My Iron Lung

Energy infrastructure, carbon capture, hydrogen, data centres, water.

Peak Cluster is four cement and lime producers across sites in Staffordshire and Derbyshire — Holcim, Tarmac, Buxton Lime, and Breedon — that together account for roughly 40% of UK cement and lime output. The plan is to capture around three million tonnes of CO₂ per year from these plants and transport it via a purpose-built underground pipeline to Spirit Energy’s Morecambe Net Zero facility, storing it in the depleted North and South Morecambe gas fields in the East Irish Sea.

In July 2026, the Chancellor announced £59.6 million in funding — £28.6 million from the National Wealth Fund, £31 million from a private consortium. The first CCS investment the National Wealth Fund has made. Around 1,500 new jobs. Green industry. Net zero. The bunting is up.

The cement case is genuine. CO₂ from cement production comes from calcination — the chemical process of heating limestone — not from the energy source. You cannot fix it with solar panels. If you are going to capture carbon anywhere, industrial process emissions in cement are the place where it makes the most defensible sense.

That defensibility is load-bearing. Everything that follows depends on it.

Sources: peakcluster.co.uk; gasworld; Carbon Herald; Yahoo News; Spirit Energy.

The Morecambe gas fields were discovered in 1974. They have been producing since 1985. Reservoir pressure in North Morecambe has dropped from 12.41 MPa to 0.82 MPa — a 93% collapse. Two of the field’s unmanned platforms have already been decommissioned, the world’s largest vessel brought in to remove the structures. The wells are plugged and abandoned. The gas ran out.

Nobody shut these fields down for the climate. The rock is empty. And the decommissioning bill for the Morecambe complex — seven platforms, including a three-platform manned central processing facility — would be enormous.

Spirit Energy’s own strategy language is instructive. Their stated priority is “meeting and de-risking our decommissioning obligations and pursuing strategic energy transition opportunities from our existing assets.” One sentence. Two purposes. When Centrica sold its interest in the Cygnus gas field, approximately half the transaction value — £99 million of £215 million — was specifically the transfer of decommissioning liabilities. Half the deal’s value was getting rid of the cleanup bill.

The UK’s total CCS commitment is £22 billion over 25 years — roughly £800 per household. Independent analysis by IEEFA puts the real cost at up to £408 billion by 2050. Approximately 75% of CCS subsidies will be paid by consumers through environmental levies on household energy bills. More than £50 billion has been earmarked for projects accounting for only 8% of the UK’s 2050 target.

The global track record: close to 90% of proposed CCS capacity in the power sector has failed or been suspended. No existing project has consistently captured more than 80% of carbon, despite industry claims of 95%. CCS sequestered roughly 0.1% of global energy-related CO₂ in 2022. As of September 2025, there are zero operational CCS projects in the United Kingdom.

The companies lobbying hardest for CCS are the companies whose business model it extends. Equinor, BP, and ExxonMobil attended 24 of 44 external ministerial CCS meetings in 2023. An MIT study found the US oil and gas industry spent nearly $1 billion since 2005 lobbying for CCS subsidies, with the primary motivation being not climate benefit but protecting fossil fuel projects.

The same week the CCS funding was announced, the incoming Prime Minister approved new drilling at the Jackdaw and Rosebank oil and gas fields. The developer, Adura, is a 50:50 joint venture between Equinor and Shell. At St Fergus in Scotland, Jackdaw’s gas comes onshore through the same terminal where the Acorn CCS cluster sends CO₂ back offshore. Same terminal. Same Shell. Gas in one direction. Carbon out the other.

The Peak Cluster pipeline runs from Derbyshire through Staffordshire, Cheshire, and out to the Wirral coast. It is classified as a Nationally Significant Infrastructure Project, with compulsory land acquisition powers. A permanent protected strip above which no tree planting, deep cultivation, drainage, or development will be permitted. Once the corridor exists, it can carry CO₂ today, hydrogen tomorrow, and whatever comes next. More than 70% of onshore pipelines can be reused for hydrogen transport.

None of this requires the CO₂ to stay underground. None of it is contingent on the climate outcome. The pipeline gets built. The subsidy flows. The decommissioning bill disappears. The infrastructure is permanent.

Sources: ResearchGate; Offshore Technology; Spirit Energy; Centrica; CMS; IEEFA; Byline Times; DeSmog; InfluenceMap; MIT; GB News; Rigzone; Inspenet; Farmers Weekly; IOGP Europe.

Engineering Greenness documented what data centres consume — four certificates, each converting a material question into a ledger entry. This section asks the question that piece did not: if the technology to build data centres without water already exists, why are these facilities being designed to use millions of litres a day?

In North Devon, a proposed data centre campus would cover 850 acres of productive farmland inside a UNESCO Biosphere Reserve. 1.5 gigawatts of compute. 1.8 gigawatts of battery storage. Eleven million gallons of water per day. The developer is Xlinks, whose £20 billion Morocco–UK subsea power cable was rejected by government in 2025. Same site. Same company. Same investors — TAQA, TotalEnergies, Octopus Energy. Energy companies, not technology companies. The use case changed. The site was always the asset.

At Didcot, RWE holds an abstraction right to draw up to 20 megalitres of water per day from the River Thames. The right was granted for Didcot A Power Station. The power station was demolished. The abstraction licence survived. RWE is building a data centre specifically to utilise the licence. The data centre is the vehicle for the water right, not the other way around.

Meanwhile, the technology to eliminate water use entirely already exists. It is not experimental. It is not awaiting validation. It is being deployed commercially, at scale, by the largest infrastructure operators on Earth.

There are three families of zero-water cooling, each commercially proven.

Bridge Data Centres and Chindata Group, working with Vertiv, have deployed X-Cooling — a waterless system using ambient air and advanced thermal controls — at hyperscale in China’s Hebei Province. The projected water saving is 1.2 million tons per year for every 100 megawatts of capacity.

The honest counterargument is energy. Evaporative cooling exploits a thermodynamic advantage: the phase change from liquid to vapour absorbs significant heat with relatively modest electricity. Air-cooled and closed-loop systems use more power — roughly 10 per cent more on average, rising to 25–35 per cent during peak summer conditions. That is a real tradeoff. Every operator listed above has concluded that the water saving justifies the energy cost. Several of the newer architectures — NVIDIA’s high-temperature inlet design, immersion cooling — are actually reducing total energy consumption, because running hotter eliminates the enormous overhead of chilling to traditional ambient temperatures. Cooling has historically consumed up to 40 per cent of a data centre’s total electricity.

The companies building the largest data centres on Earth have adopted zero-water cooling as their design standard. NVIDIA has made it the reference architecture for next-generation AI infrastructure. The technology is scalable, commercially deployed, and — as of 2026 — the direction the entire industry is moving.

The question is not whether you can build a data centre without water. You can. Microsoft does. Google does. Oracle does. NVIDIA’s reference design assumes you will. The question is why facilities proposed for the United Kingdom are still being designed to consume millions of litres a day.

The boring answer is cost. Evaporative cooling is cheaper to build. The developers chose the cheap option and externalised the water cost onto the community. That is probably partly true.

The second answer is that water abstraction licences are tradeable financial instruments. Secure the licence through planning. Retrofit to closed-loop cooling later. The licence remains as a permanent asset. The compute depreciates. The water right does not. Under the same tokenisation architecture documented in Everything Is Working Exactly As Designed, water abstraction rights sit on the same rails as biodiversity credits and carbon credits. The data centre does not need the water. The data centre needs the licence.

There is a third answer. It requires a different section.

The H100 Fife project is laying a hydrogen distribution network — new pipes running parallel to the existing gas mains — to supply around 300 homes with hydrogen produced by an electrolyser powered by offshore wind. It is described as a trial. The infrastructure being laid is a prototype for national rollout. You do not lay parallel networks for something likely to remain limited.

H100 Fife is in Fife. The proposed 600-megawatt data centre at Auchtertool — documented in Engineering Greenness — is in Fife. Three wind farms, a solar farm, and battery storage sit within two miles of the village. The EGL4 subsea cable sends electricity generated in Fife to England. The hydrogen heating trial is being built in the same county as the data centre that would consume half of Scotland’s household electricity.

Electrolysis — the production of hydrogen from water — requires water not as a coolant but as a feedstock. The water is consumed in the process. A facility with 1.5 gigawatts of compute, 1.8 gigawatts of battery storage, eleven million gallons of water per day, grid connections sized for industrial-scale energy conversion, and investors who are energy companies rather than technology companies is plausibly an energy production facility that includes compute, not a compute facility that happens to need energy.

A planning application for a “data centre” has an established pathway. An application for an industrial chemical production facility triggers different regulations, different environmental assessments, different Health & Safety Executive involvement, and different COMAH thresholds. Hydrogen production at scale is classified as a hazardous installation. A data centre is not. Same building. Different label. Different regulatory burden.

Water abstraction for cooling is a routine infrastructure request. Water abstraction as industrial feedstock for chemical production falls under a different regulatory category. In a water-stressed country, the framing of why you need the water determines whether you get it.

And the CCS infrastructure handles the CO₂ waste from hydrogen production. Without CCS, blue hydrogen is just another carbon-intensive process. With CCS, it is “low carbon.” The £22 billion CCS programme does not just capture cement emissions. It enables the hydrogen economy. The pipeline corridors are designed as multi-purpose. The terminal at Morecambe accepts delivery by pipeline, ship, and rail. The infrastructure does not care what it carries.

The planning system does not ask what happens inside the building. It asks about external impacts — traffic, noise, visual, drainage, energy demand. If those impacts are identical regardless of whether the facility produces compute or hydrogen or both, the planning system does not require you to explain the interior. You do not need to hide what you are doing. You just do not need to say.

Energy companies are building data centres. Data centre operators are becoming energy suppliers. The boundary between the two is dissolving in both directions, and the planning system has no category for what they are becoming.

Xlinks — backed by TAQA, TotalEnergies, and Octopus Energy — pivoted from a subsea power cable to a 1.5-gigawatt data campus. RWE is using a water abstraction licence from a demolished power station to build a data centre at Didcot. Centrica has purchased an 850-megawatt gas plant in South Wales in an area tipped to become a data centre hotspot. UK Power Networks — the regulated last-mile electricity distributor — is installing distributed compute nodes in people’s homes. Octopus Energy has committed £200 million to Deep Green, a company whose data centres heat swimming pools and leisure centres. In Finland, Microsoft’s data centre region near Helsinki will provide roughly 40 per cent of the district heating demand in Espoo, the country’s second-largest city. Google’s data centre in Hamina, Finland, is expected to supply 80 per cent of the town’s annual heating demand — free of charge.

These are not technology companies dabbling in energy. They are energy companies that have concluded that compute and heat are co-products of the same process.

The heat they produce is substantial. Cambridge University research, using two decades of satellite data across more than 8,400 data centres worldwide, found that data centres create a measurable heat island effect, raising nearby temperatures by an average of 2°C and as much as 9°C. The effect is detectable up to ten kilometres away. In Slough — Europe’s largest data centre hub, with 30 to 40 major facilities clustered near the town centre and an estimated energy demand of around one gigawatt — a weather station close to the data centre campus recorded 36.7°C during the June 2026 heatwave. Another station in the centre of town, further from the tech park, reached only 34.7°C. Two degrees. Same town. A few blocks apart. The planning system assesses noise, traffic, visual impact. It does not assess thermal output as an environmental impact.

At the domestic scale, the convergence is already operational. In an Essex garden shed, a retired RAF sergeant heats his two-bed bungalow using a HeatHub — a unit containing more than 500 immersion-cooled computers processing data for commercial customers. The waste heat transfers to his hot water system. His energy bill dropped from £375 a month to £40. The electricity powering the compute is paid for by the customer whose data is being processed. The homeowner pays £5.60 a month. UK Power Networks aims to deploy 100,000 of these systems annually by 2030.

The shed is a data centre. The data centre is a heating system. The homeowner is a node in a distributed computing network. The boundary between domestic space and industrial function has dissolved — not through coercion but through a genuinely good deal on heating. The system never scowls.

In January 2026, the UK government published its response to the consultation on heat network zoning — designated areas in English towns and cities where heat networks are expected to provide the lowest-cost low-carbon heating option. Using powers in the Energy Act 2023, the zoning framework includes a requirement to connect: certain types of buildings and low-carbon heat sources can be required to connect to a network within a prescribed timeframe. New builds. Non-domestic buildings above an energy threshold. Buildings already communally heated. Potential sources of waste heat — including data centres, categorised as “low temperature, near constant, recoverable heat sources” — may also be required to supply the network. Ofgem took over regulation of heat networks from 27 January 2026.

The UK currently supplies roughly 3 per cent of its heat through heat networks. The government target is 20 per cent by 2050 — an expansion projected to unlock more than £80 billion of investment. Veolia has a £1 billion pipeline of heat network projects across London, Bristol, Yorkshire, and Cambridgeshire. One industry estimate suggests that by 2035, UK data centres will generate enough waste heat to heat more than 6 million homes — if the infrastructure is built to capture and transport it.

The infrastructure does not yet exist. The UK has almost no district heating outside central London. The pipes have not been laid. The homes, in many cases, have not been built.

Which returns us to North Devon.

The Xlinks data campus is proposed for an 850-acre site between Great Torrington, Weare Gifford, and Huntshaw. The campus would occupy roughly a third of that land. The site was originally secured for the Morocco-UK Power Project — a 3,800-kilometre subsea HVDC cable that would have delivered 3.6 gigawatts of Moroccan solar and wind power to the UK grid via a landing point at Alverdiscott. The UK government withdrew support in June 2025. Xlinks withdrew its Development Consent Order and pivoted to data centres — repurposing both the site and the grid-connection planning work.

The site’s advantages are entirely infrastructure-driven. The existing Alverdiscott 400kV substation provides a grid connection point originally surveyed for 3.6 gigawatts — more than double the data campus requirement. A grid connection of this scale typically takes seven years to secure; the planning work is already done. The 1.8-gigawatt battery energy storage system, filed as a separate planning application, would sit behind the meter and provide grid-balancing revenue from National Grid’s ancillary services market. The Torridge catchment provides water access. The land is already under Xlinks’ control. The investors — TAQA, TotalEnergies, Octopus Energy — are energy companies, not technology companies. If the project proceeds as a Nationally Significant Infrastructure Project, planning consent is determined by the Secretary of State, not Torridge District Council. As of 30 June 2026, no planning application has been submitted.

Every locational advantage is an energy infrastructure decision. Not one is a decision about how to heat homes.

The population of the surrounding area is a few hundred people. There is no district heating infrastructure. There are no housing developments within pipe range of the waste heat. If this facility is built as currently described — without zero-water cooling, without waste heat capture — it will vent heat equivalent to hundreds of megawatts into the North Devon atmosphere and consume millions of litres of water per day for evaporative cooling. The Cambridge heat island research examined facilities one-sixth this size.

There is an obvious question. Why not present this as a single integrated development?

A 1.5-gigawatt energy-compute-hydrogen-housing-heat complex. Zero-water cooling producing waste heat at 45–60°C, piped through insulated underground heat highways to new-build housing on the adjacent 550 acres of unused land, zoned for mandatory connection under the Energy Act, regulated by Ofgem, with hydrogen electrolysis consuming the abstracted water and carbon capture handling the waste. One Environmental Impact Assessment. One public consultation. One planning decision. One document showing the full picture.

The answer is financial, not logistical.

Separated, each element accesses its own funding stream. The data centre gets NSIP fast-track planning. The battery storage gets grid-balancing revenue from National Grid ancillary services. The housing gets green belt release under local plan review, affordable housing grant, and Homes England subsidy. The heat network gets Green Heat Network Fund money. The hydrogen production gets hydrogen strategy funding. The water abstraction gets its own licence from the Environment Agency, assessed in isolation from the other water demands on the same site. The carbon capture links to the £22 billion CCS programme. The Climate Change Levy — an environmental tax on energy consumption, introduced to penalise heavy use and fund cleaner alternatives — is discounted by 92 per cent for data centres under a Climate Change Agreement, saving the largest facilities an estimated £80 to £100 million a year on a tax designed to discourage exactly what they do. The biodiversity net gain credits come from a separate offset market, purchased from land elsewhere. Each element has its own three-letter acronym, its own consultation, its own regulator, its own public benefit narrative.

Presented as one development, the total public subsidy would be visible in one column. Separated, it is distributed across half a dozen government departments, three regulators, and multiple planning consents. Nobody is required to produce the sum. Nobody does.

And each approval creates a fait accompli for the next. The grid connection already exists — that justifies the data centre. The data centre already exists — that justifies the housing, because workers need to live somewhere. The housing already exists — that justifies the heat network, because residents need heating. The heat network already exists — that justifies the mandatory connection, because Ofgem needs to regulate it. The waste heat already exists — that justifies the zoning, because why waste it.

Each step is reasonable. Each step is individually beneficial. Each step is individually consulted and individually approved. At no point does any single document show a private developer extracting public subsidy from six different government programmes to build an integrated energy-compute-housing complex that would never have received consent as a single application — because as a single application it would have required a single Environmental Impact Assessment, a single water abstraction total, a single energy demand figure, and a single public consultation in which the community could see the entire thing and respond to the entire thing at once.

This is not conspiracy. This is the ordinary operation of a planning system that evaluates applications in isolation because that is what it is designed to do. The developer who files one integrated proposal gets one EIA, one consultation, one chance for opposition to organise. The developer who files six separate applications gets six individual approvals and forces opposition to fight on six fronts simultaneously, each governed by different regulations, different timescales, and different decision-makers.

You do not need coordination to explain this. You need a spreadsheet and a planning system that cannot see wholes.

The word for it, in every sector the Green by Numbers series has documented, is the same word. The BNG metric measures habitat types, not animals — remove the ponies, the score stays the same. The EPC algorithm scores inputs, not outcomes — a cold expensive house outranks a warm cheap one. The RAM classifies by weight, not by health impact — glass scores worse than plastic found in brain tissue. The planning system evaluates applications, not sites. Each metric produces a number. Each number goes up. Each report says gain.

The Interoperable State documented what the digital identity infrastructure carries — credentials, licences, health data, biometric profiles. Its final section documents what it will carry next: programmable money. A Central Bank Digital Currency linked to the same wallet that holds your driving licence and your NHS number. Money with conditions. Expiry dates. Spending categories. Location restrictions. Rules you did not set and cannot change.

The Surveillance Architecture documented what watches. The Science of Surveillance documented what the published research enables. This piece documents what powers all of it and where it physically sits.

Programmable digital currency requires three things: a verified identity to attach it to, a surveillance infrastructure to monitor compliance with its conditions, and compute capacity to process the transactions. The identity layer is documented. The surveillance layer is documented. The compute capacity is being built in the facilities described in this piece and in Engineering Greenness — facilities that also produce the energy the system requires to run.

The FCA’s Digital Securities Sandbox already has sixteen firms testing tokenised issuance and settlement. The tokenisation rails that carry bonds and equities will carry carbon credits, biodiversity credits, nutrient credits, water abstraction rights, and — when the digital pound arrives — currency. Every one of these instruments runs on compute. Every unit of compute runs on energy. The energy runs through the pipeline corridors being built with compulsory acquisition powers, funded by consumer levies, on the depleted infrastructure of companies that emptied the gas fields and got paid to keep using them.

The drilling feeds the hydrogen. The hydrogen feeds the grid. The grid feeds the data centre. The data centre runs the AI. The AI processes the data. The data comes from the digital identity infrastructure that the Prime Minister cancelled on a Monday and the Permanent Secretary un-cancelled on a Tuesday. The smart meters on every house feed consumption data to the same system. The BNG registers, the Land Registry covenants, the environmental compliance architecture — all digital, all requiring compute, all running on energy, all converging in the same facilities.

This is not documented in any single planning application, any single regulatory submission, or any single government strategy document. Each component is approved separately. Each is defensible on its own terms. The convergence is visible only to whoever places the documents next to each other.

Nobody places the documents next to each other.

The planning authority assesses external impacts. The Environment Agency regulates water abstraction. The Health and Safety Executive oversees hazardous installations. Ofgem regulates energy. COMAH covers major accident hazards. Each sees the application that falls within its jurisdiction. None sees the facility.

The building that produces energy, runs AI, holds a water licence, generates tradeable credits, and will process programmable currency is approved as a data centre because the planning system classifies buildings by what they look like from the outside. The internal function is not a required field. Nobody is lying. The label is narrower than the function.

The question is not why the secrecy. The question is who asks. And the answer, documented across this series and its companion publications, is: nobody. Not because the question is suppressed. Because the system does not have a box for it.

The same mechanism operates at every scale. A Sudden Sense of Liberty? documented how the announce-celebrate-clarify sequence works because nobody reads across the headlines. This piece documents how the pipeline-water-planning sequence works because nobody reads across the regulators. Same principle. Different substrate. The confusion is temporary. The infrastructure is permanent.