AGNIT CEO on why AI data centres need gallium nitride chips
Everyone's busy drooling over NVIDIA's latest H100s, convinced that pure compute muscle is the only bottleneck in the AI race. But let's be real: all that silicon glory is useless if your data centre is guzzling electricity like a frat boy at a kegger, melting down faster than an ice cube in August. The real unsung hero isn't just about processing zeroes and ones, it's about getting the *power* to those zeroes and ones without literally setting the building on fire. If you're not thinking about energy efficiency, you're not thinking about AI infrastructure, you're just building an expensive space heater.
The critical insight, as Hareesh Chandrasekar, co-founder and CEO of AGNIT Semiconductors, articulated at the Bengaluru Tech Summit, is that AI data centres demand more than sheer processing capability; they require profoundly efficient electricity delivery. Gallium nitride (GaN) chips offer a compelling solution by significantly improving power conversion efficiency. Unlike traditional silicon, GaN can operate at higher voltages and temperatures, switching much faster and with lower resistance, thereby reducing energy loss and heat generation within power supply units. This translates directly into less wasted energy, lower cooling costs, and a smaller environmental footprint, making GaN a pivotal technology for scaling AI infrastructure sustainably.