AI Isn’t Waiting. Neither Can the Way We Build.
AI isn’t waiting for utilities. Utilities aren’t waiting for construction. Construction isn’t waiting for design.
And yet too many AI data centers are still delivered as if those timelines run one after another — design, then procure, then power, then build. They don’t. Not anymore.
After several years on the front lines of hyperscale development, I’ve become convinced the biggest bottleneck in AI infrastructure isn’t compute. It’s alignment — getting power, design, procurement, and construction to move as one system, fast enough to keep pace with demand.
The old playbook is breaking
We’re still building data centers the way we did a decade ago, and AI has rewritten the assumptions underneath. Everyone frames this as a story about scale. The real story is density. Racks that sat at 8–15 kW for years are landing at 100+ kW, and we’re actively planning for 150–300 kW. At those densities, air gives up: direct-to-chip liquid cooling, rear-door heat exchangers, and CDU infrastructure become the baseline, while AI UPS architectures and medium-voltage distribution move from edge case to standard. None of this is a future-state slide — it’s being deployed today, which is why “we’ll value-engineer it later” is no longer a safe answer.
Power is the new currency — and currency has a supply chain
“Power is the new currency” is true — but currency only matters if you can get your hands on it. Developers now choose sites by available transmission, utility willingness, and interconnection timelines, often before anything else. When the grid can’t keep up, the project makes its own power: behind-the-meter natural gas, microgrids, fuel cells, gas-to-power, and temporary generation until permanent service arrives. Add real water constraints, and “where can I get power and water, and how fast?” becomes the first design decision, not the last. The teams that win treat rapid interconnection and distributed generation as core scope from day one.
Build for how owners actually think
It’s tempting to make this a construction-methodology debate, but owners aren’t optimizing for elegant construction — they’re optimizing for economics: speed to revenue, utility certainty, cost certainty, scalability, and entitlement risk. Speed to revenue governs the rest. Every week a building sits un-energized is megawatts not monetized — so schedule isn’t a construction concern, it’s the business model. Build for how owners think, and you stop selling a method and start solving for their clock.
Precon stops being a phase and starts being the strategy
If the bottleneck is alignment, the highest-leverage work happens before anyone breaks ground. On our best projects, preconstruction stops being a phase and starts being the strategy — where speed, cost, and certainty are won or lost. That means precon carries real weight: utility coordination, procurement planning, long-lead equipment strategy, modular evaluation, labor-market analysis, schedule modeling, and phased turnover that gets the first megawatts online while the rest of the campus is still rising. Modular and prefab are powerful tools, but only as good as the planning around them — assembled doesn’t beat built unless procurement, power, and labor were sequenced to make assembly possible.
And the quiet constraint: who is going to build all this?
Behind power and water sits a third constraint — people. The buildout assumes a skilled workforce that doesn’t automatically exist in every market these campuses land in. Self-perform and vertically integrated delivery give you control over the trades that matter most; regional workforce planning, prefabrication partnerships, and national labor mobility put qualified people in front of the work when a local market runs short. The firms that can answer “who builds this, and where do they come from?” move faster than the ones assuming the labor will be there.
Why this is exactly where Build Group is built to win
AI compressed time-to-decision; our industry has to compress time-to-deployment to match. That doesn’t happen by running the old sequence faster — it happens by collapsing it, running power, design, procurement, construction, and workforce as parallel, coordinated tracks. That is the work we’re built for.
We’re vertically integrated and self-perform, so the trades that drive density and schedule are ours to control — not subcontract and hope. We treat precon as the strategy, aligning utilities, long-lead equipment, and modular decisions before a shovel moves. And we bring a workforce we can mobilize nationally into whatever market the next campus lands in. Put together, that’s a single accountable team collapsing the timelines everyone else treats as separate — and turning “as fast as possible” into a plan owners can actually bank on.
AI isn’t waiting. Neither are we.
Cody Bedell
President, Build Group
