Networks

Why a monopoly is a market-design question

Every market on this site clears on top of a network somebody else owns, plans and rations. That network is a regulated monopoly: its revenue is set by a regulator rather than by competition, its costs are recovered from users who cannot shop elsewhere, and its capacity is allocated by rules rather than by price. None of that makes it less of a design problem. It makes it a design problem in which the usual corrective — if the price is wrong, someone enters — does not operate, so the rules have to be right the first time.

Three instruments do the work, and they are substitutes for one another far more than the institutional separation suggests. Connection terms decide who may use the network and on what conditions. Network tariffs decide who pays, how much, and whether the amount responds to anything the user can control. Flexibility procurement — the subject of its own page — buys behaviour change instead of building. A system that sets its connection terms and tariffs badly will find itself buying expensive flexibility to correct a problem its own rules created, and a system that builds copper without pricing its use will pay for capacity that sits idle for all but a few hours a year.

The constraint has moved

For two decades the binding constraint on European power markets was generation adequacy, and network capacity was assumed. That assumption has failed. Renewable projects totalling around 1,700 GW were sitting in connection queues across sixteen European countries in 2024 — several times the entire installed capacity of the largest member state — and in the Netherlands the regional operators' waiting lists ran to more than fourteen thousand requests for offtake alone, with tens of gigawatts more queued at transmission level. The projects exist, the capital exists, the policy support exists. What is missing is a place to plug in.

That changes what market design is for. A capacity mechanism cannot help a plant that cannot connect; a flexibility market cannot relieve a constraint the tariff is actively creating; a bidding-zone review cannot fix a queue. The network layer has become the binding one, which is why it deserves its own treatment rather than a paragraph inside somebody else's page.

First come, first served, and why it broke

Almost every European system began by allocating connection capacity in the order applications arrived. The rule is transparent, non-discriminatory and administratively simple, and it fails badly under scarcity for a reason that is entirely predictable: when capacity is free to request and cheap to hold, the queue fills with options rather than projects. A developer with a site and a hope occupies the same place as a developer with planning consent, finance and a turbine order, and because the first has no reason to withdraw, the second waits behind it. The queue stops being a schedule and becomes a portfolio of unexercised options on the grid.

The remedies now being implemented across Europe share a shape. Great Britain rebuilt the process wholesale: Ofgem approved the system operator's connections reform in April 2025, replacing the ordered queue with a gated model in which projects must demonstrate readiness and alignment with the national capacity plan before receiving a firm offer, with the reissued offers phased across 2025 to 2027. The Netherlands moved from pure chronology to a prioritisation framework, and its regulator has pushed operators towards flexible contracts, time-bound transport rights and locally negotiated capacity sharing. At Union level, the Grids Package presented in December 2025 pointed the same way: go beyond first come first served, apply first ready first served, publish transparent maturity criteria, set development milestones with penalties for missing them, and clean the queue regularly.

The design tension is not subtle. Chronology is fair and useless; readiness tests are useful and discretionary. Every maturity criterion is a judgement about what a serious project looks like, made by an operator with limited information and enormous consequence, and every one of them advantages incumbents and well-capitalised developers over newcomers. The honest position is that a readiness test is a rationing rule wearing engineering clothes, and it should be designed, monitored and appealed as such.

Firm, flexible, and what a connection actually sells

The deeper move is to stop selling one product. A firm connection is an option on every hour of the year: the network guarantees the user's full capacity whenever they want it, which requires the system to be built for a coincidence of peaks that almost never happens. A flexible or curtailable connection sells access subject to conditions — a cap during defined hours, an obligation to reduce when the constraint binds, a share of a jointly managed capacity — and in exchange the user connects years earlier.

Britain's connect-and-manage regime, Dutch time-bound and alternative transport rights, and the flexible connection agreements now standard for distribution-connected renewables in several countries are all variations on the same bargain: the connectee accepts risk that the network operator would otherwise have to build away. Whether that bargain is fair depends on things the contract usually leaves vague — how many hours of curtailment, under what notice, compensated or not, and for how long before reinforcement arrives. A flexible connection with an unbounded curtailment obligation and no sunset is not a product, it is a queue by another name.

Two different bills

Network charging is two questions that are frequently confused. The first is the connection charge: what a new user pays to be joined to the system. Under a shallow regime it pays only for the assets dedicated to it — the cable to the substation — and the network absorbs any reinforcement its arrival triggers. Under a deep regime it pays for the reinforcement too. Shallow charging is fast and encourages connection, and it socialises the cost of accommodating whoever turns up. Deep charging makes the newcomer face the cost it causes, and it can make the first project into a constrained area pay for infrastructure the next five will use for free — which is why most systems land somewhere in between, and why the boundary is fought over constantly.

The second is the use-of-system tariff: what everyone pays each year to keep the network standing. Its structure is where the real design content sits, because a network's costs are almost entirely fixed and sunk while its tariffs are usually levied per kilowatt-hour. That mismatch is the origin of most of the pathologies: a household that halves its consumption halves its contribution to a cost that did not change, and a heat pump or an electric vehicle pays for the network in proportion to energy rather than to the peak it adds.

Making the tariff say something

The reform direction across Europe is to move charges from energy towards capacity and towards time. Germany's §14a regime is the clearest worked example: since January 2024 network operators may reduce controllable devices — heat pumps, wallboxes, home batteries — to a floor of 4.2 kW when the local network is at risk, and in exchange the customer chooses among reduced network-charge modules, with transitional arrangements tightening from mid-2026 and an energy management system required where several devices sit behind one connection. The Netherlands has pushed time-bound transport tariffs for large users and contractual alternatives to firm capacity. France recovers a substantial share of renewable connection cost through regional schemes with a shared per-megawatt levy, so that the cost of building out a region is spread across the projects that use it rather than landing on whoever arrives first.

Each of these is an attempt to make the bill respond to the thing that actually drives network cost, which is coincident peak at a particular location. Each also runs into the same three objections, and a design is only credible once it has answered them: cost-reflectivity is regressive when the households least able to shift load pay the most; granularity is unusable if the customer cannot understand or automate the response; and any tariff sharp enough to change behaviour is sharp enough to be gamed by whoever can move fastest. The last of these is the one that connects back to the rest of this site — a locational tariff and a local flexibility market are two ways of buying the same behaviour, and a system that runs both without reconciling them will pay twice.

The regulated asset base and its bias

Network revenue in most of Europe is set by allowing the operator to recover its costs plus a regulated return on its asset base. That model built the twentieth-century grid and it contains a bias that matters enormously now: an operator earns a return on capital it deploys and no return on capital it avoids. Faced with a choice between reinforcing a substation and procuring flexibility to defer it, the regulated incentive points at the copper. Britain's total-expenditure approach — treating capital and operating spend in one pot so that the operator keeps a share of savings however they are made — is the standard answer, and it is only a partial one; the calibration of what the operator keeps decides how hard it actually looks for the cheaper option.

The asymmetry of regret

The harder problem is timing. Network assets take five to fifteen years to deliver and last forty; the demand they serve is now driven by electrification and industrial siting decisions nobody can forecast at that horizon. A regulator that approves investment only against demonstrated need will be systematically late, because need is demonstrated by a queue that has already formed. A regulator that approves anticipatory investment will occasionally pay for assets that turn out not to be required.

Those two errors are not symmetric, and pretending otherwise has been the central mistake of the last decade. A stranded network asset costs its capital, spread over a large customer base, and depreciates quietly. A missing network asset blocks gigawatts of investment that has already been financed, delays decarbonisation by the length of a planning cycle, and — as the queues show — is now doing exactly that at continental scale. The December 2025 Grids Package acknowledged as much, pressing regulators to permit anticipatory investment where it is justified. The design question that follows is who bears the risk when the anticipation is wrong, and the honest answer is that consumers do, which is precisely why the decision needs a transparent test rather than a case-by-case negotiation.

Three instruments, one constraint

A network operator facing a constraint has three ways to resolve it, and the interesting question is never which is best in the abstract. Build when the constraint is structural, growing and permanent: no tariff and no market will make a chronically undersized corridor adequate, and flexibility procured year after year against a permanent deficit is an expensive way of financing a wire. Price — through connection terms or a time-and-location-sensitive tariff — when the constraint is predictable and the response is something users can plan around: a tariff is cheap to run, needs no procurement machinery, and reaches every user including the small ones a market cannot economically enrol. Procure when the constraint is occasional, uncertain in timing, and the response has to be dispatchable on the day.

Read that way, the instruments sort by how predictable the need is and how much control the operator requires over the response — and the failure modes are equally systematic. Building against an uncertain need strands capital. Pricing against an unpredictable one produces a signal nobody can act on. Procuring against a permanent one buys the same relief forever and never fixes the cause. The recurring European error is the third: a flexibility market standing in for an investment decision that keeps being deferred, which is comfortable for everyone involved and solves nothing.

Where this page connects

The network layer is where several other arguments on this site terminate. The alternative to a local flexibility market, on the Flexibility page, is a connection agreement or a tariff — the two instruments described here. The capacity that a resource-adequacy assessment counts, on the Adequacy & Capacity page, is capacity that can reach load, which is a network question wearing a generation label. And the congestion management described on the Energy Markets page — bidding zones, capacity calculation, redispatch — is what happens when the network is already built and the market has to be run inside its limits. This page is about the decisions taken before that point.