Rimward · Fleet economics · September 2026

What should solar security scale with?

A portfolio can be modest in megawatts yet contain many operating units. But counting units is only the beginning: the physical estate and the work required to monitor it also matter. Three public sources show why a single €/MW figure is an incomplete specification.

2,808 RTE entries ≥ 1 MW
1,683 IGN footprint observations
CRE: 2021–2025 tender budgets
Separate cohorts · snapshot 30 September 2026

3.75×Register entries per MW · below 5 MW versus 5 MW and above

The smaller-capacity group has about 37 register entries per 100 MW, against 10 for the larger group.

That is a structure-of-the-register result, not a measured workload multiplier. An electrical entry is not necessarily a separately managed physical site.

It nevertheless changes the first budget question. If credentials, visits or escalation tests are managed separately at each operating site, a capacity-only view can conceal repeated work. Whether that happens in a particular portfolio must be checked against its inventory.

The many-small-installations problem

In this selected register cohort, entries from 1 to under 5 MW are 66% of the count but 34% of capacity. Their combined capacity is about 4,974 MW. A capacity-weighted portfolio report gives them much less prominence than an entry-weighted one.

Share of 2,808 RTE entries and 14.6 GW · common 0–100% scale

Share of entries
Share of capacity

1–<2 MW655 entries · 918 MW
Share of entries: 23.3% Share of capacity: 6.3%
2–<5 MW1,197 entries · 4,056 MW
Share of entries: 42.6% Share of capacity: 27.8%
5–<10 MW542 entries · 4,167 MW
Share of entries: 19.3% Share of capacity: 28.5%
10–<20 MW408 entries · 5,193 MW
Share of entries: 14.5% Share of capacity: 35.5%
20 MW and above6 entries · 279 MW
Share of entries: 0.2% Share of capacity: 1.9%

Lower bounds included, upper bounds excluded; values independently rounded. The ≥1 MW threshold and removal of aggregated entries define this selection. It is not the full French solar market or a validated count of operating sites.

The physical estate needs its own inventory

IGN changes the unit of observation: its aerial-image dataset maps the extent of panel groups. Among 1,683 objects photographed in 2021–2023, the median observed footprint is 3.78 hectares; the 90th percentile is 15.60 hectares. These are different observations from the RTE cohort, not matched plants.

Share of 1,683 mapped footprint objects · 2021–2023 imagery · not fence area

Under 1 ha19.2% 323 footprint objects
1–<5 ha41.7% 701 footprint objects
5–<10 ha19.6% 330 footprint objects
10–<25 ha16.5% 278 footprint objects
25 ha and above3.0% 51 footprint objects

Mapped panel extent is not a fence line, camera-coverage requirement or confirmed operating site. Ground, floating and agrivoltaic objects are included; roof and canopy installations are excluded by the product definition.

The useful connection is a question for the site assessment: what must actually be seen—gates, access routes, panel groups and equipment—and with which usable views? Hectares alone cannot answer it either.

Only 41 retained objects have declared peak capacity. We did not validate a plant-level RTE–IGN join, so this study does not invent a hectares-per-MW curve. The missing match is a reason to maintain a verified site model, not to treat the two datasets as interchangeable.

Recurring costs are under pressure

CRE compares candidate-declared operating budgets across different ground-mounted tender cohorts in 2021 and 2025; reported increases vary by cost category.

Reported change, % · declared budgets, not realised invoices

Total operating costs+25% Ground-mounted tender projects
Insurance+103% Reported component change
Electricity-selling fees+268% Reported component change
Land rent+56% Reported component change

Each percentage uses its own base; these are not cost shares. The report does not attribute insurance growth to theft or demonstrate security-related savings.

This is not a national security-price benchmark. It makes the procurement question more demanding: which recurring work is included, which charges vary with camera estate or operator attention, and when does an on-site callout become an additional cost?

Build a quote around the work, not one denominator

The three sources describe electrical scale, observed geometry and financial context. A comparable service specification needs the operating details that sit between them.

Which sites and views must be served separately?
Reconcile electrical entries with operating compounds. Record critical approaches, existing cameras, usable day/night views and unknown coverage—not only MW or hectares.
What recurs, and what is contingent?
Separate site onboarding and maintenance, staffed verification, optional remote-action capability and private callout terms. A common annual site charge F contributes F × verified site count ÷ total MW to the fixed-per-site cost component; it is not the whole bill.
What outcome and exposure are actually measured?
Ask for observation hours, review time, evidence sufficiency, attempted remote actions and subsequent attendance. Include failures and unknowns. Quiet alerts or a person leaving a view are not, by themselves, proof of theft prevention.

The budget should follow a verified service scope

These sources do not select a winning security technology. They show why generation capacity, physical footprint and recurring costs cannot substitute for one another.

For an asset manager, the next decision is concrete: obtain an inventory-backed scope and price the repeatable work separately from contingent response. Shared operating sites or pooled resources may reduce repeated work; complex views or frequent reviews may increase it. No national tariff, dispatch saving or insurance benefit follows from these public data alone.

Sources, cohorts and limits

RTE: electrical entries: Edition 31 July 2026, fetched 30 September. Solar entries with maxpuis ≥ 1,000 kW: 3,354 candidates, minus 546 names containing agrég. The 2,808 retained entries have nbinstallations = 1 and unique EIC identifiers. Capacity uses the reported maxpuis field, not assumed connection capacity. The entry-density ratio remains 3.75 after excluding 14 Bas-Rhin entries; this does not validate physical-site identity.
IGN: independent imagery observations: Version 1.0 package labelled 2021–2023, nominal delivery 1 November 2025, retrieved 30 September 2026. Of 1,762 layer records, 79 with earlier imagery years are excluded. Retain 1,683 distinct geometries and their surf_parc values in m², divided by 10,000. One identifier repeats across two distinct observations and is not used to count physical sites. The product groups nearby panel areas and includes installations in construction; operation status is not established here.
No cross-cohort normalization: The RTE selection and IGN imagery do not have a validated installation-level match. Only 41 retained IGN objects have declared capacity; that field is peak MWp, different from the RTE active-power concept. No area/MW, fence-length or camera-count estimate is published. Area percentiles describe observed objects, not uncertainty intervals or the complete 2026 fleet.
CRE: financial context: Report 2026-01, pages 16–17: candidate-declared budgets for retained PPE2 projects, with changing project cohorts. These are not realised incumbent costs or security invoices. The four reported changes are shown without multiplying them into the RTE cohort.