BEAD Construction Starts This Summer. Most Subgrantees Are Not Staffed For It.

For four years, BEAD was a paperwork program. Allocation formulas, challenge processes, subgrantee selection, final proposals. The organizations that did well were the ones with strong grant writers and patient lawyers.

That phase is over. All 56 states and territories submitted Final Proposals, 53 have NTIA approval, and the first BEAD-funded projects are entering construction this summer. Most build activity runs 2026 through 2028, with builds in the largest states, Texas among them, stretching to 2030.

The skill that wins a subgrant and the skill that delivers one are not the same skill. A lot of the awards went to regional ISPs, electric cooperatives, and municipal utilities who have built networks before, but not at this scale, not on this timeline, and never under this compliance regime.

Here is what changes now, and what to check before your first shovel.

BEAD is not a normal build. Funded projects carry obligations that most regional operators have never had to document:

Build America, Buy America sourcing. Domestic content requirements on materials, with documentation to prove it, in a market where fiber and optical components are already supply-constrained by AI data center demand. Sourcing decisions that used to be a procurement conversation are now a compliance conversation, and a waiver request has a timeline attached.

The letter of credit. Before signing a subgrant agreement, awardees post an irrevocable standby letter of credit or a surety bond against the subaward. It can step down as take rates hit thresholds, but for a small operator it ties up real capital at exactly the moment construction costs peak, and the paperwork to reduce it is yours to drive.

Labor compliance, self-certified. The 2025 restructuring removed the workforce development scoring from the application phase, but subgrantees now self-certify compliance with federal labor and employment law. Self-certification is not less exposure, it is exposure with your signature on it and no state reviewer catching problems before they become findings.

Detailed and ongoing reporting. Location-level progress against the specific broadband serviceable locations in your award. This is not a quarterly narrative. It is address-level data, and it has to reconcile.

The 25 percent non-federal match. Most awards require it, and it has to be documented and drawn correctly.

None of these are technically hard. All of them consume engineering and program hours that were not in the original staffing model, because in a normal build they do not exist.

The bigger risk is more mundane.

You won an award to serve several thousand locations across a service territory that was chosen precisely because nobody wanted to build there. That means low density, difficult terrain, long drops, and permitting across multiple jurisdictions with no shared process. Every one of those factors multiplies the engineering hours per passing relative to a suburban overbuild.

Meanwhile, the same team doing that design work is still running your existing network, which did not get smaller when the award came through.

And you are hiring into the tightest network engineering market on record. Enterprise Management Associates put the share of organizations struggling to hire and retain network engineers at 52 percent this year, up from 26 percent in 2022. Every BEAD subgrantee in your state is recruiting from the same pool at the same time, along with every hyperscaler building in the region.

The slip is not dramatic. Design falls two months behind, permitting starts late because permits follow design, construction crews are scheduled against dates that design cannot support, and you pay crew standby while your reporting shows locations passed against a milestone you are going to miss.

Run this before construction, not after the first slip.

1. Hours, not headcount, split by where the work happens. Model the total engineering hours the award requires, then divide them into work that requires someone on site (OSP design walkouts, splice supervision, field validation, construction oversight) and work that does not (transport and IP design, configuration and turn-up, service validation, routing policy, monitoring design, documentation, reporting). A single blended estimate hides the shortage and, worse, hides the fact that these two categories have completely different fill options.

2. Your actual available capacity. Take current team hours and subtract everything committed to running the existing network. In most organizations that leaves 5 to 20 percent, not the number the org chart implies.

3. Permitting lead times by jurisdiction, mapped to design completion dates. Permits cannot start before design finishes. If you have twelve municipalities and one designer, sequence it now.

4. Make-ready and pole attachment exposure. This is where schedules die. Pole attachment regulation is itself contested right now, with NARUC’s Telecommunications Committee passing a resolution in July arguing the FCC lacks authority to preempt state pole attachment rules over policy disagreements. Assume the process is slower than your plan.

5. Material sourcing against BABA, with lead times. Fiber and high-density optical supply is tight because AI data center demand overtook telecom as the primary growth driver for optical cable this year. Order windows that were comfortable in 2023 are not comfortable now.

6. Documentation ownership, named. As-builts are the first deliverable to slip and the most expensive to reconstruct. Someone’s name goes on this or it will not happen.

7. Reporting reconciliation. Confirm your location-level progress data can actually reconcile to the BSL list in your award. If your field systems and your award data do not share a location identifier, fix that before you have 4,000 records.

8. Surge plan for the off-site work, defined in advance. Field crews you will source locally, and that market is what it is. The off-site engineering load is the part you have options on, and it is the part that quietly consumes your senior local people if you do not plan for it. Decide now, not in month four at spot rates with no time to onboard anyone.

Item eight is the one that separates programs that hold their dates from programs that do not, for a reason that is easy to miss.

The instinct during a build is to point every available engineer at the field, because the field is visibly behind. What actually happens is the opposite. The logical work does not go away, so your senior local engineers end up splitting their week between a construction site and a config backlog, doing neither well, and the design queue becomes the thing that holds up permitting, which holds up crews.

Spot contract engineering does not fix it either. A contractor who arrives in month four does not know your network, your standards, or your territory. They spend six weeks learning on your clock and then leave with the context. Repeat that three times over a two-year build and you have paid a premium for discontinuity.

The alternative is to establish the off-site capacity before you need it, with a team that learns your standards during the planning phase and is already productive when the backlog appears.

That is what Mikkena provides. Network engineering from our team in Kosovo, covering the logical layer: design and OIC engineering, provisioning and service activation, access and aggregation, Metro Ethernet and MPLS transport, NOC and ENOC operations, and the migration engineering that moves services onto new infrastructure without outages. We do not send crews. What we do is make sure the people who can be on site are actually on site, instead of clearing a config queue at 9 p.m.