Bid coverage when only a few subs can do the work

On a data center, bid coverage is not a count of names on a list. It is a count of crews that can be on site at the peak, in your window, under your terms. The electrical and mechanical contractors who can perform at this scale are few, busy and selective. They are choosing which GCs and programs to bid as carefully as the GC is choosing them. A precon team that measures coverage the commercial way, by how many proposals arrived, can have three bids on a package and no real coverage. It can also have one bid and be fine.

This piece is for precon managers, estimators and procurement leads buying electrical, mechanical, controls and fire packages on data centers. That usually means negotiated or design-build programs for hyperscale and colocation owners, often under a master agreement with a work authorization for each phase. The general method for making sure every scope is covered is in the construction bid coverage guide. This is about what changes when the qualified market is a handful of firms.

Why the list is short

The constraint is people, and specifically supervision. On their 2026 earnings calls, leaders of the large public MEP contractors described the same limit in different words. EMCOR's chief executive said the binding constraint is creating enough foremen and project managers. Quanta's chief executive noted that it takes about four years to make a journeyman and saw no oversupply of craft labor coming. Comfort Systems USA's chief executive said labor, not demand, limits what its local units take on.

The contractors' own surveys say the same. In an AGC survey reported by Construction Dive, 37 percent of respondents named worker or subcontractor availability as their biggest data center challenge. The consequences show up in disputes. Consultancy HKA described a 120 MW data center in Texas where the electrical subcontractor forecast a peak of about 150 electricians, staffed around 60 percent of it, and delayed commissioning.

A short list is therefore not a procurement failure to be fixed by inviting more names. The qualified market really is small, and every firm in it is already committed somewhere.

The subs are choosing too

The same earnings calls describe how the large MEP firms decide what to bid.

• Comfort Systems USA said it lays out labor projections for current and pursued work before it bids, stays within what its units can confidently deliver, and prefers repeat customers who are constructive when problems come up (Q4 2025 and Q1 2026 calls).
• Limbach described itself as very selective, screening on customer quality, contract structure and execution risk. It prioritizes owner-direct work and pursues GC-led work selectively (Q1 2026 call).
• MYR Group said more than 90 percent of its customers are repeat clients, and that long relationships are what differentiate it (Q2 2026 call).
• Sterling Infrastructure, a site and underground contractor, said it is not trying to win every project, only the best ones (Q1 2026 call).

At the other end of the market, an HR director at a regional electrical contractor told Construction Dive that the firm has to turn down work to avoid overcommitting, and that its data center travel crews have the highest turnover in the company.

Read together, the message for a GC is plain. The GC is being prequalified by its subs. The terms that decide whether a strong electrical contractor bids are the ones the GC controls:

• payment timing and retention;
• how changes get administered;
• whether liquidated damages are flowed down at a rate the sub can live with;
• whether the schedule is realistic;
• whether there is repeat work behind this package.

Linesight's guidance on hyperscale MEP procurement names prompt payment and manageable retention as ways to attract labor. Seyfarth, writing in Construction Dive, warns that punitive damages terms push subcontractors and suppliers to raise their bids.

Coverage is capacity, not names

On commercial work, the coverage question is whether enough responsive bids arrived for every scope. On a data center, each of the usual coverage questions has a harder version:

The right-hand questions need information the proposal alone does not contain. They need the bidder's manpower curve, a view of its other commitments, and the GC's own record of how much work that sub already carries.

That last point comes up in our own conversations with GC precon teams. Estimators award to the sub with the best number without checking how much open work that trade partner already has with the company. A sub that keeps accepting work does not mean it can staff all of it. Tracking each trade partner's open workload before award is a precon job on a data center. On a smaller building it could wait for operations.

How to level with two bidders, or one

Leveling usually works by comparison: line up the proposals and ask why they differ. With two bidders, comparison tells you little. With one it tells you nothing. The reference has to come from somewhere else.

Level against the project, not against each other. This is good practice everywhere (bid leveling). With few bidders it is the only method that works. Build the scope basis from the documents and the owner's responsibility matrix, then check each proposal against it. On owner-furnished packages, that means the ten responsibility steps for each owner-furnished item.

Use your own history as the second bid. On a campus of near-identical buildings, the best benchmark for building three is buildings one and two:

• labor per switchgear lineup and per generator set;
• change volume by scope;
• commissioning findings.

Filter the history by project type. Data center productivity is not a useful check on a university building, and the reverse is also true.

Check the manpower plan like a line item. Compare the proposed peak headcount, supervision ratio and ramp against the schedule and the scope. A price that assumes a crew the sub cannot staff is not a low price.

Prefer questions to ranking. With few bidders, the purpose of leveling shifts from choosing a winner to making the chosen proposal complete. Every silent or qualified item becomes a clarification to resolve before award.

Know what reopens. Allen Matkins, writing on campus programs, notes that the GC's fee and general conditions are often fixed for longer, while subcontract pricing is renegotiated sooner. Each work authorization is a repricing point, and the basis of the last price is what makes the next negotiation fair.

Build coverage before the bid

On this building type, coverage is won months before a package goes out. Linesight's guidance for hyperscale MEP procurement describes the practices:

• Share a 12 to 24 month pipeline lookahead with core MEP partners.
• Award across multiple buildings or years of a campus.
• Use standard bid forms and consistent scopes.
• Run prequalification once, uniformly, on safety, licensing, bonding and finances.
• Identify regional subcontractors early, to avoid insufficient bidder participation.

Two further practices come from the contractors themselves. An electrical contractor working on large technology-company data centers wrote on Hacker News that his firm almost always teams with a local contractor to reach local union labor. EMCOR has said union affiliation helps because tradespeople travel for per diem. Coverage on a large program is often a partnership structure, not a list.

Job descriptions show how much of this has moved into precon. Current postings for data center preconstruction managers at mission-critical builders list trade partner procurement, bid leveling and normalization, and guidance on market pricing and "available resource." That is capacity intelligence, and it is now part of the precon job.

Share what the department knows

A GC that builds several data centers at once learns a great deal about its trade partners:

• who staffed well;
• whose manpower plan held;
• whose change orders were clean;
• who is committed where next quarter.

In our conversations with precon teams at large design-builders, the ones working this sector describe sharing subcontractor availability and pricing across the whole department. Just as often, they describe precon knowledge that still travels by asking the right senior person.

Both are true at most firms. The first is what a thin market demands. The second is what fails when the senior person is on another pursuit, or when a project team changes and the history of what was agreed with each trade goes with it.

What this means for the precon team

• Measure coverage as credible crews at the peak, per package, not as proposals received.
• Treat the GC's own terms as part of the coverage strategy, because the best subs are choosing.
• With few bidders, level against the scope basis and your own history, and use leveling to complete the proposal, not only to rank it.
• Track each trade partner's open workload with the company before award.
• Keep the basis of every subcontract price, so the next work authorization reprices from facts.
• Make sub availability, performance and pricing a shared record, not one estimator's memory.

Where Piper fits

In a thin market, what matters most is scattered across the company:

• what each trade partner priced last time, and what it actually cost;
• which exclusions a sub always writes;
• whose manpower plan held;
• what was agreed with a sub in a clarification call two phases ago.

Some of it lives in proposals and leveling sheets. Some lives in subcontracts and change logs. Much of it lives in people.

Piper is the AI operating system for preconstruction. It builds an understanding of the project, and of the company's past work, from the documents the team already produces: drawings, specifications, proposals, clarifications, leveling decisions and estimates. When a proposal arrives, Piper checks it against the project's scope basis and against what the company knows from similar work. It surfaces what the proposal includes, excludes or never addresses, with the source behind each finding, so the team can resolve it before award. The team still decides who to buy from and on what terms.

Sources

• Comfort Systems USA, Q4 2025 earnings call and Q1 2026 earnings call (labor planning before bidding, customer preference).
• EMCOR Group, Q1 2026 earnings call and Q2 2026 earnings call (supervision as the constraint, traveling union labor).
• Limbach Holdings, Q1 2026 earnings call (selectivity and owner-direct preference).
• MYR Group, Q2 2026 earnings call (repeat clients).
• Sterling Infrastructure, Q1 2026 earnings call (project selectivity).
• Quanta Services, Q1 2026 earnings call (craft labor supply).
• Construction Dive, Labor squeeze and revised workforce strategies (a regional electrical contractor turning down work) and How top contractors staff data center jobs.
• Consulting.us, As data centers boom, construction risks loom (reporting HKA case studies).
• Linesight, Optimising MEP subcontractor procurement for hyperscale data centre construction.
• Construction Dive, Tips for owners and contractors on data center contracts (Seyfarth; damages terms and bid pricing).
• National Law Review, First shovel: construction contracting in the data center boom (Allen Matkins; repricing on campus programs).
• Hacker News, comment from an electrical contractor on data center teaming.
• Job postings: Data Center Preconstruction Manager and MEP Preconstruction Manager, National Mission Critical.