7 Heating and cooling

Figure 7.1. A flock of new houses.

Figure 7.2. The water in a bath.

This chapter explores how much power we spend controlling the temperature of our surroundings – at home and at work – and on warming or cooling our food, drink, laundry, and dirty dishes.

Domestic water heating

The biggest use of hot water in a house might be baths, showers, dishwashing, or clothes-washing – it depends on your lifestyle. Let’s estimate first the energy used by taking a hot bath.

The volume of bath-water is 50 cm × 15 cm × 150 cm ≈ 110 litre. Say the temperature of the bath is 50 °C (120 °F) and the water coming into the house is at 10 °C. The heat capacity of water, which measures how much energy is required to heat it up, is 4200 J per litre per °C. So the energy required to heat up the water by 40 °C is

4200 J/litre/°C × 110 litre × 40 °C ≈ 18 MJ ≈ 5 kWh.

So taking a bath uses about 5 kWh. For comparison, taking a shower (30 litres) uses about 1.4 kWh.

Kettles and cookers

230 V × 13 A = 3000 W

Britain, being a civilized country, has a 230 volt domestic electricity supply. With this supply, we can use an electric kettle to boil several litres of water in a couple of minutes. Such kettles have a power of 3 kW. Why 3 kW? Because this is the biggest power that a 230 volt outlet can deliver without the current exceeding the maximum permitted, 13 amps. In countries where the voltage is 110 volts, it takes twice as long to make a pot of tea.

Figure 7.3. Power consumption by a heating and cooling device.

If a household has the kettle on for 20 minutes per day, that’s an average power consumption of 1 kWh per day. (I’ll work out the next few items “per household,” with 2 people per household.)

One small ring on an electric cooker has the same power as a toaster: 1 kW. The higher-power hot plates deliver 2.3 kW. If you use two rings of the cooker on full power for half an hour per day, that corresponds to 1.6 kWh per day.

A microwave oven usually has its cooking power marked on the front: mine says 900 W, but it actually consumes about 1.4 kW. If you use the microwave for 20 minutes per day, that’s 0.5 kWh per day.

A regular oven guzzles more: about 3 kW when on full. ¹ If you use the oven for one hour per day, and the oven’s on full power for half of that time, that’s 1.5 kWh per day.

Device	power	time per day	energy per day
Cooking
– kettle	3 kW	1⁄3 h	1 kWh/d
– microwave	1.4 kW	1⁄3 h	0.5 kWh/d
– electric cooker (rings)	3.3 kW	1⁄2 h	1.6 kWh/d
– electric oven	3 kW	1⁄2 h	1.5 kWh/d
Cleaning
– washing machine	2.5 kW		1 kWh/d
– tumble dryer	2.5 kW	0.8 h	2 kWh/d
– airing-cupboard drying			0.5 kWh/d
– washing-line drying			0 kWh/d
– dishwasher	2.5 kW		1.5 kWh/d
Cooling
– refrigerator	0.02 kW	24 h	0.5 kWh/d
– freezer	0.09 kW	24 h	2.3 kWh/d
– air-conditioning	0.6 kW	1 h	0.6 kWh/d

Table 7.4. Energy consumption figures for heating and cooling devices, per household.

Hot clothes and hot dishes

A clothes washer, dishwasher, and tumble dryer all use a power of about 2.5 kW when running.

Figure 7.5. The hot water total at both home and work – including bathing, showering, clothes washing, cookers, kettles, microwave oven, and dishwashing – is about 12 kWh per day per person. I’ve given this box a light colour to indicate that this power could be delivered by low-grade thermal energy.

Figure 7.6. A big electric heater: 2 kW.

A clothes washer uses about 80 litres of water per load, with an energy cost of about 1 kWh if the temperature is set to 40 °C. If we use an indoor airing-cupboard instead of a tumble dryer to dry clothes, heat is still required to evaporate the water – roughly 1.5 kWh to dry one load of clothes, instead of 3 kWh. ²

Totting up the estimates relating to hot water, I think it’s easy to use about 12 kWh per day per person.

Hot air – at home and at work

Now, does more power go into making hot water and hot food, or into making hot air via our buildings’ radiators?

One way to estimate the energy used per day for hot air is to imagine a building heated instead by electric fires, whose powers are more familiar to us. The power of a small electric bar fire or electric fan heater is 1 kW (24 kWh per day). In winter, you might need one of these per person to keep toasty. In summer, none. So we estimate that on average one modern person needs to use 12 kWh per day on hot air. But most people use more than they need, keeping several rooms warm simultaneously (kitchen, living room, corridor, and bathroom, say). So a plausible consumption figure for hot air is about double that: 24 kWh per day per person.

This chapter’s companion Chapter E contains a more detailed account of where the heat is going in a building; this model makes it possible to predict the heat savings from turning the thermostat down, double-glazing the windows, and so forth.

Figure 7.7. Hot air total – including domestic and workplace heating – about 24 kWh per day per person.

Warming the outdoors, and other luxuries

There’s a growing trend of warming the outdoors with patio heaters. Typical patio heaters have a power of 15 kW. So if you use one of these for a couple of hours every evening, you are using an extra 30 kWh per day.

A more modest luxury is an electric blanket. An electric blanket for a double bed uses 140 W; switching it on for one hour uses 0.14 kWh.

Cooling

Fridge and freezer

We control the temperatures not only of the hot water and hot air with which we surround ourselves, but also of the cold cupboards we squeeze into our hothouses. My fridge-freezer, pictured in figure 7.3, consumes 18 W on average – that’s roughly 0.5 kWh/d.

Air-conditioning

In countries where the temperature gets above 30 °C, air-conditioning is viewed as a necessity, and the energy cost of delivering that temperature control can be large. However, this part of the book is about British energy consumption, and Britain’s temperatures provide little need for air- conditioning (figure 7.8).

Cambridge daily temperature for 2006 and 2025, seven-day means with the daily range shaded, against the 15.5 degree heating threshold. The two years track closely; 2025 averages 11.6 against 11.1 in 2006.

Figure 7.8. Cambridge temperature: MacKay’s 2006 against 2025. Redrawn in the 2026 revision. Heating degree days fell from 1890 to 1766, about 7%, which is part of why British gas demand fell without anyone insulating anything — and a reminder that some of the apparent efficiency gain in the national statistics is weather. Summer still barely reaches the point where cooling would pay, so MacKay’s judgement that Britain has little need of air-conditioning survives.³

An economical way to get air-conditioning is an air-source heat pump. A window-mounted electric air-conditioning unit for a single room uses 0.6 kW of electricity and (by heat-exchanger) delivers 2.6 kW of cooling. To estimate how much energy someone might use in the UK, I assumed they might switch such an air-conditioning unit on for about 12 hours per day on 30 days of the year. On the days when it’s on, the air-conditioner uses 7.2 kWh. The average consumption over the whole year is 0.6 kWh/d.

Figure 7.9. Cooling total – including a refrigerator (fridge/freezer) and a little summer air-conditioning – 1 kWh/d.

This chapter’s estimate of the energy cost of cooling – 1 kWh/d per person – includes this air-conditioning and a domestic refrigerator.

Figure 7.10. My domestic cumulative gas consumption, in kWh, each year from 1993 to 2005. The number at the top of each year’s line is the average rate of energy consumption, in kWh per day. To find out what happened in 2007, keep reading!

Society also refrigerates food on its way from field to shopping basket. I’ll estimate the power cost of the food-chain later, in Chapter 15.

Figure 7.11. Heating and cooling – about 37 units per day per person. I’ve removed the shading from this box to indicate that it represents power that could be delivered by low-grade thermal energy.

Total heating and cooling

Our rough estimate of the total energy that one person might spend on heating and cooling, including home, workplace, and cooking, is 37 kWh/d per person (12 for hot water, 24 for hot air, and 1 for cooling).

Evidence that this estimate is in the right ballpark, or perhaps a little on the low side, comes from my own domestic gas consumption, which for 12 years averaged 40 kWh per day (figure 7.10). At the time I thought I was a fairly frugal user of heating, but I wasn’t being attentive to my actual power consumption. Chapter 21 will reveal how much power I saved once I started paying attention.

Since heating is a big item in our consumption stack, let’s check my estimates against some national statistics. Nationally, the average domestic consumption for space heating, water, and cooking in the year 2000 was 21 kWh per day per person, and consumption in the service sector for heating, cooling, catering, and hot water was 8.5 kWh/d/p. ⁴ For an estimate of workplace heating, let’s take the gas consumption of the University of Cambridge in 2006–7: 16 kWh/d per employee. ⁵

Totting up these three numbers, a second guess for the national spend on heating is 21 + 8.5 + 16 ≈ 45 kWh/d per person, if Cambridge University is a normal workplace. Good, that’s reassuringly close to our first guess of 37 kWh/d.

What happened to heating

A section added in the 2026 revision. This is the largest single item in the book’s demand stack, and it is the one that has moved most since 2008 — in two different ways. The heat a British house needs has fallen. And the energy required to deliver that heat has fallen much further, because the machine changed.

The demand fell

MacKay’s 37 kWh/d was written when Britain burned 3.52 EJ of gas a year. In 2025 it burned 2.20 EJ, a fall of 37%, and per head the fall is steeper still because the population grew: from about 43 to 24 kWh/d per person of gas for everything — power stations, industry and homes together.⁶ For homes specifically, the government’s own accounting puts gas consumption per household down 43% since 2008, with three-quarters of what remains going on space heating.⁷

Very little of that is virtue. Some is insulation and condensing boilers, which is the real efficiency MacKay’s Chapter E is about. Some is warmer winters. A great deal of the sharpest movement — a 17% drop in 2022 alone — is people turning the thermostat down because they could not afford the gas, which is a reduction in comfort rather than in waste, and belongs in a different column of the ledger.

The heat pump changes the arithmetic, not the heat

The important change is not the number of kilowatt-hours of heat. It is that heat and energy have come apart.

MacKay’s 24 kWh/d of hot air is a fact about the building — its walls, its windows, the temperature difference across them. A gas boiler at about 90% efficiency needs roughly 27 kWh/d of gas to supply it. A heat pump does not make heat; it moves it, and at a seasonal performance factor of 3 it needs about 8 kWh/d of electricity to deliver the same 24.

Apply that across the chapter. The 36 kWh/d of hot water and hot air, delivered by heat pump rather than combustion, becomes about 12 kWh/d of electricity; with the 1 kWh/d of cooling, the whole of this chapter falls from 37 kWh/d to roughly 13. At the 2.65 that British field trials actually measure rather than the 3.0 assumed here, it is nearer 15 — still a reduction by more than half.

That is a bigger change to the British balance sheet than anything on the production side of this book, and it comes from replacing a machine rather than from building anything new. MacKay saw it — Chapter 21 makes the case for heat pumps directly, and this edition’s Chapter M explains why the accounting works: a heat pump is charged only for the electricity it draws, while a boiler is charged for the whole of the fuel it burns.

The same machine, elsewhere in the house

The heat pump has quietly taken over an appliance this chapter also costs: the tumble dryer.

MacKay puts a tumble-dryer load at 3 kWh, and advises using an airing cupboard instead at about 1.5 kWh. A heat-pump dryer recirculates its air and recovers the latent heat rather than venting it, and uses roughly half the electricity per cycle — about 265 kWh a year for an 8 kg machine against 561 kWh for a condenser. So the machine now does what MacKay’s airing cupboard did, without the airing cupboard.

The regulation went further than the market. Since 1 July 2025 only heat-pump dryers may be placed on the EU market at all, condenser and vented models having been designed out by the ecodesign efficiency floor, and the label was rebased from the old A+++ scale to a plain A–G.⁸

It is worth noticing what that implies for the argument above. The same technology that Britain is not installing in its boiler cupboards has already been installed, by regulation and without controversy, in its laundry rooms — because there the competing appliance also ran on electricity, so no price ratio stood in the way. The obstacle to heat pumps was never the heat pump.

Europe has done it. Britain has not.

The gap between those two sentences is the most striking thing in the current data.

European heat pump sales rose 11% in 2025 to about 2.63 million residential units, bringing the installed stock to roughly 28 million. Germany passed a milestone: heat pumps took 50% of the space-heating market for the first time, on 299 000 units, with gas boilers down to 44%. France sold 528 000 and Italy 423 000. Measured against households, Norway, Finland and Sweden all exceed 30 sales per 1000 households.

The United Kingdom and Poland are below five per 1000.⁹ Britain recorded 51 886 certified retrofit installations in 2025, a 7% increase, and passed a cumulative 250 000. For a country with roughly 28 million homes, that is a stock replacement rate which would take several centuries.

The ratio tells the same story from the other end. Across Europe, fossil boilers still outsell heat pumps 2.1 to 1 — an improvement on 2.4 the year before. Worldwide it is 3.8 to 1, with just under 13 million boilers sold in 2025, 90% of them gas-fired.

So gas heating is in decline, and the decline is real, but it is a decline in volume per house rather than a replacement of the method. The boiler is still what gets fitted when the old one fails, in most of Europe and overwhelmingly in Britain.

Why Britain has not, in one chart

The machine is not the obstacle. A heat pump is cheaper to run than a gas boiler exactly when its seasonal performance factor beats the ratio of what a household pays for a kilowatt-hour of electricity to what it pays for a kilowatt-hour of gas. The European Commission puts modern heat pumps at three to five times the efficiency of a gas boiler, and notes that where electricity costs more than about three times gas, the boiler stays competitive.¹⁰

Household electricity price divided by household gas price, per kWh including taxes, for 31 European countries, against the ratio of 3 at which a typical heat pump breaks even. Sweden 1.3, Netherlands 1.5, France 1.8, Italy 2.0; Germany 3.2, United Kingdom 3.6, Belgium 3.9, Romania 5.1.

Figure 7.12. Where a heat pump is cheaper to run than a gas boiler. Household prices per kWh including all taxes and levies, Eurostat 2025 second half, with the United Kingdom from the Ofgem cap. Added in the 2026 revision.

The chart sorts the argument out. Sweden is at 1.3, the Netherlands 1.5, France 1.8, Italy 2.0 — in all of them a heat pump at a performance factor of 3 more than pays. Germany is at 3.2 and the United Kingdom at 3.6, both above the line, so in both a heat pump can be more efficient in physics and still cost more to run. Germany nonetheless reached 50% market share, which tells you what a subsidy can do; Britain, without one on that scale, sits below five installations per thousand households.

The Nordic countries that lead on heat pumps mostly do not appear here at all, because they have little or no household gas distribution to compare against. That is the same point from the other side: their heat pumps are not competing with gas, and never were.

This is chapter 28a’s argument arriving in the boiler cupboard. The obstacle is not the technology and not the physics. It is that a country’s tax and levy structure decides which kilowatt-hour is expensive, and Britain has chosen to load its levies onto the one that heat pumps run on.

How much worse, and for whom

A review for the UK Collaborative Centre for Housing Evidence gathered every available British field trial and demonstrator project and reached the same conclusion by a different route, using the same measure — it calls it the electricity-to-gas price ratio, EtGPR — and it is worth quoting because it is more pessimistic than anything above.¹¹

Its finding is that a heat pump saves money against an efficient gas boiler only if the price ratio is no more than 3:1 and the unit achieves an SPF of at least 3.0. Both conditions, not either. In Britain, where the ratio has historically sat above that, no field trial has yet demonstrated that heat pumps beat efficient gas boilers on running cost. In typical existing housing a heat pump is about 36% more expensive to run; even in the ideal case — a correctly sized unit reaching SPF 3.0 — it is about 9% more expensive.

Two details sharpen this. The first is that SPF 3.0 is not what British installations achieve. The Energy Saving Trust’s monitoring of 700 heat pumps found a mean of 2.65 for air-source and 2.81 for ground-source, and at the price ratio then prevailing the SPF needed merely to break even was 2.82. The second is that reaching 3.0 presupposes a house at EPC band C or better — cavity and loft insulation, double glazing. Homes at band D and below are unlikely to get near it, and that is most of the British stock. The fabric comes first; the machine cannot rescue a leaky house.

There are two ways out, and the report identifies both. One is the price ratio, which is policy. The other is that adding solar panels and a battery flips the arithmetic even at British prices, making the heat pump cheaper to run than the boiler. That is chapter 6’s roof and chapter 28a’s storage doing the same job from the household’s side: the surplus that cannibalises the market price is worth most to whoever can consume it where it is made.

The one genuinely encouraging trend in the report is the direction of travel. Over twenty-five years heat pumps have steadily become more competitive as the technology, the installation and the commissioning have improved — the air-source mean rose from 2.2 to 2.65 between the two Energy Saving Trust trials, and ground-source from 2.3–2.5 to 2.81. The machine is getting better at roughly the rate the argument needs. The price ratio is the part that is not moving.

What the machine costs to put in

Everything above is about running cost. There is a second number, and in the country where heat pumps first outsold gas boilers — taking 50% of space-heating sales in 2025 — it has gone badly wrong. Every figure below is installed cost before any grant.

A specification-matched comparison of the same 8 kW air-to-water system in a comparable existing house gives €29 719 net in Germany against €12 095 in the United Kingdom — a gap of 146%. Add tax and it widens, because Germany charges 19% VAT on a heat pump installation and Britain charges zero: €35 366 against €12 095, or about £30 500 against £10 400, a gap of 192%. Across Europe the German range of €23 000 to €40 000 stands against €9000 to €20 000 in comparable countries, with Poland at the bottom near €8000.¹²

Between a third and a half of the gross gap is ordinary cost structure — the component ranges are wide, and the midpoint is nearer a third. German equipment is dearer, at €9000–18 000 a unit against €3000–8500 for the Asian manufacturers that hold larger shares in Britain, France and the Netherlands — though for the matched system the equipment difference is €2000–4000, not the span of those ranges. German practice requires a concrete foundation where most of Europe wall-mounts, at €976–2700 against under €500, and electrical connection work runs €1800–4000 where elsewhere it is minimal. VAT accounts for a further quarter of the gap on its own — €5647 of it.

But the interesting part is what the remaining two fifths is, and when it appeared. Before 2020 Germany was only modestly dearer than its neighbours: about €17 000 for systems up to 15 kW in 2017–19, against roughly €12 500 in Britain and €11 000 in Sweden. Then German costs approximately doubled, while British costs rose 18% and Swedish 14% over the same years and through the same energy-price shock. Material costs rose about 20%. Something else moved.

What moved was the subsidy. In 2020 German support switched from a grant based on the capacity installed to one based on a percentage of the cost, and mean customer costs jumped from around €17 000 to around €40 000 at precisely that point. The study’s authors conclude the change in subsidy design was the single largest driver of the increase.¹³

The mechanism is arithmetic anyone can check. Under the current scheme a household receives 30 to 70% of eligible costs, capped at €30 000 of eligible cost. At a 50% rate, an installer who raises the price by €5000 costs the household €2500 and gains €5000. The buyer’s resistance is halved exactly where the seller’s incentive is doubled. A subsidy proportional to price is, in part, a subsidy to price. Two qualifications belong with that. The 30–70% rates and the €30 000 ceiling are the current scheme’s, while the doubling being explained happened from 2020 under earlier parameters; and above the ceiling the household bears the whole of any further rise, so the mechanism works below the cap and cannot by itself explain a mean that now sits above it.

German heat pump sales then fell 46% in 2024 — though that fall is usually attributed to demand pulled forward into 2023 and to uncertainty over the heating law, rather than to the cost premium alone. Prices have since come off about €4000, to near €36 000, as installer availability improved and cheaper equipment entered the market.

Set that beside figure 7.12 and the two countries fail in opposite ways. Britain installs cheaply — €12 095, about £10 400 before the £7500 Boiler Upgrade Scheme grant — and then charges a household 3.6 times as much for the electricity as for the gas, so the machine is affordable to buy and expensive to run. That British grant is a fixed sum rather than a percentage, which is exactly the design Germany moved away from in 2020. Germany has the better running-cost ratio of the two at 3.2, and has made the machine cost three times as much to buy. Neither obstacle is in the physics, and neither is in the heat pump. One is a levy structure and the other is the design of a grant — which is the shape this edition keeps finding, arriving this time in two different boiler cupboards.

The simpler machine, which this chapter already describes

There is a second kind of heat pump, it is the one Europe actually buys, and MacKay has already put it in this chapter — as an air conditioner.

His cooling section describes “a window-mounted electric air-conditioning unit for a single room” that “uses 0.6 kW of electricity and delivers 2.6 kW of cooling”. That is a coefficient of performance of 4.3, and it is the same box: run the cycle the other way and it heats. An air-to-air heat pump has no water circuit, no cylinder and no radiators. It is a unit on an outside wall, a unit on an inside wall, and a hole between them.

That difference dominates the cost. In Britain a single-room air-to-air system runs about £1800 to £2600 installed, against £8000 to £15 500 for the air-to-water system that replaces a boiler.¹⁴ The expensive part of an air-to-water retrofit is rarely the heat pump; it is the wet system around it — larger radiators or underfloor pipe to carry heat at a flow temperature the pump can reach efficiently, and a cylinder to put somewhere.

This is what the Nordic figures are mostly made of. Air-to-air units are the most common type sold across European markets, ahead of air-to-water, and Norway — at 48 sales per thousand households, the highest in Europe — is overwhelmingly an air-to-air market, with about 95% of its units used for heating rather than cooling. Finland is at 33.

And it explains the shape of figure 7.12, including the countries missing from it. The Nordic countries adopted heat pumps fastest not because they were greener but because they were replacing direct electric heating, where the comparison is a resistance element at a coefficient of 1. Against that, any heat pump wins at any electricity price, and no electricity-to-gas ratio stands in the way. Britain’s problem is not that heat pumps are hard; it is that Britain is replacing gas.

The limits are real and worth stating. An air-to-air unit makes no hot water, so something else must, and this chapter’s 12 kWh/d of hot water stays where it is. It heats the space it is in, so a house of many small rooms needs several units or accepts an uneven house. And it is invisible to policy: the £7500 Boiler Upgrade Scheme grant does not cover air-to-air at all, so Britain subsidises the expensive option by thousands of pounds and the cheap one not at all.

There is a by-product that MacKay could reasonably dismiss and his successors cannot. The same unit cools in summer. This chapter concludes that Britain has “little need for air-conditioning”, and figure 7.8 says that is still nearly true — but an air-to-air installation delivers cooling whether or not it is wanted, at no extra capital cost, which quietly changes the calculation for the warm weeks the chapter treats as negligible.

Ground, water, and the season

The chapter so far has dealt with heat pumps that draw from the air. Two other sources matter, and one of them answers a problem this book raises elsewhere.

Ground source. A borehole gives the pump a source at a steady 6 to 8°C rather than an outdoor air temperature that collapses exactly when heat is most wanted. In Sweden, where bergvärme is ordinary domestic equipment, rated SCOP runs between about 4.0 and 5.5 — four to five and a half units of heat per unit of electricity. British field trials measured 2.81.

That gap is worth explaining rather than splitting, because it is partly real and partly definitional. SCOP is a rated figure for space heating alone, measured on a standard test; SPF is measured in service and includes domestic hot water. Hot water has to be delivered at a much higher temperature than a radiator circuit, so including it drags the whole-year average down — which is why SPF is the more honest number for what a household actually gets, and why it is the one this chapter uses. The rest of the gap is genuine: Swedish installations are designed around low flow temperatures by decades of habit, and the British trial measured an immature market.

The price of ground source is the drilling, which is most of the capital cost and the reason it remains a minority. Chapter 16 explains why the common name for these, “geothermal heat pumps”, is a misnomer: the heat is stored sunshine, not the Earth’s, by a factor of about 150.

Water source. Where a lake, river or sea is available it is better still, because water carries heat laterally to the collector instead of waiting to be conducted through soil.

What it replaces decides the economics, not what it draws from. A Swedish study of the Saltsjöbaden Grand Hotel, which heats with two lake-source heat pumps and two oil boilers, costed replacing one of the oil boilers with a third heat pump: a saving of about 200 MWh a year for an investment near 500 000 kronor, paying back in about two and a half years. It was the largest single measure in the study, ahead of relighting, ventilation rebalancing and window replacement, and the whole package came to roughly 500 MWh a year, a 30% cut.¹⁵

Swedish figures put the same comparison more generally: bergvärme pays back in about four to five years against oil or direct electric heating, and eight to ten against district heating — and those were the numbers before the 2022 gas shock.¹⁶

Two and a half years against a British heat pump that costs 36% more to run than the gas boiler it replaces. The machine is the same. The difference is entirely in what sits on the other side of the comparison — oil at Saltsjöbaden, direct electric resistance in the Nordic air-to-air market, and cheap gas in Britain. A heat pump is not economic or uneconomic in itself; it is economic against some fuels and not others.

Arlanda, where the season itself is stored

The deeper problem with heating is not efficiency but timing: the demand arrives in January and the surplus in July, and chapter 26 is about the difficulty of moving energy between the two.

Stockholm Arlanda Airport does exactly that, using an aquifer. Since the summer of 2009 it has run what it describes as the world’s largest aquifer thermal store, in the Brunkebergsåsen ridge, divided into warm and cold sections. In summer, cold groundwater is drawn from the cold store to supply the airport’s cooling; the water comes back at around 20°C and is returned to the warm store. In winter that stored warmth is used for melting snow on the aprons and preheating ventilation air, and the cooled water goes back to the cold side, ready for the following summer. No groundwater is consumed — the same volume is returned.

The system saves about 19 GWh a year, which Swedavia puts at the consumption of 2000 single-family homes, at an airport whose total energy use is comparable to a town of 25 000 people.¹⁷

What makes it worth a section in this chapter is not the size. It is that Arlanda is storing the seasons themselves, at a scale and cost that work, in ordinary ground. Chapter 26 treats seasonal storage as the hardest unsolved problem in a renewable system, and chapter 28a shows that the value of any storage lies in the spread it can arbitrage. An aquifer store arbitrages the largest spread available anywhere in northern Europe: summer against winter. It cannot be sited everywhere, since it needs the right geology and a matched demand for both heat and cooling. But where those coincide, the answer to “what do you do with July’s surplus heat in January” turns out to be: leave it in the ground and come back for it.

Gas heat pumps, and the case for a hybrid

Two different machines get called a gas heat pump, and only one of them is interesting now.

A gas absorption heat pump burns gas to drive an absorption cycle instead of a compressor. It is a real product, sold commercially, with a coefficient of performance around 1.3 to 1.5 — so between about 45% and 65% more heat out of a unit of gas than a condensing boiler at 0.9 manages. And when this book was written, it was the better machine. Chapter 20 costs British electricity at 500 grams of CO₂ per kilowatt-hour, which was right for 2008. Work it through:

Grams of CO₂ per kWh of heat delivered	On MacKay’s 2008 grid	On Britain’s 2025 grid
Condensing gas boiler, 90%	203	203
Gas absorption heat pump, COP 1.4	131	131
Electric heat pump, SPF 2.8	179	43

In 2008 the gas heat pump beat the electric one. In 2025 it loses by a factor of three. Nothing about either machine changed. The grid did — British electricity fell from about 500 grams a kilowatt-hour to about 120 — and a technology that made perfect sense became pointless without ever failing at anything.¹⁸

That is the clearest instance in this book of a conclusion expiring rather than being refuted, and it is a warning about every other comparison here that has a carbon intensity of electricity buried in it. Chapter 20’s own 500-gram figure is used in several places and each of them needs the same correction.

The hybrid is a completely different argument, and it survives.

A hybrid is an ordinary electric heat pump working alongside the gas boiler that is already there, with a controller choosing between them hour by hour. The heat pump runs in mild weather, which is where most of the year’s heat goes; the boiler covers the coldest days and the hot-water peak.

The case for it is not carbon per kilowatt-hour, where it is plainly worse than a full heat pump. It is peak. Electrifying British heat completely takes peak electricity demand from about 50 GW to somewhere near 115 to 120 GW, with electric cars adding more on top. That is less a generation problem than a wires problem: every substation and every street cable in the country.

The Freedom Project put hybrids into 75 homes in Bridgend and measured them. The heat pump supplied 40 to 58% of annual heat, cutting carbon by 30 to 48% — over a tonne per home a year — while the coldest hours stayed on gas. The project’s estimate of avoided distribution network reinforcement, for South Wales alone, was £1.3 billion by 2050.¹⁹

And a hybrid works in the houses that full electrification cannot reach. It needs no radiator replacement, no electrical supply upgrade and no particular fabric standard, because the boiler covers whatever the heat pump cannot. That is precisely the problem chapter 21 quantifies: 52% of English homes reach EPC band C and 56% reach A to C, so about two in five sit below the level this chapter’s arithmetic wants.

The objection is real, and it is about permanence. A hybrid keeps the gas connection, which keeps the gas network, which keeps the network’s fixed costs spread across a shrinking number of customers — a problem that gets worse as it succeeds. And a hybrid house is not a decarbonised house; it is a half-decarbonised house, potentially for ever, because no moment ever arrives at which somebody rips out a working boiler.

British policy has taken a side. Hybrids are excluded from the Boiler Upgrade Scheme, which pays £7,500 towards a heat pump only if the fossil boiler goes entirely. Whether that is right depends on what one believes the binding constraint to be. If it is carbon per house, hybrids are a trap. If it is the number of houses converted before 2050, and the cost of the wires to do it, they may be the only route to the two homes in five that chapter 21’s figures place below the band a heat pump alone wants.

The standards moved faster than the stock

Two changes in the rules are worth recording, and one caveat matters more than either.

The recast Energy Performance of Buildings Directive, in force since 28 May 2024, requires all new buildings in the EU to be zero-emission from 2030 and new public buildings from 2028, with no on-site fossil fuel use. Subsidies for fossil-fuel boilers were to end in January 2025 and the boilers themselves are earmarked for phase-out by 2040; member states have until 29 May 2026 to transpose it.²⁰

LEED v5, opened for registration in April 2025, restructured the best-known voluntary standard around carbon: half of all available points now relate to decarbonisation, and every project must complete a carbon assessment, a climate-resilience assessment and a human-impact assessment.²¹

The caveat is the one MacKay would have insisted on. Both of these govern new buildings, and this chapter is about the stock. Britain builds roughly 200 000 homes a year against a stock of 28 million, so a standard applying only to new construction touches under 1% of the problem annually. The 37 kWh/d in this chapter is being spent, overwhelmingly, in buildings that already exist and that no building standard will ever reach. What reaches them is insulation and a different machine in the cupboard.

Notes and further reading

---

1. An oven uses 3 kW. Obviously there’s a range of powers. Many ovens have a maximum power of 1.8 kW or 2.2 kW. Top-of-the-line ovens use as much as 6 kW. For example, the Whirlpool AGB 487/WP 4 Hotplate Electric Oven Range has a 5.9 kW oven, and four 2.3 kW hotplates. www.kcmltd.com/electric oven ranges.shtml www.1stforkitchens.co.uk/kitchenovens.html

2. An airing cupboard requires roughly 1.5 kWh to dry one load of clothes. I worked this out by weighing my laundry: a load of clothes, 4 kg when dry, emerged from my Bosch washing machine weighing 2.2 kg more (even after a good German spinning). The latent heat of vaporization of water at 15 °C is roughly 2500 kJ/kg. To obtain the daily figure in table 7.4 I assumed that one person has a load of laundry every three days, and that this sucks valuable heat from the house during the cold half of the year. (In summer, using the airing cupboard delivers a little bit of air-conditioning, since the evaporating water cools the air in the house.)

3. Daily temperatures for 52.205 N, 0.119 E from the Open-Meteo archive, which serves the ERA5 reanalysis. Reanalysis is not a station record: it is a model reconstruction on a grid, so individual days will differ from what a Cambridge thermometer read. It is used here because it gives the same treatment to both years, which is what a comparison needs, where the Cambridge NIAB station record is monthly. Heating degree days are computed against a 15.5°C base, the convention in UK energy statistics, as the sum over the year of (15.5 − daily mean) where positive. Two years are two years and prove nothing about climate; the point of the figure is the shape of the demand, not a trend.

4. Nationally, the average domestic consumption was 21 kWh/d/p; consumption in the service sector was 8.5 kWh/d/p. Source: Dept. of Trade and Industry (2002a).

5. In 2006–7, Cambridge University’s gas consumption was 16 kWh/d per employee. The gas and oil consumption of the University of Cambridge (not including the Colleges) was 76 GWh in 2006–7. I declared the University to be the place of work of 13 300 people (8602 staff and 4667 postgraduate researchers). Its electricity consumption, incidentally, was 99.5 GWh. Source: University utilities report.

6. Energy Institute, Statistical Review of World Energy 2026: UK gas consumption 3.52 EJ in 2008 and 2.20 EJ in 2025; Europe 22.55 to 17.36 EJ; Germany 3.22 to 2.83 EJ. Per-person figures use populations of about 62 million in 2008 and 69.3 million in 2025, and cover all gas use, not only domestic heating.

7. Department for Energy Security and Net Zero, Energy Consumption in the UK and the subnational gas consumption statistics: domestic gas consumption per household down about 43% since 2008, down 17% between 2021 and 2022 and a further 6.4% between 2022 and 2023, with 75% of domestic gas going to space heating in 2024. The 2021–22 fall reflects both a warm year and the price shock, and the statistics do not separate the two.

8. European Commission ecodesign and energy-labelling measures for household tumble dryers, applying from 1 July 2025: the efficiency floor admits only heat-pump machines to the EU market, and the label reverts to an A–G scale with consumption expressed per 100 drying cycles. Consumption figures — about 265 kWh a year for an 8 kg heat-pump dryer against 561 kWh for a comparable condenser — are manufacturer and label data on the standard test programme, which uses a cotton load at a defined moisture content; real households load, dry and sort differently, so the ratio travels better than the absolute numbers. See https://energy.ec.europa.eu/news/new-measures-more-energy-efficient-household-tumble-dryers-1-july-2025-07-01_en.

9. European Heat Pump Association market data for 2025: sales across 16 European countries up 11% to about 2.63 million residential units, installed stock about 28 million; France 528 000, Italy 423 000, Germany 299 000 with 50% year-on-year growth and a 50% share of the space-heating market against 44% for gas boilers; Norway, Finland and Sweden above 30 sales per 1000 households, the United Kingdom and Poland below five. The European boiler-to-heat-pump sales ratio was 2.1:1 in 2025 against 2.4:1 in 2024. UK figures are MCS-certified retrofit installations, 51 886 in 2025 against 48 677 in 2024, passing 250 000 cumulative. Counts differ between sources depending on whether new-build and non-certified installations are included, so the UK total is sometimes quoted higher.

10. European Commission, “5 things you should know about heat pumps”, 5 December 2025: https://energy.ec.europa.eu/news/5-things-you-should-know-about-heat-pumps-2025-12-05_en. Modern heat pumps three to five times more efficient than gas boilers; bill savings from 20% to over 60% depending on local prices, and up to 80% in the Netherlands replacing an old boiler; gas boilers remain competitive where electricity costs more than about three times gas; the electricity-to-gas price gap has narrowed across most EU countries since 2019; Finland has 524 heat pumps installed per 1000 households, which is a stock figure and not comparable with the sales-per-1000 figures above. On the grid, the Commission reckons 52 million heat pumps in Europe by 2030 would be about 5% of annual electricity demand and 9% at peak — the clearest available answer to the objection that the wires cannot take it. Price ratios in figure 7.12 are computed from Eurostat household electricity (band 2500–4999 kWh/yr) and gas (band 20–199 GJ/yr) prices for the second half of 2025, both including all taxes and levies; the UK is the Ofgem default-tariff cap for July–September 2026 at 26.11p and 7.33p per kWh, a ratio of 3.56, which was 4.30 on the April–June cap. A seasonal performance factor of 3 is used as the break-even; real installations vary with the building, the emitters and the installer, and a poorly commissioned system can fall well below it.

11. Nicholas Harrington, The running cost of domestic heat pumps in the UK: a comprehensive evidence review and evaluation, UK Collaborative Centre for Housing Evidence, University of Glasgow, March 2024: https://housingevidence.ac.uk/wp-content/uploads/2024/03/The-running-cost-of-heat-pumps-v7.pdf. Funded under the EPSRC FASHION project (EP/V042033/1) and peer reviewed. The 36% and 9% figures are running-cost comparisons against an efficient gas boiler, all else equal, and are sensitive to the price ratio assumed; they are not carbon comparisons, on which heat pumps win regardless. The Energy Saving Trust field-trial means quoted are from Lowe et al. (2017) analysing 700 heat pumps monitored between October 2013 and March 2015 under the Renewable Heat Premium Payment scheme, against the Trust’s own earlier 2010 trial. Note that a running-cost disadvantage is not an argument against heat pumps on any other ground, and that the report’s own conclusion is that the binding variable is the price ratio rather than the technology.

12. Marek Miara and Jan Rosenow, “Why are heat pump installation costs so much higher in Germany? Evidence from a European comparison”, Energy & Buildings 369 (2026) 117998, https://doi.org/10.1016/j.enbuild.2026.117998. The specification-matched bilateral figures are theirs: a defined 8 kW air-to-water system in a representative existing home at €29 719 net in Germany against €12 095 in the United Kingdom, becoming €35 366 against €12 095 gross once 19% German VAT is applied against a UK zero rate, a gross gap of €23 271. Component contributions to that gap: equipment €2000–4000, foundations €1000–2200, electrical connection €1500–2500, labour €500–3000, VAT €5647. The pre-2020 baseline of about €17 000 for German systems up to 15 kW against roughly €12 500 in Britain and €11 000 in Sweden, the subsequent doubling in Germany against rises of 18% and 14%, and the attribution of the jump to the 2020 switch from capacity-based to percentage-based support, are from the analysis commissioned by the German federal economics ministry that the paper draws on. German averages are corroborated across two independent sources — a consumer-organisation survey of 160 actual offers in 2024–25 giving a mean near €36 000 across a €20 000–63 000 range, and a federal subsidy dataset of some 300 000 cases giving about €36 500. Three cautions. Sterling conversions use the mid-2026 rate of 1.16 euros to the pound given in chapter I and are rounded to the nearest £100; the paper’s own figures are in euros and are the primary ones. The subsidy attribution is a conclusion drawn from timing and from comparison with countries that had the same shocks and not the same policy, not from a controlled experiment — the authors are careful about this, and the residual category also contains market dynamics and an early refrigerant transition. And the German 46% sales fall is for 2024, whereas the market-share figure earlier in this chapter is for 2025, when sales recovered; the two are consistent, and reading either alone would mislead. The paper does not claim the cost premium caused that fall, and neither does the text: the drop is commonly attributed to demand pulled forward into 2023 and to uncertainty over the heating law, with cost as one contributor among several.

13. Marek Miara and Jan Rosenow, “Why are heat pump installation costs so much higher in Germany? Evidence from a European comparison”, Energy & Buildings 369 (2026) 117998, https://doi.org/10.1016/j.enbuild.2026.117998. The specification-matched bilateral figures are theirs: a defined 8 kW air-to-water system in a representative existing home at €29 719 net in Germany against €12 095 in the United Kingdom, becoming €35 366 against €12 095 gross once 19% German VAT is applied against a UK zero rate, a gross gap of €23 271. Component contributions to that gap: equipment €2000–4000, foundations €1000–2200, electrical connection €1500–2500, labour €500–3000, VAT €5647. The pre-2020 baseline of about €17 000 for German systems up to 15 kW against roughly €12 500 in Britain and €11 000 in Sweden, the subsequent doubling in Germany against rises of 18% and 14%, and the attribution of the jump to the 2020 switch from capacity-based to percentage-based support, are from the analysis commissioned by the German federal economics ministry that the paper draws on. German averages are corroborated across two independent sources — a consumer-organisation survey of 160 actual offers in 2024–25 giving a mean near €36 000 across a €20 000–63 000 range, and a federal subsidy dataset of some 300 000 cases giving about €36 500. Three cautions. Sterling conversions use the mid-2026 rate of 1.16 euros to the pound given in chapter I and are rounded to the nearest £100; the paper’s own figures are in euros and are the primary ones. The subsidy attribution is a conclusion drawn from timing and from comparison with countries that had the same shocks and not the same policy, not from a controlled experiment — the authors are careful about this, and the residual category also contains market dynamics and an early refrigerant transition. And the German 46% sales fall is for 2024, whereas the market-share figure earlier in this chapter is for 2025, when sales recovered; the two are consistent, and reading either alone would mislead. The paper does not claim the cost premium caused that fall, and neither does the text: the drop is commonly attributed to demand pulled forward into 2023 and to uncertainty over the heating law, with cost as one contributor among several.

14. UK installed-cost ranges for 2025–26 from trade and comparison sources: a single-room air-to-air system about £1800–£2600, whole-house multi-split £4000–£9500, against £8250–£15 500 for a typical air-to-water installation. These are advertised market ranges rather than audited survey data and vary widely with property and installer. Air-to-air units are reported by the European Heat Pump Association as the most common type sold across European markets, ahead of air-to-water; Norway recorded 48 sales per 1000 households and Finland 33 in 2024, with about 95% of Norwegian air-to-air units used for heating. The Boiler Upgrade Scheme grant of £7500 covers air-to-water and ground-source heat pumps and biomass boilers, not air-to-air, which is one reason UK air-to-air installations do not appear in the MCS-certified counts quoted earlier and why those counts understate the number of heat pumps actually running in British homes. MacKay’s own air-conditioner figures — 0.6 kW electricity for 2.6 kW of cooling — are from this chapter’s cooling section and imply a coefficient of performance of 4.3, though a cooling COP and a heating SPF are not directly comparable.

15. Tobias Alm, Energieffektivisering av Saltsjöbaden Grand Hotell, examensarbete 30 hp, Uppsala universitet, 2017: https://www.diva-portal.org/smash/get/diva2:1140546/FULLTEXT01.pdf. The hotel is heated by two lake-source heat pumps and two oil boilers; replacing one boiler with a third heat pump is costed at about 200 MWh/year saved for roughly 500 000 SEK, a payback near 2.5 years at an assumed electricity price of 1 SEK/kWh, with the full package of measures reaching about 500 MWh/year or a 30% reduction. This is a master’s thesis and a single building, not a controlled study, and the payback depends on the assumed electricity price and on oil being the incumbent; it is quoted as a worked example of the point being made rather than as a general result.

16. Swedish figures for bergvärme are from ordinary domestic practice as summarised in the Swedish-language literature: rated SCOP typically 4.0–5.5, payback of roughly four to five years against oil or direct electric heating and eight to ten against district heating, with the payback figures predating the 2022 energy price shock and therefore understating the case since. On the SCOP–SPF distinction: SCOP is defined for space heating on a standardised seasonal test, while SPF as used in the British field-trial literature covers the installed system in service including domestic hot water, which requires higher flow temperatures and lowers the annual average. The two numbers are not interchangeable and the gap between 4–5 and 2.81 is not all a difference in performance.

17. Swedavia’s description of the Arlanda aquifer thermal energy store: operating since summer 2009 in the Brunkebergsåsen ridge, described as the world’s largest of its kind, divided into warm and cold sections, saving about 19 GWh a year — which Swedavia equates to the consumption of 2000 single-family homes — at an airport using energy comparable to a town of 25 000. Return water is around 20°C and is used for snow melting and ventilation preheating; the groundwater volume is returned rather than consumed. Figures are the operator’s own and the 19 GWh is a saving against a counterfactual system rather than a metered output, so it is not directly comparable with a generation figure. Aquifer stores require permitting for groundwater use and suitable hydrogeology, and are not a technology that can be sited at will.

18. Gas absorption heat pumps achieve a gas utilisation efficiency of roughly 1.3–1.5 on manufacturers’ figures for air-source domestic and light-commercial units, against about 0.9 for a condensing boiler at realistic return temperatures; both figures are on higher heating value. The carbon comparison uses 183 g CO₂ per kWh of gas at the meter, MacKay’s own 500 g per kWh of electricity for 2008, and about 120 g per kWh for Britain in 2025 — the latter derived from the generation mix in the Energy Institute’s Statistical Review of World Energy 2026, treating biomass as zero at the stack, which flatters it. The electric heat pump uses the 2.81 seasonal performance factor measured in the Electrification of Heat trial rather than a manufacturer figure. All of these are annual averages: the gas heat pump’s advantage is largest in the coldest weather, when the electric machine’s performance falls and the grid is dirtiest, so the comparison is less lopsided at the moment of peak demand than the annual table suggests. The wider point stands regardless of the exact numbers — the ordering reversed because of the grid, not the machines.

19. The FREEDOM project, led by Wales & West Utilities with PassivSystems and Delta-EE, installed hybrid heat pump and boiler systems in 75 homes in Bridgend, south Wales, and reported in October 2018. Heat pump share of annual heat was 40–58% with carbon reductions of 30–48%, and the estimate of avoided electricity distribution reinforcement for South Wales was about £1.3 billion by 2050. That last figure is a modelled counterfactual produced by a gas network operator, which has an evident interest in a technology that keeps houses connected to the gas grid, and should be read with that in mind; the measured heat-pump shares are the more reliable part of the study. Peak electricity demand estimates for full heat electrification vary considerably with assumptions about fabric, control and diversity — published figures for a fully electrified GB peak range from about 100 to 125 GW against roughly 50 GW today, and actively managed heat pumps with thermal storage reduce it materially. The Electrification of Heat trial’s average seasonal performance factor of 2.81 across 428 heat pumps is DESNZ’s, and includes hybrids in the wider trial.

20. Directive (EU) 2024/1275, the recast Energy Performance of Buildings Directive, in force 28 May 2024: zero-emission new public buildings from 2028 and all new buildings from 2030, with no on-site fossil fuel use; an end to incentives for fossil-fuel boilers from January 2025 and a phase-out earmarked for 2040; transposition due by 29 May 2026. As a directive it binds member states to an outcome, not to a method, and national implementations differ.

21. LEED v5, US Green Building Council, opened for registration 28 April 2025: roughly half of available points allocated to decarbonisation, with mandatory carbon, climate-resilience and human-impact assessments and a five-year update cycle. LEED is a voluntary certification scheme rather than a building code, and certification measures design intent and modelled performance, which is not the same as metered energy in use.