Use them where you make your own electrons.

There is one number that decides almost every solar business case we model, and it is not the price of a panel. It is the fraction of generation that is consumed where it is made.

A unit used on site avoids the full delivered price of electricity: wholesale, network charges, policy costs and supplier margin. For a business that is somewhere in the region of 25 to 40p, depending on tariff, site and half-hour. The same unit pushed out to the network earns a wholesale-linked export rate, often between 5 and 15p depending on the contract and the hour. Three or four to one, for the same photon, decided entirely by which side of the meter it lands on. (VAT sits on top of all of it, but a VAT-registered business recovers it, so it is not part of the saving.)

We have written the mechanics of the sell-side out in full before, in SEG, PPAs and the rise of symmetric pricing. The conclusion there was that export markets are getting cleverer, and that the interesting question has shifted from "what is my export rate?" to "what is my spread, and what can I do with it?" This piece is the other half of that answer. The best spread available to almost anyone in Britain is still the one between retail and wholesale, and you capture it by using the electricity rather than selling it.

The waste is now structural

The grid problem in Britain is no longer a shortage of generation. It is a shortage of wire in the right places at the right times. When the network cannot move what has been generated, the system operator pays the generator to stop, and the bill for that has run past a billion pounds a year, with NESO expecting it to stay heavy until the transmission build-out lands later this decade. That money buys nothing. It buys the absence of electricity.

At the site level the same problem wears a different hat. It shows up as an export cap on your connection offer, and it is why we argued in East-West arrays that on a constrained site the right question is revenue per hectare rather than yield per panel, and that a tall midday peak has become a liability rather than a prize. It shows up again in playing the DNO: you cannot choose your network operator, and you often cannot buy your way to a bigger connection inside a development timeline that anyone will fund.

So there are three ways to deal with a constrained connection. Shrink the array, which throws away the cheapest thing you own, the land you have already secured. Wait for reinforcement, which is somebody else's programme. Or put the load next to the generation.

Local first: the parish-scale private wire

Start with the boring version, because it is the version that pays. A farm, an estate or a council site with a decent array, a battery, and within a kilometre or two a set of buildings that already buy electricity at retail: a poultry unit, a cold store, a packhouse, a joinery, a leisure centre, a small industrial terrace, a village hall, an EV charging hub.

A private wire between generator and consumer, not using the public distribution network, sits under the class exemptions in the 2001 Order rather than under a supply licence. That is a real route, not a loophole, but it has edges: capacity limits, much tighter treatment where domestic customers are involved, and the fact that your offtaker will still want a grid backup supply behind the private wire. Within those edges the arithmetic is simple. The generator sells at a price above what export would have paid; the consumer buys at a price below what their supplier charged. Both sides are better off, and the difference they are splitting is the cost of moving electricity that never needed moving. The hard parts are wayleaves, metering, and writing a contract that survives a change of tenant. None of them are exotic.

This is the same logic that runs through the community bond and through local net zero: if the value is created in a place, there is no good reason for all of it to leave that place. It is also the reason we keep saying, in why 1% of the UK is all we need, that Britain has a coordination problem rather than a land problem. The generation and the demand are frequently within sight of each other and contractually strangers.

Then, and only then, the compute shed

Now the awkward part. Even with good local offtake, a well-sized rural array will spend hundreds of hours a year producing power that nobody nearby wants and the connection will not take. Summer middays, weekends, holidays. Those hours currently produce nothing at all.

Compute is the one significant load that can be moved to the electrons rather than the other way round. A containerised data hall arrives on a lorry and can be commissioned in months rather than years, and, crucially, it is interruptible: certain classes of work, training runs, batch rendering, scientific jobs, some proof-of-work, can be throttled to zero within seconds and picked up later. That is a very unusual property for a load, and it is exactly the property a constrained solar site needs. It is not free of paperwork, mind. Containers in a farmyard are still development, cooling fans are still a noise condition, and a rural site with no fibre is a compute site in name only.

The order matters, and we would put it strongly: compute is the buyer of last resort, not the anchor tenant. Serve the homes and businesses around you first, because their electricity is worth more, because it builds the local case for the array, and because a neighbour who buys your power does not object to your planning application. Then let the machines have what is left, at a price that beats zero and beats curtailment.

Done in that order, an emergent grid is not a slogan. It is a stack of contracts: retail-ish prices to local offtakers, an interruptible price to on-site compute, export as the residual, and a battery arbitraging between all three.

What it does to the investment case

For the investor, three things change at once, and they compound.

The third one is underrated. A merchant solar farm has a single revenue line exposed to a market that is deliberately trying to make midday power worthless. A microgrid with local offtake plus an interruptible load has a portfolio, and the value of a portfolio is that its worst hour is no longer zero. That is what a lender is actually pricing.

It also changes what you should build. If your surplus has a buyer on site, the case for packing more capacity onto the same acreage improves again, which loops straight back to the East-West argument, and to keeping the ground live underneath in the way we describe in vineyard agrivoltaics and the hop garden that became a power station. And if the site is ruined ground rather than farmland, a capped landfill on the edge of a town is very close to an ideal microgrid host: no agricultural argument, and demand within a couple of miles. The purest version of the same host, though, is an airport: one owner, kilometres of mown boundary that can never be built on, and a large continuous load already inside the fence.

The honest objections

A private wire is a civil engineering job and a legal one. Crossing a road, a railway or someone else's field turns a cheap cable into an expensive one, and a wayleave that takes a year. Route first, spreadsheet second.

Offtaker credit is real. You are replacing a licensed supplier with a local business, and if that business closes, your revenue closes with it. Two or three offtakers beats one, and the contract needs to say what happens when a building changes hands.

Compute is a business, not an appliance. Somebody has to buy the hardware, keep it busy, and accept that its value falls quickly. The version we would back is the one where the landowner hosts, sells power and takes rent, rather than the one where a farm becomes an accidental technology company.

And islanding, running when the grid is down, costs real money in protection and control. Most sites do not need it. Say so before someone designs it in.

The cheapest electron in Britain is the one that never enters the public network.

Our customers already know this, because it is the line in their model that made the project work. Everything above is the same idea taken one field further: build the generation, wire in the neighbours who will actually use it, and put something on site that is delighted to take whatever is left over rather than paying someone to switch it off.

If you own land with a constrained connection and businesses within a mile or two, that is the shape of the conversation. We will start with what the network looks like at your boundary, which you can check at a postcode before spending anything.

Prompted by Gridless's "Building the emergent grid", which makes the case in an African off-grid context. The British version is different in one important way: we have a network everywhere, and the constraint is its capacity rather than its absence.