Field notes
Notes on AI data center strategy, renewable compute and what it takes to bring 10-100 MW of AI capacity online in Bhutan.
Global AI compute demand is outrunning conventional data center delivery. Bhutan's own AI strategy, renewable hydropower base and phased development model make the country a serious candidate for the next wave of AI infrastructure.
Read the note
GPU-as-a-Service, reserved AI clusters, bare metal, wholesale capacity and colocation aren't the same decision wearing different names. Here's how to match an AI compute purchasing model to the workload — and how to diligence any provider's claims.
Read the note
AI GPU racks now run hotter than air cooling was ever built to handle. Here is why liquid cooling for AI data centers has become a design requirement, and how Rinchen's double-loop system is engineered around it.
Read the note
A 10 MW pilot doesn't earn its way to 100 MW on a forecast — it earns it by clearing five stop/go decision gates, one verified phase at a time.
Read the note
A Bhutan hydropower AI data center isn't just a renewable-tariff story. It's a grid-reliability story, a phased-delivery story, and a sovereign-compute story — and the three are built to reinforce each other.
Read the note
Sovereign AI is usually told as a story about the largest economies. It's just as real for a small, hydropower-rich nation with a national AI strategy and a phased plan to build sovereign AI compute at home.
Read the note