Eight months after its founding, C2i Semiconductors has a pitch that fits the moment: artificial intelligence is burning electricity at an unsustainable rate, and the Bengaluru-based startup thinks it can help put out the fire—or at least turn down the heat.
The company announced February 15 that it had raised $15 million in Series A funding, led by Peak XV Partners with participation from Yali Capital and TDK Ventures. That brings total capital raised to $19 million. The thesis underpinning the round is simple, if audacious: data centers running AI workloads waste enormous amounts of energy converting electricity from grid to chip, and C2i believes its "grid-to-GPU" power delivery platform can cut those conversion losses by roughly 10 percent. At scale, that translates to about 100 kilowatts saved per megawatt—real money when power now dominates the operating budget for any serious data center operator.
Whether the company can deliver on that promise remains an open question. But investors are placing their bets at a time when the industry's power appetite looks increasingly untenable.
Power Costs and the Economics of AI Infrastructure
Peak XV Partners has been making aggressive moves in AI infrastructure. Managing Director Rajan Anandan framed the C2i investment around what he sees as a defining shift: power costs, not server hardware, are now the primary expense line item. The firm views C2i's systems-level approach, which rethinks every stage of power delivery from the facility bus to the processor itself, as a potential fulcrum for reshaping data center economics.
Yali Deeptech/Yali Capital, which led C2i's earlier $4 million round in November 2024, returned for the Series A. That initial round also attracted Lip-Bu Tan, who remains listed on C2i's website as both investor and advisor. TDK Ventures joined for this latest financing. None of the investors have issued standalone press releases as of mid-March, perhaps letting the fundamentals speak for themselves.
The Numbers Behind the Urgency
The projections are stark. Bloomberg New Energy Finance estimates U.S. data center power demand could hit 106 gigawatts by 2035, with global data center electricity consumption reaching 1,200 terawatt-hours in the same timeframe. A December 2025 analysis from TechCrunch, drawing on BNEF data, suggested planned data center additions imply nearly triple current electricity demand by 2035. Goldman Sachs Research, for its part, forecasts data center power demand surging 165 to 175 percent by 2030 compared to 2023.
Against that backdrop, even a 10 percent efficiency gain compounds in a hurry. C2i also claims its platform can improve GPU performance by around 3 percent and extend server lifetimes—though these figures remain company assertions awaiting independent validation in production environments.
Building Silicon, One Tape-Out at a Time
What exactly is C2i building? The company describes its offering as a plug-and-play, system-level platform that integrates control, conversion, and packaging across the entire power chain. The architecture targets 800-volt class inputs, a design choice gaining traction among hyperscale AI platforms, and aims to replace the multi-stage power distribution networks that bleed energy at each conversion step.
The startup's roughly 65-person engineering team—led by founders Ram Anant as CEO and Preetam Tadeparthy as CTO—is racing toward tangible proof points. C2i expects two chip designs back from fabrication between April and June, with production handled by Tower Semiconductor in Israel and GlobalFoundries facilities in either Singapore or Dallas. A third tape-out is slated for July.
The company has opened discussions with what it describes as three to four enterprise server customers to help define next-generation platforms. Qualification and validation at data center operators and hyperscalers are planned for later in the year, assuming silicon returns on schedule.
What the Fresh Capital Will Fund
The new funding will accelerate development of the grid-to-GPU system and bankroll customer-facing operations in the United States and Taiwan—two centers of gravity where data center infrastructure decisions get made and chip ecosystems run deep. Bengaluru will remain the engineering hub.
C2i also states it received approval under India's Design Linked Incentive Scheme, a government program backing semiconductor design efforts, though an official beneficiary list confirming the company's inclusion has not been publicly located.
A Crowded Field, a Certain Problem
C2i is hardly alone in chasing power efficiency gains. Vicor, Infineon, Navitas, and Empower Semiconductor are all racing to address similar challenges, deploying vertical power delivery modules and wide-bandgap semiconductors to squeeze waste from the system. The competitive landscape is thick.
But the company has positioned itself at the convergence of two undeniable forces: AI's appetite for electricity is growing exponentially, and the industry will pay handsomely for anything that shrinks the power bill. Silicon expected in a matter of months could validate C2i's approach—or expose it as another ambitious pitch struggling to cross the chasm from theory to production. The hyperscalers, for now, are watching closely. They have little choice.
