Key Takeaways

  • When Crusoe planned its Abilene site to build over 200 megawatts of compute capacity, they committed to delivering in one year while competing bids required two and a half years.
  • Vendor lead times for medium voltage power distribution centers reached 100 weeks, stalling data center construction across the market.
  • Crusoe bypassed vendor delays by manufacturing electrical distribution centers in-house, slashing delivery times from 100 weeks down to 28 weeks.
  • Lochmiller relies on the idiot index, evaluating the ratio of raw material costs to finished component prices to decide when to build hardware internally.

The 100-Week Vendor Trap

When Crusoe planned its Abilene project to deploy over 200 megawatts of capacity, standard general contractors told them the schedule was impossible. Every traditional proposal required multiple years just to get switchgear on-site.

“When we set out to build the first two buildings in Abilene, the first it was a little over 200 megawatts of compute capacity, we had sort of committed to doing this in one year and the next closest bid was 2 and 1/2 years,” Lochmiller recalls.

The single biggest roadblock was not securing land, water, or grid interconnects. It was medium voltage power distribution centers. When Lochmiller canvassed traditional electrical vendors, they quoted a 100-week lead time. That meant waiting nearly two full years for a single class of electrical boxes before compute racks could draw power.

“The supply chain to support large scale AI data centers is sort of like a game of whack-a-mole,” Lochmiller says. “At different points different bottlenecks come into play and I think that's actually one of the critical aspects of being vertically integrated is that we're able to navigate a lot of those critical bottlenecks.”

Building Power Hardware in 28 Weeks

Instead of accepting a two-year delay, Crusoe chose to make the equipment internally. They created an internal electrical manufacturing unit, mapped the bill of materials, sourced base parts directly, and assembled the power distribution centers on their own floor.

“We said how quickly could we make this ourselves,” Lochmiller says. “The team did a bunch of work, sourced components to make these things and we were able to do it in 28 weeks.”

Cutting lead times from 100 weeks to 28 weeks allowed Crusoe to hit its one-year delivery target in Abilene. In AI infrastructure, speed to power dictates who wins deals. When top AI labs need gigawatt-scale clusters immediately, waiting two years for electrical assemblies is a non-starter.

Using the Idiot Index to Spot In-House Wins

Lochmiller evaluates whether to outsource or manufacture by looking at the price markup over raw materials. He borrows a mental model from Elon Musk to evaluate every hardware choke point.

“Elon has this very interesting framing of this,” Lochmiller explains. “He calls it the idiot index in terms of what is the cost of raw materials and then what is the end product relative to that cost of just the raw materials going in.”

When a vendor sells a medium-voltage box for hundreds of thousands of dollars despite containing basic sheet metal, breakers, and copper cabling, the index is sky-high. A massive markup combined with an extended delivery delay indicates supplier pricing power, not genuine manufacturing difficulty. For Crusoe, that spread justified building an entire internal manufacturing unit to control their own destiny.

What to Do With This

Audit your primary hardware or software dependency this week and calculate its raw component cost against your vendor's price quote. If an outside supplier quotes a lead time longer than six months with a 5x markup over the underlying parts, map out the bill of materials to build the module in-house before signing their contract.