There is a device attached to your building that has been in continuous commercial deployment, in recognisably the same form, since 1889. No other consumer-facing technology comes close. The telephone has been reinvented four times. The car engine has been redesigned repeatedly. The electricity meter spun a metal disc for a hundred years.

Then, over roughly the last two decades, it became a computer — one with firmware, a communications stack, a data model, and a backend system it reports to. Most people never noticed, because the box looks much the same and still sits in the same cupboard.

The interesting question is not how the technology changed. It is why it took a century, and then happened all at once.

The problem of selling something invisible

When Thomas Edison began supplying electricity commercially, he faced a problem that had no precedent: how do you bill for a product the customer cannot see, count, or verify?

His first answer was to avoid measuring it at all. Early customers were charged per lamp — a flat fee for each fixture connected, regardless of how long it burned. This was administratively simple and commercially useless, bearing no relationship to actual consumption.

His second answer was chemistry. Edison's meter, patented in 1881 (US patent 251,545), used the electrochemical effect of current. A sealed cell containing an electrolyte and metal electrodes was installed at the premises. As current flowed, metal was transferred from one electrode to the other at a fixed, known rate. At the end of the billing period a meter reader physically removed the electrodes and weighed them on a laboratory balance; the difference in mass was the customer's consumption.

Two details are worth pausing on.

The first is that the meter was deliberately calibrated so that bills could be rendered in cubic feet of gas. Edison was not selling a new product category — he was selling a substitute for gaslight, and he priced it in the units his customers already understood. The unit of account followed the incumbent technology, not the physics.

The second is that the meter was unreadable. The utility could only determine consumption in a laboratory, and the customer could not verify anything at all. Mishandling of the plates produced billing errors in at least a few documented cases, and there was no way for a customer to check the result. A meter that neither party can read is a trust problem wearing the costume of a measurement problem. Edison later bolted on a counting mechanism to make readings easier, and other electrolytic designs followed — the Siemens-Schuckert hydrogen meter, the Schott & Gen. Jena mercury meter — but they shared the same fundamental defect. Worse, electrolytic meters could measure only ampere-hours, which is only proportional to energy if the supply voltage stays constant.

Alternatives multiplied through the 1880s. Ayrton and Perry described a pendulum meter principle in 1881; Hermann Aron independently built one in Germany in 1884. Ferranti offered a mercury motor meter in 1885 with a register resembling a gas meter's, which had the significant advantage that a customer could walk up and read it.

The disc

The design that won emerged from the alternating-current side of the industry, and it emerged three times in about eighteen months.

Oliver Shallenberger, chief electrician at Westinghouse, received a patent for an AC meter on 14 August 1888. Elihu Thomson at General Electric developed a recording watt-hour meter around 1889 using an ironless commutator motor, which worked on both AC and DC and avoided the weighing problem entirely. And in the autumn of 1889, Ottó Bláthy — already co-inventor of the ZBD transformer at the Ganz Works in Budapest — patented the AC watt-hour meter that became the template for everything after it.

The principle is elegant enough to state in a sentence: two electromagnets, one driven by voltage and one by current, induce eddy currents in an aluminium disc, and the disc rotates at a speed proportional to the power passing through the circuit. A gear train counts the revolutions. A permanent magnet provides braking so the relationship stays linear. The meter integrates power over time mechanically, with no electronics and no external power source, and it displays its result on a dial anyone can read.

Ganz showed the first Bláthy meter at the Frankfurt Fair in autumn 1889 and was selling induction kilowatt-hour meters by the end of that year. The early units were mounted on wooden bases, weighed 23 kg, and spun at 240 revolutions per minute at full load — fast enough that the count was hard to read. By the early 1920s the weight had fallen to roughly a tenth of that.

Then, essentially, the design stopped changing. Shallenberger's induction approach was refined into modern watt-hour form by 1894. Accuracy improved. Manufacturing got cheaper. But the AC kilowatt-hour meters installed in the 2000s still operated on Bláthy's principle — a hundred and ten years of the same physics, in the same product, sold to the same customers.

The century of stasis

This is the part of the story that engineering histories tend to skip, and it is the part that actually explains the timeline.

The induction meter did not survive a century because it could not be improved. Electronics capable of replacing it existed by the 1970s. It survived because three forces held it in place, and none of them were technical.

It was a legal instrument, not a product. An electricity meter determines a payment between two parties, which makes it subject to legal metrology — the same body of law governing petrol pumps and butchers' scales. In most European countries a meter must be type-approved by a national metrology authority before it can be installed, and verified periodically thereafter. Changing the design means re-entering an approval process measured in years. This is a powerful brake, and a deliberate one: the whole point is that the instrument you are billed by cannot be quietly altered.

The business model demanded nothing more. A regulated monopoly selling a commodity at a flat regulated tariff needs exactly one number per customer per billing period. Not when the electricity was used, not what it was used for — just the total. The induction meter delivered that number with no operating cost, no power supply, and a service life of decades. Any additional data would have been information the utility had no commercial use for and no tariff structure to apply it to.

Nobody was competing for the customer. Where the supplier cannot be switched, there is no need to reconcile consumption between competing retailers, no need for accurate switching reads, and no need for a shared data format between companies. The absence of a market is the absence of an integration problem.

Stasis was rational. The meter was not backward; it was correctly sized to the industry around it.

What broke

Three things changed the requirement at roughly the same time.

Liberalisation split the vertically integrated utility. Once the wires company and the retailer are different legal entities, and a customer can switch retailers, meter data stops being an internal accounting detail and becomes a settlement record exchanged between competitors. It now needs to be timely, disputable, and standardised.

Time-of-use pricing made timing billable. If electricity costs different amounts at different hours, a single monthly total is no longer a valid basis for a bill. The meter must record a time series, which means it needs a clock, memory, and a tariff calendar someone can update.

Distributed generation reversed the flow. A house with rooftop solar exports power. A meter designed to count in one direction cannot describe that relationship, and a utility that cannot measure export cannot pay for it.

Each of these turned "how much" into "how much, when, and in which direction" — and that is a computer, not a disc.

The Italian exception

The first mass deployment did not come from a regulator or a technology vendor. It came from an Italian utility solving a labour problem.

Enel began its Telegestore rollout in 2001 and had replaced roughly 30 million meters by 2006 — at the time, and for years afterwards, by far the largest single-utility deployment in the world. The original goal was not the smart grid. It was to eliminate the workforce that walked door to door reading mechanical meters, to detect theft, and to reduce the cost of physical site visits for connections and disconnections.

The architecture is worth noting because it prefigured everything since: meters communicating over the power line itself to roughly 400,000 concentrators installed in secondary substations, which in turn reached a central head-end system over public telecoms networks. The link speed was 2400 bit/s, half-duplex — a rounding error by modern standards, and entirely sufficient, because the payload was small and the latency requirement was hours.

Italy's regulator made installation mandatory in 2006, and by 2011 roughly 95% of Italian metering points were covered — the European Union's 2020 target reached nearly a decade early, by a country that started before the target existed.

The communications protocol Enel used was proprietary. It was standardised later, retroactively, once other utilities wanted to buy the same thing. This is the pattern the rest of the industry then spent fifteen years working through in the opposite order.

The actual hard problem

Here is the part that surprises people who assume the difficulty in smart metering is the meter.

A meter is a straightforward embedded device: measurement front end, microcontroller, non-volatile memory, radio or power-line modem, tamper detection. None of that is research. The hard problem is that a Swiss-manufactured meter, a German head-end system, a French meter data management platform and an Italian billing system must all agree on what a "meter reading" is — its identity, its timestamp semantics, its unit, its quality flags, its relationship to a metering point, a customer, a contract, and a physical asset.

That agreement is the Common Information Model. Its origins are older than smart metering and come from a different problem entirely: EPRI convened a Control Center Application Programming Interface task force in 1993, motivated by the observation that no single vendor could supply a utility's entire control-room stack, and that integrating multiple vendors was consuming the value of buying from them. The resulting semantic model, expressed in UML, was adopted by IEC Technical Committee 57 and published as IEC 61970-301 for energy management systems.

The model was then extended outward. IEC 61968 carried it from the transmission control room into distribution — asset tracking, work scheduling, customer billing — and IEC 61968-9 specifically defines interfaces for meter reading and control. IEC 62325 extended it again to electricity market participants. Together these describe, in a single consistent object model, an industry that had previously described itself in several hundred incompatible ones.

This is why metering became a distributed systems problem. Not because the meter got clever, but because thirty million endpoints reporting into an ecosystem of separately procured, separately owned software systems is an integration problem of a kind the electricity industry had never previously faced. The engineering effort went into the semantics, not the silicon.

Arriving late, on purpose

The European Union set its target in the Third Energy Package of 2009: member states were to roll out smart metering to at least 80% of electricity customers by 2020, in cases where a national cost-benefit analysis came out positive, with those analyses due by September 2012.

By the Commission's own 2014 assessment, the commitments amounted to around €45 billion for close to 200 million electricity meters — covering roughly 72% of European consumers, short of the headline 80%. At that point Finland, Italy and Sweden between them accounted for close to 45 million meters already installed, about 23% of everything envisaged across the EU. Germany, after running its cost-benefit analysis, opted for a selective rollout rather than a universal one.

The 80% target was not met EU-wide by 2020.

The reasons are the same three forces that produced the century of stasis, arriving in modern dress. Legal metrology still requires type approval and verification, now for a device with updatable firmware — which raises a genuinely difficult question about what exactly was approved. Data protection law applies, because half-hourly consumption data reveals when a household is occupied, awake, cooking, or away, and that is personal data under any reasonable reading. And utility procurement runs on multi-decade asset cycles that do not accelerate because a directive says so.

None of these are engineering failures. They are the cost of the meter being a legal instrument in a regulated industry rather than a consumer gadget — the same property that kept it unchanged for a century.

What the box on the wall is now

The device measures energy in both directions, timestamps it, stores an interval series, applies a tariff calendar, detects tampering, reports outages, accepts remote firmware updates, and can disconnect the supply on command from a system hundreds of kilometres away. It is a networked embedded computer with a regulatory certificate, sitting in a cupboard, doing a job that a spinning aluminium disc did adequately for a hundred years.

Whether that is progress depends on what you think it is for. As a billing instrument, the improvement over the disc is marginal — both produce a number. Its value is in what the old meter could not express at all: that electricity consumed at 3 a.m. and electricity consumed at 6 p.m. are different products, that a building can be a supplier as well as a customer, and that a grid running on weather-dependent generation needs to know what its edge is doing more often than once a month.

Edison priced electricity in cubic feet of gas because that was the frame available to him. The induction meter counted a monthly total because that was what the industry was structured to sell. The meter has always described the commercial arrangement of its era rather more precisely than it described the physics — and that, more than any change in the technology, is what actually changed.