What the PUD's own records actually show.

What actually comes out of the wireless line — and what's left to save

Fair question to ask before we say they could have saved: what has to come out of wireless first? Here's the whole list, straight from the District's own 2025 Wholesale Telecom budget — and to be scrupulously fair to the Board, we charge wireless the entire broadband system's debt, even though that same debt built the fiber network too.

The one thing the budget won't split — and how we stay honest about it: the District runs wireless and fiber with one 8-person crew and one set of network software, and it does not break out wireless's share of that labor. So this ~$782K is not free cash — wireless's slice of the crew still comes out of it. But two facts bound the answer hard: (1) that same crew also runs a fiber/transport business twice wireless's size, and (2) after paying every cost, every salary, and all of that debt, the audited telecom segment still cleared an $808K surplus in 2024. Even if wireless absorbed a full revenue-proportional third of all remaining telecom operating cost, it would still leave far more than the ~$113K/yr (10%) a replacement fund needed. So the answer to "could they have been saving into a capital account?" is yes — unambiguously. The room was always there. (Source: 2025 Wholesale Telecom Proposed Budget — line 810/811 debt service, 712 wireless subscriber units, 085 Canopy site leases; audited statements, Note 9.)

The same waterfall, every year — watch the leftover grow

Now the breakdown for each year: wireless revenue split into debt, reinvestment, the allocated crew & overhead, and what's left to save. The green slice grows — from ~$158K in 2021 to ~$394K by 2025 — because as the District cut reinvestment from $330K toward $26K while revenue and overhead held roughly flat, the room to set money aside kept widening. Every bar shares one scale.

How to read this honestly: the actual, documented parts are wireless revenue and reinvestment — straight from the District's budgets, by year. The crew, the overhead, and the broadband debt (which built the fiber network too) are allocated to wireless by its revenue share — crew and overhead held flat (the budgets show them roughly flat) — so the exact height of those bands is an estimate. But the trend is not an estimate: because revenue, debt, and overhead barely moved, the green "left to save" slice grows almost entirely because reinvestment fell — and that decline is documented fact. Every year clears the ~$113K a 10% replacement fund needed; by 2025 the leftover is more than triple it. One honest edge: part of why the room "grew" is simply that the District spent less and less maintaining the network — so this is the capacity to save and the evidence of under-investment, at the same time. (2024's ~$313K reconciles with the audited surplus-share of ~$270K above; the small gap is depreciation timing. Wages from the 2025 budget; other overhead ~$900K, illustrative.)

And you didn't need a savings rate at all — just bank what was left

This one isn't a hypothetical percentage — it's the actual leftover from the chart above (the green slices), added up year by year. Set that money aside as it came in, and the running total passes the $1.2M replacement cost in five years — with no grant, no loan, and no rate increase:

Cumulative running total of the wireless line's leftover: $158K → $336K → $626K → $940K → $1.33M by 2025, clearing the $1.2M cost. You wouldn't necessarily bank every dollar — but even setting aside half of it funds the replacement inside the equipment's ~10-year life. And if you distrust the allocation model entirely, here's the floor that needs no model at all: 10% of gross wireless revenue is ~$113K/yr — less than the leftover the line produced in every single year above — so a sinking fund was never a stretch. Kept honest: "leftover" is modeled (wireless revenue minus its actual reinvestment, minus a revenue-share allocation of the broadband debt and the shared crew/overhead), so treat the heights as a careful estimate, not a measured wireless margin. The cumulative growth, though, is driven by the documented reinvestment decline. It reconciles with the audited record: 2024's ~$313K matches the ~$270K surplus-share above (gap = depreciation timing), and the whole segment cleared an $808K surplus that year. And a bottom-up rebuild from the District's own billing and budgets lands near this same ~$270K independently — so this isn't margin-by-assumption alone; the one input still worth confirming by records is how the shared crew's hours split.

The line they paused is their largest — and one of only three that grew.

Every recurring revenue category, 2021 against 2025, from the monthly billing. Rural wireless started in a near-tie with the carriers' dark fiber — then the big carrier contracts stepped down as they expired, and wireless has been the District's single largest telecom line every year since 2022.

Full-year billed revenue by category, ex pass-through tax; one-time construction and fees excluded. The carrier declines are contracts stepping down as they expired — a normal shape for wholesale deals — which is exactly the contrast: the wireless line is ~3,000 households choosing it monthly, and it's the category that held. Source: monthly billing reports (FOI), sources/analysis/yearly_revenue_by_category.csv.

What the rural-wireless customers can actually buy — capped at 20 Mbps

Active wireless subscribers by speed plan, from the District's monthly billing reports (May 2026).

The network has never offered a plan faster than 20 Mbps in its entire life. And the 20 Mbps ceiling is the only tier that grew — customers climbing as high as they can and finding nowhere to go — while a majority started on the slowest 3–7 Mbps plans (51.6% in 2021, still 43% today). Wholesale prices never changed in five years. The shelved upgrade is what brings 250–500 Mbps. Why a single rate left money on the table →

Demand didn't fall — it climbed into the ceiling.

Same reports, every month since January 2021: subscribers by plan. The slow tiers empty out as customers move up — and the 20 Mbps tier, the fastest plan the network sells, grew from 635 to a peak of 1,033 even while the total shrank. That is not a market losing interest. That is a market out of headroom.

All 65 monthly “Wireless” reports, January 2021 – May 2026. Blended revenue per subscriber ran ~$29 to ~$31/month across all five years — wholesale prices never rose; revenue held because the mix moved up. This is the demand signal the District's letter never mentions: the only tier that grew is the one at the cap. Parsed dataset: sources/analysis/wireless_tiers_monthly.csv.

Five years on, the District serves more last-mile connections — not fewer.

The letter's premise is a service in decline. Add up every retail connection the District bills — wireless and fiber — and the total grew: 3,315 (Jan 2021) → 3,392 (May 2026). Households didn't leave the public network; some moved from its wireless to its fiber — a migration the District itself builds toward. The question was never demand. It's which technology serves the homes fiber hasn't reached — and that is still 4 out of 5 of these connections.

Units are billed access connections from the monthly reports: wireless service tiers plus fiber retail connection codes (residential and small-business E-codes; carrier circuits excluded). Wireless for January 2023 is interpolated (that report is missing from the District's production). Some of the fiber growth is former wireless customers upgrading — which is the point: they stayed on the public network. Parsed dataset: sources/analysis/bycode_long.csv.

They didn't just cut the budget. They stopped hooking people up.

Every new District-provided wireless hookup bills a $40 install charge (code W05, set by Res 1768). Count them, year by year:

District-billed installs of District-provided subscriber modules — a conservative proxy for new-connection activity (it excludes provider-supplied radios, and Res 1768 waives the fee for legacy-system conversions). The collapse tracks the budget: the “Wireless Subscriber Units” capital line bought 600 radios in 2021, 250 in 2025, 40 in 2026. A network stops growing when its owner stops connecting people to it — and then the subscriber decline gets cited as the reason to stop connecting people to it. Source: monthly “by Code” billing reports (FOI); budget workbooks.

Where the telecom money comes from (2025)

Wireless is the single largest retail line. The big carrier categories (dark fiber, wavelength, business Ethernet) are wholesale deals with national carriers — not the rural retail service this decision affects.

Now break it out by year — and watch wireless on its own

"Telecom revenue" bundles two different businesses: wholesale fiber and transport deals with national carriers, and the rural retail service this decision is actually about. Split each year into its parts and one line stands out — wireless is the District's single largest retail line, about $1.1M a year, steady and slightly growing — while the bigger carrier/transport money is shrinking. So the upgrade isn't competing with fiber for scarce dollars; it's reinvestment in the segment's most stable earner. The bars share one scale, so you can watch the total edge down while wireless holds.

Full 12-month billing by category, 2021–2025 (excludes pass-through tax); each year's segments sum to that year's audited operating revenue. "Carrier / transport" = dark fiber, wavelength, business Ethernet, colocation — wholesale deals with national carriers. "Local fiber" = retail fiber connections + wholesale internet sold to local ISPs. Source: PUD monthly billing reports.

The wireless line earns ~$1.1M a year. Here's what the District put back into it.

This is the comparison that actually matters — not the whole telecom segment, just wireless against its own reinvestment. Revenue barely moved. Capital fell off a cliff: from about $1 reinvested for every $3 the wireless line earned down to $1 for every $14.

Wireless revenue = full-year billing on wireless service tiers and radios. Reinvested = the wireless capital line in the District's own telecom budgets (FOI). The $1.2M upgrade is about one year of wireless's own revenue. One honest caveat: these are wireless revenue and wireless capital — the District has never published the wireless line's standalone operating margin (telecom is reported as one combined segment). That figure is exactly what the records request asks for.

Sixty-five months of wireless billing. Look for the decline.

Every month of wireless billing the District produced, January 2021 through May 2026, in $thousands. The last twelve months on record average $91.7K — higher than the first twelve ($89.8K). Half a decade of a public network doing its quiet, steady job.

Cell = that month's wireless billing in $thousands (darker = more), from the District's monthly reports (FOI), excluding pass-through tax; range $83K–$99K. Kept honest: subscriber counts did decline (−16% from the 2022 peak — the tier chart above shows where and why); the revenue record stayed inside a ±9% band the whole time. Whatever “steady decline” the pause was answering, it is not visible in the District's own billing.

The timing, overlaid: investment fell first, customers followed

Put the two lines on one timeline and the order is unmistakable. Reinvestment (left axis) was cut 34% by 2023 — and that year the subscriber base (right axis) still stood at 3,236 (Oct 2023) — within two subscribers of its all-time high (3,238, Aug 2022). Customers didn't begin leaving until mid-2024 — roughly a year and a half after the network's funding was pulled. And note the gray marker: Starlink arrived in 2021, yet the base grew for three years after. The District's own spending cut set the slide up first; the satellite churn and NCI's RSP decline came later.

Left axis (rust): budgeted wireless reinvestment, by year — the District's own telecom budgets. Right axis (navy, zero-based — no exaggeration): active wireless subscribers, monthly. The reinvestment line is drawn as an annual step. Why this isn't circular: you might think capital fell simply because there were fewer customers to equip — but the timing rules that out. The deepest cut (2023) landed while the customer base was at its healthiest, so that cut was a choice, not a response to lost demand. And the one investment that would have kept the network competitive — the platform upgrade — was never budgeted in any year. The result was a network frozen near 20 Mbps while the alternatives raced ahead; the customer losses came after.

Same towers, same rates, seven different stories.

Monthly wireless subscribers for each reseller, all 65 reports. On the exact same “declining” network, the providers that kept selling it kept growing — Will Connect up 63%, BHPcom up 26% — while most others drifted single digits. That spread is a sales-and-support story, not a technology verdict — and it's exactly why a decision about this platform needed the providers in the room.

January 2021 → May 2026, per-panel scale (each starts at zero). Percentages are first-month to last-month. Two providers grew double digits on aging gear the District had stopped feeding — the demand is there where someone sells to it. NCI Datacom's line reflects a shift in its PUD-wholesale account, concentrated in Dec 2025–Jan 2026; NCI also operates its own network, and the record does not say why its wholesale count moved — so neither do we. Source: monthly “Wireless” report cuts (FOI), sources/analysis/wireless_rsp_tier.csv.

Last in line — by written policy.

When something breaks, whose service gets fixed first? The District's own rate resolution answers in plain text. This is the service-response priority order from Resolution 1768 (February 2022), verbatim:

Res 1768, “Response to service related issues will be offered in order of service type with service priority offered in the following order…” — and, elsewhere in the same resolution, wireless is designated “Reasonable Effort”: connection speeds “cannot be guaranteed.” None of this is scandalous by itself — triage exists everywhere. But when the argument for the pause is that wireless customers drift away from a service that underperforms, it matters that under-service was the written plan: last for repairs, best-effort by definition, and — per the charts above — near-zero for reinvestment. Customers didn't quit a network the District fought to keep sharp. Source: Resolution 1768 (Feb 14, 2022), sources/resolutions/.

Why customers actually leave — and what it would take to keep them

Three of the District's retail providers shared their cancellation records. Start with what the numbers don't say: across all three, Starlink accounts for about 15% of cancellations — the single biggest competitor customers name, but still a minority of why people leave. Most go for ordinary reasons: moves, nonpayment, seasonal vacancy, or no reason given. A full reason-coded log from one of the three (below) shows it plainly: of 174 cancellations, 27 (15.5%) named Starlink — and "left for a competitor" (about a third) overstates the loss, because some of those customers switched to another local provider on the same PUD network, so they never left the system at all. Starlink is real, but it stayed a minority even through its cheapest-ever pricing. The lesson isn't panic or denial — it's that the network proved resilient, and where Starlink is winning, only a modern network holds those customers. Waiting fixes neither.

From three retail providers' own cancellation logs (the chart shows one provider's full reason-coded breakdown). Counts only — no customer is named, and the providers aren't identified here (the source records are held privately for the Board). "Other / unspecified provider" includes customers who didn't name where they went, customers who moved to another local provider on the PUD network (and so never left the system), and cellular (T-Mobile, Verizon) or in-town wired options. Data-quality note: only this one provider's log is reason-coded record-by-record; the other two reported totals. All three exclude NCI (itself ~half the county-wide RSP decline), so the ~15% describes the network's resilient majority — and it's a floor for Starlink switches, since some unattributed departures may also have gone to satellite. One more blind spot: none of the logs codes deaths as a cancellation reason — deceased customers land in "moved away," "no reason given," or "vacancy" — so death-driven churn is invisible in this chart. See the aging-service-area factor below.

What changed isn't how many leave — it's why

Same calendar months, one year apart (so seasonality can't be the cause). The share leaving because they moved barely budged (~21%). What rose is competition — from 15% to 55% of cancellations — though that bucket mixes other local providers (some customers stayed on the PUD network) with cellular and satellite; Starlink itself went 6% → 20% (2 of 33 cancellations → 11 of 56) — the single biggest but still a minority. People aren't suddenly moving more; they now have a reliable alternative, and a frozen network gives them less reason to stay.

Jan–May 2025 (n=33) vs Jan–May 2026 (n=56), one provider, identical months. The 2025 sample is small — read the direction, not the decimal; the full-year figures move the same way.

Seven local providers resell PUD wireless

For most of them, wireless is the majority of what they buy from the District. This is the "RSP wireless activity" the May 4 letter was addressed to.

Full-year 2025, from the monthly "by RSP" billing reports. Across all services, the District's FY2024 audit reports 8 retail providers and 3,800+ end users on the network; the seven above are those reselling wireless.

The cost-benefit, to scale

Set the cost against what the telecom business already earns. Taking the District's full stated $1.2 million at face value, a one-time upgrade is about 1.5 years of the segment's annual surplus and roughly one year of the revenue the wireless line brings in — a single capital decision, not a recurring drain.

At the District's full $1.2M figure, that's about one year of the revenue the wireless line earns and roughly 1.5 years of the segment's $808,166 annual surplus (audited, 2024); the segment's net position was $9,121,260 at 12/31/2024. All source documents are on the documents page — check the math yourself.