For a decade, Scotland’s grid charging regime has been framed as an offshore wind story. That framing is now out of date. With the UK Government’s £750m Edinburgh supercomputer under construction at EPCC, the Lanarkshire AI Growth Zone designated in February, CoreWeave and DataVita’s £2.5bn Lanarkshire compute campus, and Lenovo’s new Edinburgh AI R&D hub all landing at once, Scotland is on the cusp of becoming one of the most concentrated locations for AI compute in Europe.
The single biggest risk to that outcome, and to the opportunity for further data centre growth, is not planning, not chip supply, not skills, and not even the Scottish Government’s second thoughts on environmental impact. It’s transmission charging.
Why compute is now a locational choice
Data centres — and hyperscale AI compute in particular — are electricity businesses first. A modern GPU-dense AI facility burns power at a rate that would have been unthinkable a decade ago. Ireland, until recently the default European destination for hyperscale build-out, has effectively paused new grid connections for data centres. Parts of west London are in the same position. That has pushed the compute industry to look for locations with three things at once: available power, renewable generation on the doorstep, and a stable price signal.
On the first two, Scotland wins outright. Scottish renewables output already exceeds Scottish demand for large parts of the year. The AI Growth Zone in North Lanarkshire is pairing DataVita’s 500 MW hyperscale build-out with 1 GW of private-wire renewables. The £750m EPCC facility is being built next to ARCHER2, powered from the same low-carbon grid mix.
But on the third, the price signal, Scotland has a real problem.
The TNUoS drag on Scottish compute
Transmission Network Use of System charges apply to generators, not to data centres directly. But they pass through the system in three ways that matter for anyone locating compute in Scotland.
First, they lift the cost of the underlying electricity. A generator in North Scotland pays around £26 per kilowatt of capacity a year, and one in East Scotland around £17. Along the English Channel and South Coast, generators receive a payment of around £4 per kilowatt for the same service. That differential adds up to £17 per MWh — and up to £27 per MWh once transmission losses are counted — to the cost of Scottish generation compared with a project in the south of England. That gap does not disappear when the power is sold to a Scottish data centre; it is embedded in the wholesale price signal Scottish generators are willing to bid.
Second, they distort the Contracts for Difference auctions that finance Scotland’s future clean-power supply. Offshore wind projects in Scotland cleared at £89.49 per MWh in the most recent Allocation Round 7. Every additional pound of transmission cost is a pound the developer has to recover in that strike price — meaning Scottish projects either bid higher and lose the auction, or win with a strike price that will not cover their costs and then cannot raise the debt to build. The National Energy System Operator’s own forecasts show TNUoS costs for parts of Scotland doubling by 2030. A 500 MW northern Scottish offshore wind farm’s annual TNUoS bill is on track to rise from roughly £13m today to £27m by the end of the decade.
Third, and most directly for compute: the UK Government’s own AI Growth Zone policy has priced in a £24/MWh electricity discount for a 500 MW data centre in Scotland, compared with £16/MWh in Cumbria and £14/MWh in the North-East. That figure is Whitehall’s own admission that Scottish electricity currently sits at a locational disadvantage large enough to require a corrective subsidy to attract hyperscale investment. A subsidy inside an AI Growth Zone is not a substitute for structural reform. It is a workaround — and it is one the Scottish Government seems to have been content to accept, rather than demand that the underlying charging regime be fixed.
What this means for the £750m supercomputer
The Edinburgh national supercomputer is a public research asset. Its viability is not determined by CfD arithmetic. But its operating economics — a facility that will run in the top 5 supercomputers in the world at launch, at around 50 times the compute of ARCHER2 — depend on Scottish electricity being competitively priced over a 15–20-year lifetime. EPCC has been clear that the £750m envelope covers the machine, the building, the staff, and the electricity to run it. Rising Scottish transmission-linked costs eat into every one of those line items.
The bigger question is what surrounds the national supercomputer. Edinburgh, alongside Bristol’s Isambard-AI and Cambridge’s Dawn, is one of three anchor nodes in the AI Research Resource that the UK Government has committed to expanding twentyfold by 2030. Around that public spine, the commercial cluster is now being built in Scotland — CoreWeave / DataVita’s £2.5bn compute campus in Lanarkshire, Lenovo’s Edinburgh AI hub, and the £15bn AI Pathfinder in North Ayrshire named in Scotland’s AI Strategy 2026-2031. That is roughly £20bn of announced tech infrastructure whose location decision was made on the assumption that Scottish power stays cheap and clean.
The reform window, and the opportunity for Scottish Ministers
Ofgem’s Transmission Network Use of System Review reached its final report in April, with the regulator’s preliminary view that a single GB cap and floor on the wider TNUoS £/kW charge is the most suitable intervention. That is the right direction. What matters for compute investors is a firm timetable.
Data centre location decisions are not five-year decisions; they are one-to-two-year decisions. The pipeline of hyperscale build-outs looking at Scotland now is being priced against an assumption about Scottish electricity costs in the late 2020s and early 2030s. If Ofgem’s cap-and-floor decision lands in 2029, it arrives after the current wave of AI Growth Zone investment has already committed capital, or gone somewhere else.
This is the point where Scottish Ministers could provide much needed input. TNUoS is a UK-wide regime, but the political standing to force a reform timetable on to Ofgem and Whitehall sits squarely with the devolved government whose growth sectors are most damaged by the status quo. There is no evidence they are using it. Scottish Government focus would seem to be on layering additional environmental and planning conditions onto the same offshore wind, oil and gas and grid infrastructure projects that a fixed TNUoS regime would make investable.
Consent bought the option. Reform is what exercises it.
For Scotland’s tech industry, TNUoS is no longer an energy-sector issue. It is now, alongside grid connection queues and planning acceleration, one of the three variables that determines whether the £750m supercomputer sits at the centre of a UK compute cluster or as a standalone research asset in a country that lost the surrounding industry to Cumbria and Teesside.
The UK Compute Roadmap sets out a national ambition; Scotland’s AI Strategy 2026-2031 sets out the Scottish delivery vehicle. Neither document can substitute for a decision on transmission charging that arrives in time to be priced into 2027 and 2028 investment decisions. Consent for offshore wind and designation of AI Growth Zones bought Scotland the option to host the UK’s compute economy. It did not exercise it.
The uncomfortable fact is that the growth industries Scotland claims to want — offshore wind at scale, hyperscale compute, energy-intensive manufacturing, and the U-turn on oil and gas — all depend on the same reform to a UK charging regime that works against realising these opportunities. Failing to reform TNUos is a key reason why the £1.7bn of Ocean Winds spend behind Caledonia, and the £20bn commercial compute pipeline sitting behind the £750m supercomputer, are still options and not commitments.
