Kakinada AI Factory
Our flagship AI Factory. 1.2 GW in three 400 MW tranches: Vera Rubin Ultra NVL576, AMD Helios, and 400 MW held for hyperscaler and enterprise load. 800VDC power architecture engineered around the rack-scale systems from day one. Connected via dedicated fibre along the Chennai–Vishakhapatnam Expressway in partnership with NHAI, anchoring south India's largest planned AI compute footprint.
1.2 GW Planned Capacity
One architecture,
every campus.
Kakinada AI Factory is built on the same power and cooling design as every other HyperNext site, so a workload moves between them without re-engineering.
800VDC to the rack
Two conversion stages instead of six. Around 95% of the power drawn reaches the chip.
How it works →33 kV MVSST
Medium voltage reaches the rack through one solid-state stage. No traditional transformer in the chain.
The design decisions →Direct-to-chip liquid
A sealed glycol loop with dry coolers. No evaporative make-up water.
The cooling system →2,800 kg/m² floor
Specified for Vera Rubin and AMD rack-scale systems, the highest floor loading in India.
Why it was decided first →Sizing a workload for Kakinada AI Factory?
Tell us the rack count, the density and the date you need it live.
