START: Sajan Khosa, CCO & Co-Founder, Liquid Energy "AI inference powered by sunlight"

Listen on Apple Podcasts Listen on Spotify Watch on YouTube Subscribe to START

‍

"What Elon's doing in space, we're doing here on Earth."

That's how Sajan Khosa describes Liquid Energy (YC F26) and its NanoDC: a solar-powered autonomous data center built for AI inference.

NanoDCs are built at a factory and shipped to the site. Each unit generates its own power using solar and batteries, with panels that fold out like an accordion. The compute is cooled with patent-pending high-density air cooling and connects to the internet via Starlink.

No construction. No grid interconnect. No water supply.

Liquid Energy says NanoDC can be deployed anywhere on the planet in four weeks. Units are shipped to the site ready to be turned on in 30 minutes.

Sajan says they can go from the first conversation to live inference in six weeks, compared with four to 12 months for traditional compute arrangements.

Customers don't need to purchase the hardware. They sign a compute contract, Liquid Energy deploys the units, and additional capacity can be added by simply deploying more.

And when asked what breaks first as AI demand scales - GPUs, the grid, or data centers - his answer was immediate: "The grid."

Sunlight in. Tokens out.

‍

🎙️ Sajan Khosa, CCO & Co-Founder, Liquid Energy on Fondo START with Guest Host: Kamran Karic, COO & Co-founder, Ploid

‍

00:45 From conversation to live inference in six weeks

01:32 Turning data centers from a construction

problem into a tech problem

02:28 Eliminating traditional infrastructure costs

03:17 Why GPUs aren't necessarily AI's biggest bottleneck

04:42 Deploying compute in harsh environments

05:14 Zero water consumption and ~85% lower cooling costs

06:29 Developing and testing the cooling technology

07:04 Why Liquid Energy is betting on solar

07:38 Why the electrical grid is AI's biggest bottleneck

Check out www.liquidenergy.world

Check out www.liquidenergy.world

‍