Peak XV Partners has invested in C2i, an Indian startup targeting the power bottleneck that now constrains AI data center construction, TechCrunch reports.

The investment underscores a shift in where AI infrastructure money is flowing. For two years, capital concentrated on GPUs and networking. Now data center operators face a harder constraint: getting enough electrical power into the building and distributing it efficiently across racks drawing ever-higher loads.

C2i is building technology to address this bottleneck from India, where a deep base of power electronics and semiconductor engineering talent has begun attracting venture attention. Peak XV, formerly Sequoia India and Southeast Asia, is among the most active investors in the region and has repeatedly backed hard-tech and semiconductor-adjacent bets as India pushes its chip and electronics manufacturing ambitions.

The specific round size and valuation were not disclosed in the report.

The commercial logic is straightforward. AI accelerators are dense power consumers, and data center operators report that grid interconnections, transformer supply, and in-rack power delivery — not chip availability — increasingly set the pace of new capacity. Startups that can improve power conversion efficiency or density address a cost line that scales directly with compute deployment.

India's role in this supply chain is growing. The government's production-linked incentive schemes and semiconductor mission have pulled in fabs and assembly plants, while engineering services firms have long served global power and analog chip designers. C2i represents the startup layer on top of that base: a company using domestic engineering to attack a global bottleneck.

Peak XV's backing signals that investors now treat power delivery as a first-order AI infrastructure category rather than a niche within electrical equipment. If power constraints persist as the binding limit on data center expansion — as most operators expect over the next several years — companies improving efficiency at the rack and facility level will compete for a market that grows in lockstep with AI compute shipments.