Indian startup C2i wants to fix power delivery in AI data centers

Indian startup C2i has raised $15 million from major fund Peak XV Partners. The size of the round signals strong confidence in a technology meant to address one of the most pressing problems in today's digital infrastructure: power shortages in data centers, especially those built for compute-hungry AI workloads. The startup's solution, which it calls "grid-to-GPU," promises to dramatically cut electricity losses as power is distributed inside the data center.

Data center power consumption is not a new problem, but the exponential growth in demand for compute to train and run AI models has pushed it to a new level. Modern graphics processing units (GPUs), the heart of AI infrastructure, draw enormous amounts of power. Meanwhile, traditional approaches to distributing electricity inside data centers lead to significant losses, reaching 15-20% of total consumption. These losses occur at several stages, from voltage transformation to the many converters and cabling systems along the way. With existing grids often unable to keep up with rising loads, and new infrastructure expensive and slow to build, every kilowatt-hour saved is worth its weight in gold.

C2i's technology targets exactly this critical link. Rather than relying on multi-stage conversions and high-loss distribution networks, the company is building a system that delivers power to GPUs with minimal losses. The company has not yet fully disclosed technical details, but the core idea is more efficient management of power flows, possibly using new materials, architectural designs, or intelligent power management systems. The goal is to get as close as possible to the ideal scenario, in which power from the grid flows directly to the GPU compute cores and skips redundant intermediate stages. This not only lowers overall energy costs but may also improve hardware stability by providing cleaner, more predictable power.

The impact of solutions like this could be substantial. First, they directly improve the energy efficiency of AI infrastructure. Lower losses mean lower total electricity consumption, which is critical given global efforts to fight climate change and pursue sustainable development. Second, they could ease the bottleneck created by the physical limits of the power grid. If data centers can run more efficiently, they will need less peak capacity, which could reduce the load on existing grids and delay or even shrink the need for large-scale investment in grid upgrades. Third, more efficient power delivery could extend hardware lifespan and lower operating costs. For AI companies themselves, that means a chance to scale operations faster and at lower cost.

The new funding and C2i's approach mark an important step in building out infrastructure for artificial intelligence. As the industry keeps pushing toward ever more powerful and complex AI models, supplying them with stable, energy-efficient power is becoming just as important as developing the algorithms themselves. By optimizing the grid-to-GPU chain, C2i's technology has every chance of becoming a key factor in overcoming today's energy barriers and opening new possibilities for scaling and deploying AI.