Nobody sets out to over-build a fab's utility distribution. It happens anyway, on almost every project — and the reason isn't carelessness. It's that manual design under uncertainty systematically rewards adding margin, and nobody is positioned to take it back out.
How the margin stacks up
A fab's utility backbone — power, process cooling, exhaust, make-up air, and the rest — has to be designed before the information it depends on is settled. The process and tool set firm up late, and every change ripples through the load on every system. Meanwhile the design work is spread across MEP consultants, spreadsheets, and CAD, with hand-offs between disciplines and firms.
At every hand-off, the rational move is to protect yourself. The process engineer rounds the tool's demand up. The consultant sizing the distribution adds design margin on top. The engineer laying out panels or mains leaves room for tools that might land later. Each decision is defensible on its own. Compounded across thousands of connections, they produce a distribution that is significantly larger than the demand it serves — and no single person can see the total, because the total only exists across a dozen spreadsheets.
Over-building isn't a mistake anyone makes. It's the sum of everyone sensibly protecting their own hand-off.
Owners pay for this twice. Once in wasted capital — equipment, cable, pipe, and cleanroom space spent serving load that never arrives. And again in rework, because a design frozen early to make the schedule gets rebuilt on paper every time the tool layout shifts.
What an optimizer does differently
Our approach starts from a different place: put every demand, every supply, and every connection in one model, and let an optimizer — not a chain of hand-offs — site and size the distribution.
The Optimizer, the engine inside FacilityConnect, evaluates millions of configurations on every run. For each candidate it checks the full picture: does every piece of equipment stay within its capacity, with diversity applied consistently rather than margin stacked locally? Among the configurations that work, it looks for the one that uses the least equipment and cable — mechanical and electrical, every system and piece of equipment in the distribution.
That last part matters. A human team can check that a design works. What it can't do — not at fab scale, not by hand — is search the space of designs that work and pick the cheapest one. That search is precisely what software is good at.
The proof point
We delivered the electrical distribution design for a leading-edge logic fab this way. The same fab had also been designed manually, which makes for a rare clean comparison — same tools, same demands, same building:
Two hundred panels is not a rounding error. It's floor space, switchgear capacity, installation labor, and maintenance scope that the fab simply doesn't carry — while serving exactly the same tool set within exactly the same capacity limits.
Optimal is a moving target — so re-run it
There's a second half to this that matters as much as the first run: the optimal design changes every time the inputs do. A distribution optimized against March's tool layout is not optimal against June's. Manual processes can't afford to redo the work, so the design drifts away from optimal as the build progresses — usually back toward margin.
Because the Optimizer runs against the living model, re-running it is cheap. Early in the project it works from projected demand and sets the utility basis of design. Once the tool layout stabilizes, it re-runs in detail. When a tool moves after that, it runs again. The design stays optimal as of now, not as of the last time anyone had budget to check.
Today, our engineers run this loop for you: you send the tool layout and the utility demands, and we return the sized mechanical and electrical distribution with the living model behind it. If that's a conversation worth having about your build, we'd like to hear from you.