AI factory engineering, by the engineers who build them.
Drybulb publishes rigorous technical writing on AI infrastructure — power systems, liquid cooling, networking, reliability, and sustainability — alongside a growing library of practical engineering tools. Written for the owners, investors, and engineering teams making high-stakes infrastructure decisions.
From the writing
Every article is built on first-principles diagrams — drawn, not stock.
Legacy AC vs. 800VDC power chains — grid to GPU
The 800VDC Rollout→
Latest articles
All articles →Before You Fund an AI Factory: The Engineering Questions That Decide the Return
An AI factory's returns depend on a few engineering answers: firm power, rack density, cooling and water, and who is responsible for what. Answered late, they come back as change orders, schedule slips or stranded capacity. These are the questions investors should ask first.
The AI Factory's Nervous System: The Case for Converging IT and OT
The power and cooling are being reinvented for the AI factory; the data layer that ties them together is still treated as an afterthought. The case for converging IT and operational technology onto one open, real-time, event-driven telemetry-and-control fabric — grid to GPU.
Certified, Self-Qualified, and Still Untested: The Case for Project-Specific BESS Testing
AI factories are making the battery load-bearing. A stack of certifications — and even NVIDIA's own self-qualification — won't tell you whether this BESS will hold your grid. Why project-specific testing is the layer of assurance no one can outsource.
The 800VDC Rollout: How the AI Factory Power Architecture Is Taking Shape
Everyone agrees 54V is dead above a megawatt. Inside the monopolar-vs-bipolar split, the reference designs, and the supplier ecosystem powering the next generation of AI factories.
Engineering tools
Free, open tools for data center and AI infrastructure engineers — a climate-based PUE calculator and a build-cost model, with more on the way.
PUE Calculator
Estimate Power Usage Effectiveness from IT load, cooling, and overhead inputs. Compare against industry benchmarks for traditional and AI-dense facilities.
Data Center Cost Model
Model the all-in cost to build and run AI-factory capacity — facility capex by discipline, grid vs. on-site gas, annual opex, and levelized $/MWh.