In-House vs Outsourced Solar O&M: Choosing an O&M Strategy for a Utility-Scale Portfolio

Most comparisons of in-house and outsourced solar O&M are built on a comparison that doesn’t hold. On one side, a technician’s salary. On the other, a contractor’s annual fee per kWp. Subtract, and the in-house column usually looks cheaper.

It looks cheaper because the salary line is the only cost being counted, and a technician is not an O&M capability. A technician is one component of a capability that also includes fault detection, engineering escalation, dispatch coordination, spare parts, vehicles, calibrated instruments, certification, safety systems, records that survive an audit, and cover for the days when the technician is ill, on holiday, or already at another site when a second fault opens.

The real question isn’t whether a salary is cheaper than a fee. It’s what level of operational capability your portfolio actually requires, and what the most efficient way to provide it is. Sometimes the answer is to build it. Often it’s to buy it. Increasingly, for portfolios in the middle, it’s to split it.

This article sets out how to work out which applies to you.

The two terms are used interchangeably in the market and they shouldn’t be. The distinction matters here because owners frequently decide to “keep O&M in-house” when what they actually mean, and what usually makes sense, is keeping asset management in-house.

Technical O&M is the physical and operational layer: monitoring and alarm triage, preventive maintenance, corrective maintenance, fault diagnosis, spare parts, HSE on site, and the reporting that evidences all of it. It is a 24/7, geographically-bound, equipment-specific function.

Asset management is the commercial and financial layer: PPA and offtake administration, insurance, lender and investor reporting, budgeting, warranty claim strategy, capex decisions, and end-of-life planning. It is an office-based, portfolio-level function.

The SolarPower Europe O&M Best Practice Guidelines — now in their sixth edition, and the closest thing the European market has to a common standard — treat these as separate disciplines with separate guidance documents. They’re separable in practice too, and they split along very different lines when you’re deciding what to build and what to buy.

Everything below concerns technical O&M.

We’re publishing the median because it gives a clearer picture of how our O&M operations perform in practice. It is based on recorded interventions between January 2025 and April 2026, with each work order’s operational data logged and timed. That gives us a measurable baseline we can track, break down and improve over time.

The real achievement for us as solar O&M providers isn’t just the number, but also the established measurement system that makes the the number possible and consistent over time.

Search for guidance on this decision and you will repeatedly find a threshold of around 1 MW, above which in-house O&M is said to become viable. It appears in vendor blogs, in aggregated summaries, and increasingly in AI-generated answers that draw from them.

It is wrong by a wide margin, and it’s worth understanding why, because the error reveals the flaw in the whole salary-versus-fee framing.

A 1 MW plant generates, very roughly, a few hundred hours of scheduled maintenance work per year. It does not generate enough work to occupy one technician full-time, let alone the two or three required to cover absence and simultaneous faults. It certainly does not justify a control room, an engineering function, a spares inventory, a CMMS, or a certified HSE management system.

What a 1 MW threshold actually describes is the point at which it becomes worth paying attention to O&M. That is a genuine threshold and a much lower one. The point at which it becomes economic to build the capability yourself is a different question entirely, and the arithmetic behind it looks nothing like a single number.

A practical field-staffing benchmark runs at roughly one field technician per 20 MW of installed capacity, with one O&M manager per 40–50 MW. Run those ratios against a small portfolio and the problem shows itself immediately.

At 20 MW, the ratio implies one technician. Not two, not a rota — one person. One person is not seven-day coverage; it’s a single point of failure who cannot take a holiday, get sick, or attend a training course without the plant going uncovered. The management ratio is worse: 20 MW implies roughly 0.4–0.5 of an O&M manager, and organisations don’t employ four-tenths of a person. You round up to one, and that one manager is now carrying a fixed cost that a genuinely large portfolio would spread across 40–50 MW instead of 20.

At 60 MW, the technician ratio gives three — enough bodies to start rotating, though still thin for illness, training and simultaneous faults across a full week. The management ratio gives 1.2–1.5, which in practice still rounds up to a single full-time O&M manager, now finally sized close to their intended workload.

Somewhere above that — the exact point depends on variables this article can’t responsibly put a number on, including control-room and monitoring software costs, cybersecurity requirements, and the spares and tooling capital discussed below — the fixed layer starts to look genuinely amortised rather than force-fit onto too few MW. Below it, in-house isn’t just more expensive, it’s structurally short of the redundancy a reliable operation requires, regardless of what the technician’s salary looks like on a spreadsheet.

And MW is only the crudest input. Geographic concentration, the minimum headcount an internal team needs before it can function at all, the fixed cost of monitoring and control-room infrastructure, and the capital tied up in spares all move the real threshold as much as capacity does — each is taken up in the sections that follow. The honest conclusion is that no single MW figure — not the 1 MW commonly cited online, and not a tidier-sounding replacement — survives contact with how differently those factors combine across real portfolios. What does hold up is the staffing math: below a certain concentration-adjusted size, the ratios simply don’t produce a viable team, and no amount of salary-line optimisation changes that.

Here is what has to exist for a utility-scale portfolio to be operated reliably. Whether you employ it or contract it, you are paying for all of it.

Continuous monitoring, alarm configuration, and the triage layer that separates a real fault from noise. A portfolio generates alerts constantly; the value is in knowing which ones matter and how quickly.

The capability to diagnose why something failed, not merely that it failed. This is chronically undervalued. Alectris made the point over a decade ago in its O&M contractor evaluation work: engineering gets discounted when a contractor’s role is limited to preventive maintenance and incident response, but as plants age, the ability to analyse deficiencies and propose technical corrections becomes the difference between maintaining performance and slowly losing it.

Technicians, with the qualifications to work on HV equipment, in sufficient number to cover the portfolio’s geography and to absorb simultaneous faults, absence and turnover.

The function that turns a detected fault into a technician standing in front of the right piece of equipment with the right part. More on this below, because it’s where most of the time actually goes.

Inventory, warehousing, and the working capital tied up in both. A critical inverter component that has to be ordered when it fails is a multi-week outage; the same component on a shelf 90 minutes away is a same-day fix.

Vehicles, thermographic cameras, IV curve tracers, calibrated instruments, PPE, and the calibration and replacement cycles for all of it.

The system that holds the asset register, the work orders, the maintenance history, and the evidence trail. This is not an administrative nicety. Warranty claims, insurance claims, lender reporting and due diligence at exit all depend on being able to produce a defensible record of what was done and when.

Initial qualification plus ongoing renewal, per technician, per equipment type.

Safety systems, method statements, permits to work, incident procedures, and the certification that evidences them. A single serious incident on a self-performed site is an owner liability in a way that it is not under a contracted arrangement.

The ability to surge when a storm takes out multiple sites at once, or when two inverters fail on the same morning in different provinces.

Somebody has to run all of the above. That person is usually not counted in the in-house calculation and is usually the most expensive line in it.

That is eleven functions. A salary-versus-fee comparison counts one of them.

There’s a specific reason dispatch deserves its own treatment: it’s the part of the chain owners most consistently underestimate, and the part where scale makes the most difference.

Alectris analysed 16 months of categorised work orders across its Italian portfolios, covering January 2025 to April 2026. The headline figure was a median intervention time of 137 minutes.

The more useful finding was what happened when that number was broken apart. A grid event, a string-level alert and a scheduled corrective task are detected, escalated and dispatched differently. Averaged together, they produce a figure that is accurate and operationally useless, because it tells you nothing about which part of the chain is slow.

Broken into stages, the chain runs roughly: detection → triage → work-order creation → assignment → travel → intervention.

Alectris measures dispatch delay at each of those stages individually rather than only reporting the blended total — the internal breakdown isn’t published here, but the method itself is the point worth taking from this. The moment a delay is attributed to a stage rather than buried in an average, it becomes something you can act on. A slow “detection” stage points at monitoring and alarm configuration. A slow “assignment” stage points at coordination and workflow tooling, not at the technician at all.

The reason this matters for the in-house decision is that each stage of that chain maps to a different item in the capability stack. Detection is a control room. Triage is engineering judgement. Work-order creation and assignment are a workflow system and a coordinator. Travel is fleet and geographic density. You cannot compress the chain by hiring a technician, because the technician only enters at the fifth stage.

An owner self-performing with a small team typically has stage five covered and stages one to four improvised. That’s where response times go.

Two properties of O&M cost explain most bad in-house decisions.

First, most of the capability is fixed cost. A control room costs roughly the same to run whether it watches 20 MW or 200 MW. So does a CMMS licence, an HSE management system, an engineering function, and the manager coordinating it. Fixed costs amortise, which is why cost per MW falls as portfolios grow. This isn’t a vendor talking point; it’s visible in the sector-wide data. IRENA’s Renewable Power Generation Costs in 2023 reports that falling O&M costs were one of the direct contributors to utility-scale solar’s global weighted-average LCOE dropping 12% year-on-year in 2023 alone — a decline IRENA attributes to a combination of falling installation costs, improved capacity factors, and reduced operating costs, not to any one of them in isolation.

Second, coverage is stepped, not linear. As the staffing ratios showed earlier, one technician isn’t coverage once holidays, sick days, training and simultaneous faults are accounted for — and two still can’t staff a seven-day rota with redundancy. The same applies to control-room operators: continuous monitoring isn’t something a person does, it’s something a rota does.

So the cost curve isn’t a smooth line you can slide along. It’s a staircase. You either fund the step that gives you genuine coverage, or you fund something cheaper that covers most of the time and fails precisely when it’s least convenient — during a storm, over a holiday, at 4pm on a Friday.

Half a capability does not cost half as much as a whole one. It costs slightly less and delivers considerably less.

A single field technician in Italy carries a gross salary in the region of €40,000/year. Layer in employer social contributions (roughly 30% on top), a vehicle, tooling and calibrated instruments, PPE, and initial and recurring HV/electrical safety certification, and the fully-loaded annual cost lands closer to €55,000–65,000 — 40–60% above the salary line most in-house comparisons start and stop at.

That gap is the whole argument in one number. Before a single hour of coverage-cliff redundancy is added — the second and third technician needed for genuine seven-day cover — the “cheap” option is already a third more expensive than it looked on the back of an envelope. (This range uses a standard Italian employer-contribution rate of roughly 30% plus typical vehicle, tooling and certification costs for the role — a market-level estimate rather than a disclosure of any individual employer’s payroll.)

The same fixed-versus-variable pattern applies to spare parts. A single-site owner has to hold enough inventory to cover their own worst case in isolation — the critical inverter component, the transformer part, the one component that turns a same-day fix into a multi-week outage if it isn’t on the shelf. A contractor running many sites pools that inventory across a warehouse network instead: Alectris operates two warehouses, one in the north of Italy and one in the south, supporting field teams across more than 130 sites nationwide, with stock positioned by where it’s actually needed rather than duplicated site by site. The capital tied up per MW under that model is lower than a single owner stocking alone would require, for the same reason an insurer needs less capital per policyholder than each policyholder would need self-insuring individually — the risk being covered is pooled, not the ownership.

For external reference points, published European benchmarks currently sit at roughly €7–13/kWp/year for utility-scale and €10–19 for commercial, with OPEX typically running 1.5–3% of initial CAPEX annually. Use these to sanity-check a quote, not to cost an in-house build. They price a service, not a capability.

Not raw MW — MW in one place. A large portfolio spread across five countries has poor in-house economics even at high capacity, because travel time and local compliance destroy technician utilisation. A comparable portfolio concentrated in two provinces has good economics, because a technician can service several sites in a day.

Utilities, industrial groups and large IPPs that already employ electrical technicians, already run an HSE system, and already hold the certifications are adding a capability at the margin rather than building one from zero. Their true incremental cost is genuinely much lower than a greenfield in-house build.

Every fault your own team diagnoses becomes institutional knowledge. Over a 25-year asset life, that compounds. It also matters at acquisition time: an owner who understands their fleet’s failure patterns negotiates better than one who has outsourced that understanding along with the work.

If operational data lives in a contractor’s system and the contract ends, you can lose the history. This is a real risk and it’s underdiscussed. It’s also solvable contractually, but only if you raise it before signing rather than after.

A contractor prices in profit. At sufficient scale, that margin is worth more than the efficiency you give up.

No scope negotiation, no change orders, no arguing about whether a task falls inside the contract. For owners running unusual equipment or frequent modifications, this is worth more than it looks on paper.

Concentration matters more than capacity. 50 MW in one region supports a very different model from 50 MW across four countries. Run the staffing ratios above against a scattered 50 MW and the technician-to-MW math holds on paper while the travel time between sites quietly erodes it in practice. Calculate MW per hour of drive time, not MW.

Asset age shifts the weight from field to engineering. New plants need routine maintenance and warranty administration. Ageing plants need diagnosis: degradation analysis, PID, connector failures, inverter obsolescence, and revamping decisions. That shift moves the balance toward whoever has genuine engineering depth, which is a harder capability to build in-house than a field team.

Equipment heterogeneity taxes in-house teams disproportionately. A portfolio acquired piecemeal carries several inverter platforms, several monitoring systems and several vintages of module. Alectris’s contractor evaluation work made this point about O&M providers and it applies identically to self-performing owners: relying on multiple disconnected monitoring systems makes consistent portfolio-level oversight very difficult, because the data sits in separate databases and cannot be aggregated meaningfully.

Growth trajectory matters as much as current size. Building an in-house capability for a portfolio you intend to double is a different decision from building one for a portfolio at its final size. Capability takes 18–24 months to mature.

Work through these in order. The answers give an indicated model rather than a definitive one, but they will surface the constraints that actually bind.

 

1. Map your capability requirement. Go through the eleven functions above and mark each: have it properly / have it partially / don’t have it. Be strict. “One person who checks the monitoring portal each morning” is not detection.
2. Cost the gaps honestly. For each gap, cost the whole step, not a fraction. Include management overhead. Include recruitment and the 6–12 months before a new hire is fully productive.
3. Calculate MW per drive-hour, not MW. Take your total capacity and divide by the number of distinct service regions. Use the largest single region as your real in-house case.
4. Establish your coverage requirement. What response time does your revenue model actually need? A plant on a merchant tariff during peak pricing periods has a different requirement from one on a stable feed-in tariff. Work backwards from lost revenue per hour of downtime.
5. Test the coverage staircase. For the coverage requirement in step 4, what is the minimum headcount that provides it with redundancy? Not the minimum that provides it on a good week. As a starting ratio: roughly one field technician per 20 MW and one O&M manager per 40–50 MW is a workable planning benchmark — then check whether your actual geography lets those ratios function as real coverage or only as a headcount that looks right on paper.
6. Compare like with like. Your fully-loaded in-house cost from steps 2 and 5, against a quoted fee for equivalent scope and equivalent SLA. If the quoted scope is narrower, adjust before comparing.
7. Decide the seam. If neither pure model fits, decide which functions you keep — and check that your split doesn’t separate detection from resolution.

Looking for O&M or EPC services for your solar assets in Italy/Greece? Book a meeting with our team today.