversion losses add up. Dan Kultran, Claros' CEO and a former CTO at defense tech firm Epirus, told Axios the full platform could deliver "30% savings in energy" by addressing waste from the grid connection all the way down to individual chips.
It's an audacious claim. But perhaps less so when you consider where Kultran cut his teeth: designing power amplifiers and advanced power systems for applications where efficiency isn't a nice-to-have, it's mission-critical.
Two Products, One Thesis
Claros develops two core pieces of hardware. The first, an integrated voltage regulator, delivers power directly to AI processors rather than distributing it laterally across a board—a seemingly small architectural change that the company says can reduce conversion losses by up to 30 percent compared to legacy designs. By March, Claros had fabricated three IVR iterations and was readying a fourth based on input from at least one unnamed customer already receiving samples, according to Technical.ly.
The second product, what Claros calls its DC-native Power Gateway, is designed to deliver 800 volts of direct current straight to server racks while integrating with batteries, renewable sources, and traditional utility feeds. The goal: eliminate the multiple AC-to-DC conversions that plague conventional systems. Claros pegs the efficiency improvement at up to 20 percent, with total cost of ownership falling by as much as 30 percent over a product's lifecycle.
These are vendor claims, of course, not independently validated figures. But they're specific enough—and the backers serious enough—to suggest something more than vaporware.
"Better power delivery at the chip level is essential to unlocking the next generation of data center performance," Irena Spazzapan, managing partner at Systemiq Capital, said in the funding announcement. Translation: without solving the power problem, scaling AI infrastructure may simply become economically untenable.
From Incubation to Independence
Claros didn't start as a typical venture-backed startup. Red Cell Partners, one of the round's co-leads, launched the company in 2025. That relationship matters—Red Cell specializes in dual-use technologies that serve both government and commercial markets, and data center infrastructure increasingly falls into that category as national security intersects with technological competitiveness.
The company surfaced publicly in February 2025 with a $9.75 million round. The latest injection, which took roughly three months to close, suggests momentum building faster than originally anticipated.
"The power infrastructure supporting AI is one of the most significant investment opportunities of our time," Paul Kwan, managing director at General Catalyst, said in a statement accompanying the March 19 announcement. That framing—infrastructure, not just technology—signals where investors see the real value. This isn't about building a better mousetrap; it's about rethinking the entire supply chain.
Scaling Up, Carefully
Claros now employs roughly 34 people split between a lab facility in Torrance, California, and offices in Northern Virginia. The company plans to add at least another ten over the coming year, according to Technical.ly, though that timeline feels conservative given the capital at hand.
The new funding will expand lab capabilities, accelerate prototyping, and—crucially—move the company closer to manufacturing at scale. Claros has partnered with Samsung for production and previously worked with GlobalFoundries on earlier prototypes. Engineering samples of the IVR are slated for delivery sometime this year, with low-volume U.S. commercial production penciled in for 2027 and high-volume manufacturing by 2028.
Those timelines are aggressive by hardware standards, especially for products that must meet the exacting reliability requirements of enterprise data centers. A single power failure can cascade catastrophically.
An Industry Under Pressure
The timing of this raise reflects broader market forces. Hyperscalers like Microsoft, Google, and Amazon are racing to build out AI capacity while simultaneously committing to sustainability targets that become harder to hit with every new GPU cluster. Startups like Claros represent a potential escape hatch—a way to expand computational capability without proportionally expanding the electric bill.
Whether the technology delivers on its promise remains an open question. But the urgency is real, the capital is committed, and the team has credibility. Claros was recently named to Technical.ly's DC RealLIST Startups and appeared on the 2025 LA Hard Tech 50, recognition that matters less for prestige than for what it signals about industry mindshare.
Kultran and his team are betting that the next frontier in AI infrastructure isn't faster chips or better algorithms—it's simply using less power to run the chips we already have. If they're right, 30 percent isn't just a number. It's the difference between sustainable growth and an unsustainable arms race.
