Heat Pumps Are Not Witchcraft

A machine that provides more energy than you feed it sounds, on first hearing, like a scam. We understand the suspicion — a healthy allergy to free lunches is the correct default setting, and it's the same instinct we apply to every glossy brochure that crosses our desk. So let's be clear about what this article is: not a sales pitch, but an explanation of a nineteenth-century machine you already own at least one of, and why its arithmetic is not merely legal but boring.

That machine is your fridge. A fridge pumps heat out of an insulated box and dumps it into your kitchen through the coils round the back. A heat pump is the same device pointed the other way: it pumps heat out of the great outdoors and dumps it into your house. If your fridge has never struck you as sorcery, you're already most of the way there.

Older than the light bulb

None of this is new, which is worth dwelling on, because "unproven technology" is one of the more confident objections we hear.

The theory arrived in 1824, when the French engineer Sadi Carnot worked out the rules governing heat engines. In 1852, William Thomson — later Lord Kelvin — pointed out that you could run the logic in reverse: instead of using heat to produce work, use work to move heat from somewhere cold to somewhere warm. He called it a "heat multiplier" and noted, correctly, that it would beat any heater that merely made heat.[1]

Four years later it stopped being theory. In 1856 the Austrian engineer Peter von Rittinger built the first working heat pump — not to warm a parlour, but to dry salt at the Ebensee salt works, where his calculations promised energy savings of up to 80% over burning wood.[1] Industrial process engineering, driven by fuel costs. Some things don't change.

By 1927, the Scottish engineer T.G.N. Haldane had patented and demonstrated a vapour-compression heat pump heating his own home and office — in Scotland, a place not noted for its gentle winters.[1] Then the same technology went into mass production as refrigeration and air conditioning, and the world quietly built billions of the things. Edison's light bulb, for reference, arrived in 1879. The heat pump predates it and has had rather less trouble with public acceptance in its fridge-shaped disguise.

How it actually works

The whole trick rests on one piece of everyday physics: the boiling point of a liquid depends on pressure. Water boils at 100°C at sea level, but at around 70°C on top of Everest. Squeeze a vapour and it condenses at a higher temperature; release the pressure on a liquid and it boils at a lower one.

A heat pump exploits this with a refrigerant chosen to boil at brutally low temperatures. Four components, in a loop:

1. Evaporator (outside): liquid refrigerant at low pressure boils at perhaps −20°C or −30°C. Even a miserable 2°C January day is hot compared with that, so heat flows from the outside air into the refrigerant, boiling it into a vapour. This is the step people trip over — "there's no heat in cold air" — but temperature is relative. Air at 2°C is stuffed with thermal energy compared with air at −25°C; the refrigerant is simply colder than the weather.
2. Compressor: an electric pump squeezes the vapour. Compressing a gas heats it — the same reason a bicycle pump gets warm. The vapour leaves at 60–80°C. This is where your electricity goes: not making heat, but concentrating it.
3. Condenser (inside): the hot, high-pressure vapour condenses back to liquid against your radiators or underfloor loops, handing over both the heat it collected outside and the heat added by compression.
4. Expansion valve: the liquid drops back to low pressure, gets very cold again, and returns to the evaporator for another pass.

Round and round, several times a minute, entirely described by thermodynamics that was settled before your great-great-grandparents were born.

A lever, not a lifestyle statement

The way we prefer to frame it: a heat pump is an energy lever. Every kilowatt-hour of electricity you put on the short end lifts a multiple of itself in heat on the long end — how big a multiple depends on the design, the emitters and the weather, and pinning that number down honestly is what the rest of this article is about. It is the only heating appliance you can buy where the laws of physics work in your favour rather than merely breaking even.

And the lever multiplies whatever you feed it. A kWh from your own solar array becomes three kWh of heat. As the grid's generation gets cleaner — and per our embodied energy article, the British grid average is now around 126 g CO₂ per kWh — the same lever multiplies progressively cleaner energy, no further action required on your part. A boiler, by contrast, burns gas at slightly under one-to-one forever, and no amount of grid progress improves it by a single percent.

So much for theory — what do they do in the field, bolted to real damp British walls by installers of variable talent? The largest UK trial to date, the Electrification of Heat project, monitored 742 air-source heat pumps in ordinary homes — terraces, semis, flats, the lot. Measured median seasonal performance: 2.80 units of heat per unit of electricity across the year, and 2.44 on the coldest days. Roughly three times the efficiency of a gas boiler, in real houses, including the badly-behaved ones — and a step up of 0.3–0.4 on the same trial's predecessor a decade earlier.[2] The spread around that median is the interesting part: system design and commissioning, not the badge on the box, decide whether you get a 2.2 or a 3.5. This is why we bang on about design.

Nobody told Norway it doesn't work

If heat pumps were a fair-weather technology, you'd expect adoption to peter out somewhere around the Alps. The opposite is true: the coldest countries in Europe like them most. In Norway, roughly two-thirds of households heat with a heat pump — in a country where winter is not a rumour.[3] Europe's installed stock stands at about 29 million units across the 21 countries the industry association tracks, with sales up 13% last year.[4]

Globally, this is now a heavy industry, not a cottage one. The IEA's Heat Pump Monitor puts 2025 sales at 108 GW of capacity, with heat pumps covering about 12% of the world's space heating in buildings, and worldwide manufacturing capacity at roughly 145 GW a year. The heat pumps already running avoided an estimated 53 billion cubic metres of natural gas in 2025 across Europe, Japan and China — without them, Europe's gas consumption would be more than 10% higher.[5] Whatever your view on the politics of energy, displacing imported gas with locally-generated electricity at a ratio of one-to-three is simply good engineering.

The UK, it must be said, is late to this party: 51,886 certified retrofit installations in 2025, up 7% on the year before.[6] Growing, but from a small base — which mostly reflects our unusually gas-dependent housing stock and pricing, not any special property of British air.

"But my house isn't modern enough"

This one deserves its own section, because it's the most confidently repeated myth in British heating: the idea that heat pumps are only for hermetically sealed newbuilds, and that anything older, draughtier or more characterful is somehow exempt from thermodynamics. The Electrification of Heat trial went looking for exactly this and its finding was blunt: all housing types are suitable for heat pumps — that's the project's own headline conclusion, drawn from installs across pre-war terraces, semis, flats and everything between. The trial also found that high-temperature air-source units, the kind fitted where big radiator surgery isn't wanted, performed with efficiencies similar to their low-temperature siblings.[2]

The reason is that a heat pump doesn't care about your EPC rating, your conservation area or the year your house was built. It cares about exactly two numbers: how fast the building loses heat on the coldest design day, and what flow temperature your emitters need to replace that heat. Specify the machine and the emitters honestly against those two numbers and any building can be served — from a super-insulated box to a genuinely draughty castle. The castle simply needs a bigger lever. Notice that nobody has ever declared a house "too draughty for a gas boiler"; the fitter just installs a bigger boiler and the fuel bill quietly says the rest. Same physics, same answer, different fuel.

What insulation actually changes is the size of the machine, the radiators and the running bill — which is why fabric improvements are worth making whatever you heat with, and why we survey and calculate heat loss room by room before specifying anything. "Your house can't have a heat pump" almost always translates as "I'd rather not do the design work."

The honest arithmetic

Here is the part the brochures skip, so we won't. A heat pump's running cost depends on the ratio between electricity and whatever fuel you'd otherwise burn. Under the current price cap, electricity costs 26.11p per kWh and gas 7.33p — about 3.6 to one, which is unhelpful.[7] And if you're off the gas grid — a large slice of our corner of the South West — your comparison is heating oil, which deserves a row of its own: kerosene has bounced between roughly 64p and 95p a litre in mid-2026 alone; we'll use 80p as a working figure, and a litre holds about 10.35 kWh.[8] Run the numbers:

So at standard tariff rates: a median-performance heat pump costs slightly more to run than a good gas boiler (break-even is a seasonal COP of about 3.2), sits level with oil at mid-range prices, and a properly designed system beats both. That's the truth, and anyone who tells you a heat pump will automatically slash your bills at today's standard prices is selling something. Note what the oil row hides, too: that 8.6p was 7p in early July and over 10p by the end of the month — oil customers don't have a unit rate, they have a lottery ticket and a delivery lorry.[8]

But the standard tariff is the worst case, because a heat pump is the ideal customer for time-of-use pricing — heat can be made when electricity is cheap and stored in the fabric, the cylinder or a buffer. Heat-pump tariffs currently offer around 13p/kWh across three off-peak windows a day in exchange for a dearer teatime peak;[9] overnight EV-style tariffs go as low as 7p.[10] Re-run the table with those:

Read that last row again: under 3p per kWh of heat — roughly a third of the gas boiler's cost — for a household that pairs the lever with a battery and a smart tariff. The honest caveats attached: those tariffs demand some scheduling discipline (or kit that automates it), and tariff rates are commercial offers that change — we've referenced today's, and you should check the current ones. On battery size: winter heating demand is large, so a token battery will only ever cover a slice of it — but that's a sizing decision, not a law of nature. We regularly size storage to carry a whole house, heat pump included, through a worst-case winter day entirely on overnight charge. Proper system design, again. Three further tilts, all real: ditching gas entirely removes the gas standing charge (about £106 a year at current cap rates[7]); pairing with your own solar feeds the lever kilowatt-hours at whatever your panels cost you; and the political elephant — UK electricity prices carry levies that gas largely escapes. If that ever gets rebalanced, as it already has in countries where heat pumps are the cheap option, the arithmetic moves decisively in one direction.

The other honest caveats: performance drops as the temperature gap widens, so the coldest day of the year is also the machine's hardest (2.44, measured, not zero — Norway, again). In a leaky house with postage-stamp radiators, the honest design answer is bigger emitters or higher flow temperatures, and the higher the flow temperature the lower the COP — a question of economics, never of feasibility, as covered above. And an incompetent installation will cheerfully turn a 3.5 machine into a 2.2 one. None of these are exotic failure modes; they're the same "design it properly" rules that apply to every other piece of building services, applied to a machine the public has only recently been asked to have opinions about.

The verdict

A heat pump is a 170-year-old idea, mass-produced for a century in fridge form, measured in the field at roughly three times the efficiency of the best possible boiler, and deployed most enthusiastically in the coldest corners of Europe. The physics is settled, the machinery is mature, and the running-cost arithmetic is a straightforward function of tariffs and design quality — favourable if you do it properly, unfavourable if you don't. That's not witchcraft. That's engineering, and engineering rewards being done well.

What happens to the box — and the refrigerant inside it — when it eventually wears out is a subject we'll take apart in a separate article.