Are O&M costs rising or falling in 2026?

Core fees have been broadly flat to slightly up with labor costs, after a decade of decline. The structural movement is unbundling: monitoring and analytics increasingly priced separately from field labor, which is healthy — it makes both legible and lets detection quality compete on its own merits.

What changes with in-house O&M?

The fee becomes headcount, and the leak question turns inward: detection and dispatch-efficiency economics are identical, and arguably more important, because there is no contract to renegotiate — only tooling to improve.

How many module cleanings per year are right?

It is a soiling-rate question, not a calendar question. Measured soiling loss against cleaning cost gives a break-even trigger per site; fixed schedules overspend in wet climates and underspend in dusty ones. Platforms that track soiling state continuously make the trigger explicit.

What does a truck roll actually cost?

Direct cost runs $700–$1,500 (labor, vehicle, travel), but the full cost includes the opportunity cost of the technician's day and, for false dispatches, the erosion of alarm credibility. Avoided-dispatch value is therefore usually understated.

Is a cheaper O&M bid ever the right answer?

Sometimes — but only after normalizing scope line-by-line and testing detection capability on your own telemetry. A bid that is 15% cheaper with a detection stack that misses string-level losses costs more in year one than the savings recover in five.

Solar O&M Costs in 2026: Benchmarks per MW and Where the Money Leaks

Virat Damania — Co-Founder & Managing Director, Ellume Technology

The scope of an O&M contract varies more than its price. Here is what the market actually pays — and why the real money is lost outside the contract entirely. Utility-scale solar O&M in North America typically runs $5,000–$9,000 per MW per year for core scope, with full-wrap agreements reaching $12,000–$17,000 per MW-year. C&I portfolios run higher per MW due to scale penalties. The larger economic leak is usually not the fee but unrecovered energy: undetected partial faults and unnecessary truck rolls commonly cost 2–4× more than any achievable fee reduction.

Key Takeaways

→Core-scope utility O&M benchmarks at $5,000–$9,000/MW-year in 2026; full wrap at $12,000–$17,000; C&I portfolios carry a scale penalty of roughly 50–80%.
→Quoted fees are incomparable without a line-item scope map — 'monitoring included' spans everything from a staffed NOC to a login nobody uses.
→The three real leaks sit outside the invoice: undetected energy loss (typically 1–3% of revenue), false-positive dispatch, and reactive major maintenance that predictive signals would have scheduled.
→The correct evaluation unit is cost per recovered megawatt-hour, not cost per MW — a contract five percent cheaper and ten percent worse-directed is a bad trade every time.
→On a representative Ellume fleet plant, one month of dispatch intelligence — five truck rolls avoided, eleven false alarms suppressed, faults clustered into single visits — saved $3,500 against an O&M fee it would take years of negotiation to match.

Every year, procurement teams grind O&M fees down by a few hundred dollars per megawatt and book it as a win. Meanwhile the same portfolio quietly loses several times that amount in energy nobody detected and site visits nobody needed. This piece gives you the fee benchmarks, because you genuinely need them for budgeting and negotiation — and then makes the argument, with fleet numbers, that the fee is the smaller figure on the table.

What does solar O&M cost in 2026?

Two honest caveats before anyone pastes this table into a negotiation. Scope definitions vary so widely that quoted fees are incomparable without a line-item scope map — 'monitoring included' can mean a staffed operations center with engineers or a portal login nobody opens; 'corrective maintenance' may or may not include consumables, crane mobilization, or inverter board-level repair. And regional labor markets move these ranges twenty percent in either direction. Treat the table as calibration, not gospel — and treat any bid that is dramatically below it as a scope question, not a bargain.

Indicative North American O&M pricing, 2026 (per MW-DC per year)
Segment / scope	Typical range	Notes
Utility-scale, core scope	$5,000–$9,000	Preventive + corrective labor, monitoring, basic vegetation
Utility-scale, full wrap	$12,000–$17,000	Adds major-component risk, guarantees, spares strategy
C&I portfolio (1–10 MW sites)	$8,000–$15,000	Scale penalty; travel time dominates
Module cleaning (per event)	$500–$1,500 per MW	Climate-dependent; 0–4 events/year
Unscheduled truck roll	$700–$1,500 per visit	The unit that clustering and triage attack

Where does the money actually leak?

Three places, none of them on the invoice. First, undetected energy loss. Partial faults below availability thresholds — dead strings, chronic derating, soiling past its trigger — typically cost a conventionally-instrumented plant one to three percent of annual revenue. On a 100 MW plant, that is an order of magnitude more than any plausible fee negotiation, recurring annually, compounding silently. The fee buys labor; this leak is about what the labor never gets pointed at, because nothing surfaced it. Second, false-positive dispatch. Alarm systems that cry wolf generate truck rolls that find nothing — at $700–$1,500 per roll — and, worse, they train dispatchers to triage by skepticism, which is how the real faults start waiting too. Suppressing nuisance alarms and clustering genuinely related faults into single visits is worth thousands of dollars per site per month, and it is pure software. Third, reactive major maintenance. An inverter failure caught as efficiency drift or a rising thermal trend is a scheduled repair with ordered parts; caught as a trip, it is an emergency with expedited freight, overtime, and lost daytime energy attached. The cost differential between the two versions of the same failure routinely exceeds the annual per-MW fee of the plant it happened on.

How Ellume Vector attacks the leaks — with the receipts

Vector's analytics page carries an O&M savings section for exactly this reason: the leaks only get managed when they are measured monthly, in dollars, next to the fee they dwarf.

•Dispatch intelligence: five unnecessary truck rolls avoided in the month — eleven raw alarms suppressed as false positives, zero false dispatches — for $3,500 in avoided cost on a single 3 MW site. The prior month, the same engine recovered 9,800 kWh of energy and generated $3,178 in total savings.
•Event clustering: three inverters (INV-044, INV-051, INV-057) faulted with signatures tracing to the same combiner row in one block. Rather than three separate dispatches, Vector clustered all three — plus a pending soiling job on the same block — into a single visit. Four problems, one truck roll.
•Loss surfacing: every active event carries its kWh and revenue impact — the plant's three string outages were costing a combined 16,156 kWh and $1,777 for the period, visible on the executive dashboard, which is what moved them up the work queue.
•Predictive scheduling: per-inverter health scores and remaining-useful-life estimates (fleet average 7.7 years on this plant) convert replacement from emergency procurement into planned CapEx with lead-time pricing.
•Recommendation economics: each AI recommendation ships with impact (kWh/day), effort, time-to-fix, and confidence — the top item on this plant read 14.1 kWh/day recoverable, a 45-minute fix, at 92% confidence — so the O&M manager sequences the week by return on wrench time.

The arithmetic that settles it: that single site's monthly savings — $3,500 in avoided dispatches alone — annualizes to roughly $14,000 per MW: more than the entire core-scope O&M fee for a utility plant, achieved without renegotiating a dollar of it. Fee negotiation compresses a number that is already thin. Detection and dispatch intelligence attack a number nobody was measuring. Only one of those has room in it.

How should owners evaluate O&M economics, then?

Move the evaluation from cost-per-MW to cost-per-recovered-MWh, and put four questions in every O&M RFP and every renewal. What fraction of energy losses does the detection stack actually see, at string granularity — demonstrated on our data, not on slides? What is the false-dispatch rate, measured? What is mean time from fault onset to detection — not fault to trip, fault to detection? And is there a monthly reconciliation of expected versus actual energy, with losses attributed and priced, that both parties see? An operator who can answer those four is worth a premium fee, because the fee was never the big number. The energy is the big number — and it has been, quietly, all along.

Frequently Asked Questions

Are O&M costs rising or falling in 2026?: Core fees have been broadly flat to slightly up with labor costs, after a decade of decline. The structural movement is unbundling: monitoring and analytics increasingly priced separately from field labor, which is healthy — it makes both legible and lets detection quality compete on its own merits.
What changes with in-house O&M?: The fee becomes headcount, and the leak question turns inward: detection and dispatch-efficiency economics are identical, and arguably more important, because there is no contract to renegotiate — only tooling to improve.
How many module cleanings per year are right?: It is a soiling-rate question, not a calendar question. Measured soiling loss against cleaning cost gives a break-even trigger per site; fixed schedules overspend in wet climates and underspend in dusty ones. Platforms that track soiling state continuously make the trigger explicit.
What does a truck roll actually cost?: Direct cost runs $700–$1,500 (labor, vehicle, travel), but the full cost includes the opportunity cost of the technician's day and, for false dispatches, the erosion of alarm credibility. Avoided-dispatch value is therefore usually understated.
Is a cheaper O&M bid ever the right answer?: Sometimes — but only after normalizing scope line-by-line and testing detection capability on your own telemetry. A bid that is 15% cheaper with a detection stack that misses string-level losses costs more in year one than the savings recover in five.