Impermanent Solution

Imagine studying engineering for years. You learn how materials fatigue and how they fail. You learn tolerances, load margins, heat dissipation, and why a weld cracks. You learn to remove a gram of steel without losing strength, to cut a watt of waste from a motor, to find the weakest part of a machine and make it stronger. Every exam, every lab report, every late night pushes in one direction: make it work better, make it last longer.

Then you are hired. Your first important assignment arrives on your desk, and it asks you to make the product worse.

Not visibly worse. Not badly made. It must still work well, well enough that the customer trusts the brand and would buy it again. The finish must be clean and the performance convincing on the first day and the hundredth. It simply needs to stop working sooner.

The task sounds absurd, until you try to carry it out. A product built carelessly fails at random: one unit dies in a week, another survives for twenty years, and the company's reputation dies with the first one. A product that fails predictably, close to a chosen date, across millions of units, is much harder to make. It requires knowing exactly how the material degrades, how heat accelerates it, how much variation the factory introduces, and how to hold all of it inside a narrow window. Making something last forever is difficult. Making something last precisely long enough may be more difficult.

This is not a thought experiment. Something very close to it happened, it was written down, and it began with a lightbulb.

A Bulb in a Fire Station

In a fire station in Livermore, California, a small hand-blown bulb hangs from the ceiling. It was made by the Shelby Electric Company, it has a carbon filament, and it has been glowing, with only a few interruptions, since 1901. It was moved to its present home, Fire Station 6, in 1976. In May 2013 a failed power supply left it dark for about nine and a half hours, and when it was reconnected it came back on. Guinness World Records lists it as the oldest known working lightbulb. ^[5]

The Centennial Light is often presented as proof that manufacturers have been hiding the secret of the eternal bulb. It is not that, and it should be handled carefully.

Its conditions are unusual. The bulb was reportedly rated around 60 watts when new; today it runs at roughly 4 watts. ^[5] It gives off a dim glow, closer to a night light than to a reading lamp. A filament run far below its designed temperature evaporates extremely slowly, and a bulb that is almost never switched off avoids the repeated heating and cooling that crack filaments in ordinary use. Carbon filament lamps of that era were also much less efficient than the tungsten lamps that replaced them. And the Centennial Light is one survivor among the millions of bulbs made around 1900, nearly all of which burned out long ago. Nobody built a museum for those.

So the bulb does not prove that modern bulbs were deliberately weakened. What it demonstrates is simpler and harder to dismiss:

We have known, for more than a century, how to make certain objects last a very long time, if we are willing to accept the trade-offs that come with it.

The interesting question is who decides which trade-offs to accept, and why.

"My Old Machine Lasted Thirty Years"

Most people have heard some version of the sentence. A parent or grandparent, standing next to a new appliance that has already failed: my old washing machine lasted thirty years. These new ones break every few years.

It is tempting to treat this as evidence. It is weaker evidence than it sounds.

The first problem is survivorship bias. The old machines people remember are the ones that survived. The ones that broke in their third year were hauled away decades ago, and nobody tells stories about them. A basement full of working appliances from 1975 says little about the appliances that did not make it to 1985. Memory keeps the survivors and quietly deletes the rest.

The second problem is that "old" did not always mean "better." Many older appliances used far more water and electricity. Some were dangerous: exposed wiring, no residual-current protection, wringers that crushed fingers, refrigerants that damaged the ozone layer. A machine that outlived its owner's patience could still be a worse machine by most measures that matter today.

And yet the perception is not simply nostalgia. When researchers actually looked, part of it held up.

In 2016 the German Environment Agency (UBA) published a study by the Öko-Institut and the University of Bonn on the lifetimes of electrical products. Among large household appliances replaced because of a defect, the share replaced within less than five years of purchase rose from 3.5% in 2004 to 8.3% in 2013. About a third of consumers surveyed said they were dissatisfied with how long their appliances lasted. ^[9]

The same study also complicated the story. Most appliances were not replaced because they broke: in many categories the leading reason was that the owner wanted a newer device, even when the old one still worked. And the researchers found no evidence that manufacturers deliberately designed products to fail early. What they found instead was that products are designed for particular price points and product cycles, and that fast innovation cycles can leave less time for thorough testing. ^[9]

Durability did change. Deliberate sabotage was not shown. Both findings deserve to be held at once.

There is also something the nostalgic sentence gets right that has little to do with how long a motor lasts. Many older appliances were easier to keep alive. They were mechanically simpler. They opened with ordinary screwdrivers. Their belts, bearings, brushes and pumps were often standard parts sold by independent suppliers, and a local repair shop could fix them with a manual and an afternoon. They did not depend on a proprietary control board, a firmware update, or a server somewhere in another country.

An old machine did not necessarily fail less often. It was more often allowed to be repaired when it did.

Naming the Thing

What the grandparent is describing, and what the imaginary engineer was asked to do, has a name.

Planned obsolescence is the practice of designing or managing a product so that it becomes unusable, unrepairable, or undesirable sooner than its technology requires, in order to prompt a replacement purchase.

The phrase was popularized, and possibly coined, in a 1932 pamphlet by a New York real estate broker named Bernard London, titled Ending the Depression Through Planned Obsolescence. London was not warning against the idea. He was proposing it as public policy: during the Depression, he argued, people were using their cars, radios and clothing for too long, and the government should assign products a legal lifespan, after which they would be retired and replaced to keep factories running. ^[8]

The proposal never became law. But by the time London wrote it, the lightbulb industry had already been doing something similar for seven years, privately, and with considerably more technical skill.

The Phoebus Cartel

On 23 December 1924, representatives of the world's major lamp manufacturers met in Geneva and signed the "Convention for the Development and Progress of the International Incandescent Electric Lamp Industry." It created a Swiss company, Phoebus S.A., to administer the agreement. ^[1,2]

The members included Osram of Germany, Philips of the Netherlands, the Compagnie des Lampes of France, Tungsram of Hungary, Associated Electrical Industries of the United Kingdom, and Tokyo Electric of Japan. General Electric of the United States took part less directly, through its international affiliates and holdings. ^[1] Between them they controlled most of the world's lightbulb production.

A cartel coordinates what competitors would otherwise fight over, and Phoebus coordinated almost everything. It divided the world into territories and assigned sales quotas. It managed patents and the exchange of technical information. It influenced prices. ^[1,3] Those arrangements were the ordinary business of an interwar cartel. What made Phoebus remarkable was one committee.

One thousand hours

Before the cartel, household incandescent bulbs commonly lasted around 1,500 to 2,000 hours or more. ^[1] Early in 1925 Phoebus established a "1,000 Hour Life Committee," and the standard life of a general-service household lamp was fixed at 1,000 hours. ^[1,2]

The rule was not a recommendation. Each member factory had to send samples of its production to a central testing laboratory in Switzerland, where the bulbs were burned until they failed and their lifetimes recorded. Manufacturers whose lamps lasted significantly longer than the standard were fined, on a schedule in Swiss francs that rose with the number of excess hours. ^[1,2] Popular retellings often quote specific amounts from this schedule; the figures vary between accounts, and the underlying cartel documents are discussed mainly in Markus Krajewski's historical work on Phoebus, so the precise numbers are best treated with caution. The existence of testing and fines is well documented.

Krajewski quotes a letter in which Anton Philips, head of Philips, complained about lamps being supplied into his company's market: "After the very strenuous efforts we made to emerge from a period of long life lamps, it is of the greatest importance that we do not sink back into the same mire by paying no attention to voltages and supplying lamps that will have a very prolonged life." ^[1]

The results showed up in the data. According to Krajewski, the average life of lamps produced by cartel members fell from about 1,800 hours in 1926 to 1,205 hours in the fiscal year 1933–34. ^[1]

The engineering of failure

This is where the imaginary engineer from the opening stops being imaginary.

The Phoebus engineers were not told to build bad bulbs. A bad bulb is easy: a thin spot in the filament, a poor seal, a flaw in the glass. Such bulbs fail unpredictably, and customers notice. What the cartel needed was harder. It needed bulbs that failed consistently, close to the same number of hours, across factories in different countries using different machines.

To do that, engineers adjusted the filament's material, thickness, shape and uniformity, and the relationship between a lamp's rated voltage and the voltage it would actually receive. ^[1,2] The physics made this lever unusually powerful. An incandescent filament is a wire heated until it glows. Run it hotter and it gives more light per watt, but tungsten evaporates from it faster, and it breaks sooner. The relationship is steep. Lamp engineers work with approximate rules of thumb of the form

where $V_0$ is the lamp's rated voltage, $V$ the voltage actually applied, $L$ and $L_0$ the resulting and rated life, $\Phi$ and $\Phi_0$ the light output, and $n$ an empirical exponent commonly taken to be around 12 or 13. ^[6,7] These are approximations, not laws of nature, but they show the scale of the effect. With $n = 12$, running a filament just 5% above its design point raises light output by roughly 19% and cuts its life by roughly 44%. A small change in design, invisible to any customer, moves the moment of death by hundreds of hours.

That is what made the 1,000-hour target achievable, and it is also what makes the story more complicated than it first appears.

What is documented, and what is disputed

Because brightness and life trade against each other so directly, there is no single "correct" lifespan for an incandescent bulb. A longer-lived bulb gives less light for the same electricity, which means higher running costs for the customer. A shorter-lived bulb is brighter and more efficient, but must be replaced more often.

Phoebus members used exactly this argument in their own defense, and not all of it was cynical. In 1951 the United Kingdom's Monopolies and Restrictive Practices Commission examined the lamp industry. It concluded that any standard life "must always represent a compromise between conflicting factors," accepted the manufacturers' view that 1,000 hours was a reasonable compromise at that time, and dismissed the allegation that the standard had been set to increase replacement sales. ^[4]

An American court read the evidence differently. In United States v. General Electric Co. (1949), a federal court in New Jersey found that General Electric and its partners had violated the Sherman Antitrust Act in the incandescent lamp industry. Among its findings, the court examined how lamp life had been set and concluded that the choice was driven substantially by profit rather than by the customer's interest alone. ^[3]

Both readings can be partly true. A 1,000-hour lamp may have been a defensible engineering compromise. The cartel documents also show companies monitoring each other's lamps, penalizing those that lasted too long, and describing long-lived lamps as a "mire" to be escaped. ^[1,2] Whatever the engineering merits of the number, it was not chosen freely by competing firms each trying to win customers. It was imposed, tested and enforced collectively, precisely so that no member could compete by selling a longer-lasting bulb.

It is also worth separating the history from its popular afterlife. Phoebus did not create an eternal bulb and then hide it. The Centennial Light is not the suppressed product that Phoebus killed. Modern bulbs are not short-lived because of a secret agreement from 1924; the cartel lost its force during the Second World War, around 1940, well before its planned expiry in 1955. ^[1] What Phoebus shows is narrower, and in some ways more unsettling: that an industry once sat down, measured its products' lifespans, decided they were too long, and organized a system to make them shorter.

Many Ways to End

Phoebus is the clean case, the one where lifespan was a number written into a contract. Most modern obsolescence looks nothing like it. A product can reach the end of its useful life in many ways that have nothing to do with a filament burning through.

Some things simply wear out. Rubber seals harden, bearings wear, capacitors dry, hinges fatigue. That is physical degradation, and every object is subject to it; the question is only how quickly, and whether the worn part can be replaced.

Increasingly, that second question decides everything. A phone whose battery has lost half its capacity is a phone with one failed component, but if the battery is glued behind a sealed glass back and a replacement costs a large fraction of a new device, the owner will reasonably buy a new one. The battery failed; the phone was discarded. The same pattern repeats when replacement parts are not sold to the public, when a product is held together with adhesive instead of screws, or when the screws that do exist require a proprietary driver. In its 2021 report to Congress, Nixing the Fix, the U.S. Federal Trade Commission described restrictions of exactly this kind (adhesives that make parts hard to replace, limited availability of spare parts, withheld diagnostic software) and found "scant evidence" to support most of the justifications manufacturers offered for them. ^[11]

Other products stop working without anything inside them changing at all. A laptop runs perfectly well until its operating system stops receiving security updates, after which connecting it to the internet becomes a risk. An app is no longer supported on an older version of the system. A peripheral no longer has drivers. A device that depends on a company's servers becomes an ornament when the servers are switched off. In 2016 Nest, then owned by Google, shut down the servers behind the Revolv smart-home hub, which had been sold with a "lifetime" service; on 15 May the hubs stopped working. Nest offered refunds after the public reaction. ^[12] The hardware was intact. Its reason for existing was elsewhere.

Some obsolescence lives in the edges. A connector changes, and a drawer of working chargers and cables becomes useless. A printer is sold cheaply and its ink sold expensively, with cartridges that the printer recognizes and third-party cartridges it may refuse. And some obsolescence is not technical at all: a phone, a car, a jacket can become socially outdated, unfashionable or embarrassing, long before it physically fails. That form of obsolescence is manufactured by marketing rather than by engineering, and it may account for more replacements than any broken part. The German study found that wanting a newer device was the most common reason for replacing many electronic products. ^[9]

Even software can blur the line between a trade-off and a trap. In 2017 Apple released iPhone updates that slowed some older models whose batteries had degraded. Apple said the feature prevented unexpected shutdowns, which is a real problem with aging lithium-ion cells. In 2020 French authorities fined the company €25 million, not for planned obsolescence, which the investigation did not establish, but for failing to tell users that an update could slow their phones and leaving them no way to undo it. ^[13] The technical measure may have been reasonable. What the customer was not told was the part that became illegal.

Short-lived is not the same as sabotaged

It would be easy to see planned obsolescence everywhere, and that would be a mistake.

Many products last less long for reasons that have nothing to do with a plan to make them fail. Cheaper products use cheaper materials, and many buyers would rather pay less now than more for a machine that outlives their next move. Miniaturization makes components harder to separate: a thinner phone is harder to open, and waterproofing often means sealing. Safety and efficiency rules change designs in ways that add parts and complexity. Electronics replaced mechanical controls because they allow features customers wanted, and electronics fail differently from mechanisms. Competition pushes prices down and margins with them. Engineering is the art of trade-offs, and durability is one variable among many.

The distinction matters because the remedy is different. A product that fails early because it was made cheaply is a question of price and information. A product that fails early because failure was chosen, or because repair was deliberately made impossible, is a question of incentives.

The Incentive Problem

It is often said that planned obsolescence is necessary for capitalism. That claim is too simple. Companies can build profitable businesses around durable products. They can charge a premium for longevity, sell maintenance and repairs, sell spare parts and upgrades, offer services and subscriptions, and win new customers with a reputation for reliability. Some of the most respected brands in several industries are built on exactly this.

But consider a company whose revenue depends heavily on selling units. For that company, a customer whose product never needs replacing has a quiet, uncomfortable property: they may never become a customer again.

From the business's point of view, the cycle that keeps it alive looks like this:

product → purchase → replacement → another purchase

From the planet's point of view, the same cycle looks different:

extraction → manufacturing → transport → consumption → waste → extraction again

The first cycle measures success by how often it turns. The second cycle pays for every turn.

Every replacement that technology did not require carries costs that never appear on the shelf price. Ore is mined for copper, cobalt and rare earths. Oil becomes plastic. Energy is spent in smelters, chip fabs, assembly lines and container ships. Packaging is made and thrown away. And at the end, the discarded object becomes waste. In 2022 the world generated a record 62 million tonnes of electronic waste, about 7.8 kilograms per person. Only 22.3% of it was documented as formally collected and recycled. The rest was stored, landfilled, burned, or handled informally, often in ways that release lead, mercury and other toxic substances. E-waste is growing almost five times faster than documented recycling and is projected to reach 82 million tonnes by 2030. ^[10]

The same German study that found no deliberate sabotage offered a precise illustration of the cost of short lifetimes. A washing machine used for five years instead of twenty has about 40% higher energy demand and global warming potential over that period, because the energy and materials spent manufacturing each new machine outweigh the efficiency gains of newer models. ^[9] Replacement is not free just because the new product is more efficient.

Changing the Incentives

If the problem is incentives, the response is to change them, and in the past decade governments have begun to try.

In 2015 France became one of the first countries to make planned obsolescence a criminal offense. Article L441-2 of its Consumer Code prohibits techniques, including software, by which a manufacturer deliberately aims to reduce a product's lifespan. ^[14] In practice the offense has proved hard to prosecute, since it requires proving intent; the Apple case above was settled under a different charge.

The European Union has taken a broader approach, aimed less at punishing intent and more at making durability and repair the default. Since 20 June 2025, smartphones and tablets sold in the EU must meet ecodesign requirements: batteries that retain at least 80% of their capacity after 800 charge cycles, critical spare parts available for seven years after a model stops being sold, and operating system updates for at least five years after the last unit is placed on the market. ^[16] From 18 February 2027, the EU Batteries Regulation requires portable batteries in most devices to be removable and replaceable by the end user with commercially available tools. ^[17] And the Right to Repair Directive, which member states had to apply by 31 July 2026, obliges manufacturers of certain products (including washing machines, dishwashers, refrigerators, vacuum cleaners and smartphones) to offer repair at a reasonable price even after the legal guarantee has expired, and extends the legal guarantee by a year when a consumer chooses repair over replacement. ^[15]

In the United States, repair legislation has advanced state by state. California's Right to Repair Act, in force since July 2024, requires manufacturers of electronics and appliances to provide parts, tools and documentation to owners and independent repair shops: for three years after production for products with a wholesale price between $50 and$99.99, and for seven years for products at $100 or more. ^[18] Other states, including New York, Minnesota and Oregon, have passed their own laws with different scopes.

None of these measures forbids a product from breaking. What they change is the arithmetic. If spare parts must exist, repair becomes possible. If batteries must be replaceable, a worn battery stops being a death sentence. If software must be supported for years, a working device is not abandoned by its maker. If the manufacturer must offer repair after the guarantee, a product that fails early becomes a cost to the company as well as the customer. Durability starts to pay, or at least failure stops paying so well.

Whether these rules will work as intended is not yet known. They are new, enforcement varies, and manufacturers adapt. What is clear is that lawmakers have stopped treating product lifespan as a purely private matter between a buyer and a seller.

No Villain Required

It would be comforting if the story needed villains. It does not.

A rational engineer, given a cost target and a deadline, chooses a cheaper capacitor, glues a panel because clips would add a millimeter, and seals a battery because the product must survive a drop into water. Each choice can be defended.

A rational manager, answerable to revenue targets, notices that repair operations lose money and new units make it, and decides that spare parts beyond a few years are not worth warehousing. That can be defended too.

A rational consumer, faced with a repair quote at 60% of the price of a new model with a better camera and a longer warranty, buys the new one. Nobody would blame them.

Each decision is reasonable where it is made. Added together, across billions of products and a few decades, they produce mountains of discarded circuit boards, drawers of dead chargers, and mines dug to replace things that did not need replacing. The system is not irrational because its participants are foolish. It becomes irrational because each participant is responding sensibly to incentives that were never designed with the whole cycle in view.

This is the real subject of the Phoebus story. The cartel is unusual only because it wrote the incentive down. Most of the time, no one has to decide that products should fail. It is enough that failure is rewarded and durability is not.

The question the story leaves is not whether companies are good or bad. It is this: what happens when an economic system can reward a product for failing before its technology requires it to?

The Desk

Go back to the engineer.

They spent years learning to do more with less. To use less material without losing strength. To waste less energy as heat. To find the weak point in a structure and eliminate it. To raise the mean time between failures. To make machines that are quieter, lighter, cleaner, more reliable. That is what the discipline is for, and in a sense it is what every tool in their training was built to do.

Now a product sits on their desk. It works. It is well made. The bearings are sound, the circuit board is clean, the motor turns without complaint, and with a new battery and a spare part now and then it could probably keep turning for decades.

Someone leans into the doorway and asks, not unkindly, the question the whole business depends on:

"How do we make them buy another one?"

The engineer looks at the machine for a while. They know exactly how to answer. That was never the difficult part.

References

[1] Krajewski, M. (2014). "The Great Lightbulb Conspiracy." IEEE Spectrum, 24 September 2014.
https://spectrum.ieee.org/the-great-lightbulb-conspiracy

[2] Krajewski, M. (2014). "Fehler-Planungen. Zur Geschichte und Theorie der industriellen Obsoleszenz." Technikgeschichte, 81(1), 91–114.
https://doi.org/10.5771/0040-117X-2014-1-91

[3] United States District Court, District of New Jersey (1949). United States v. General Electric Co., 82 F. Supp. 753.
https://law.justia.com/cases/federal/district-courts/FSupp/82/753/1755675/

[4] Monopolies and Restrictive Practices Commission (1951). Report on the Supply of Electric Lamps. HC 287. London: HMSO.
https://www.gov.uk/government/publications/report-on-the-supply-of-electric-lamps

[5] Livermore's Centennial Light Bulb Committee (n.d.). "Facts." Livermore's Centennial Light Bulb.
https://www.centennialbulb.org/facts.htm

[6] Kykta, M. (2022). "Incandescent lamp design and lifetime." AIP Advances, 12(10), 105116.
https://doi.org/10.1063/5.0101992

[7] Dunn, J. (2016). "Incandescent lamps and service life." EDN, 10 October 2016.
https://www.edn.com/incandescent-lamps-and-service-life/

[8] London, B. (1932). Ending the Depression Through Planned Obsolescence. New York. Project Gutenberg eBook #72003.
https://www.gutenberg.org/ebooks/72003

[9] German Environment Agency (Umweltbundesamt) (2016). "Lifetime of electrical appliances becoming shorter and shorter." Press release, 15 February 2016, on the study by Öko-Institut e.V. and the University of Bonn.
https://www.umweltbundesamt.de/en/press/pressinformation/lifetime-of-electrical-appliances-becoming-shorter

[10] Baldé, C. P., Kuehr, R., et al. (2024). The Global E-waste Monitor 2024. International Telecommunication Union (ITU) and United Nations Institute for Training and Research (UNITAR).
https://ewastemonitor.info/the-global-e-waste-monitor-2024/

[11] Federal Trade Commission (2021). Nixing the Fix: An FTC Report to Congress on Repair Restrictions.
https://www.ftc.gov/reports/nixing-fix-ftc-report-congress-repair-restrictions

[12] CBC News (2016). "Nest's move to stop supporting Revolv smart hub leaves customers with costly 'brick'."
https://www.cbc.ca/news/science/revolv-bricked-1.3521927

[13] Library of Congress, Global Legal Monitor (2020). "France: Watchdog Agency Fines Apple for Deceitful Practice." 28 February 2020.
https://www.loc.gov/item/global-legal-monitor/2020-02-28/france-watchdog-agency-fines-apple-for-deceitful-practice/

[14] République française (2015, amended 2021). Code de la consommation, Article L441-2. Légifrance.
https://www.legifrance.gouv.fr/codes/article_lc/LEGIARTI000044330817

[15] European Parliament and Council (2024). Directive (EU) 2024/1799 on common rules promoting the repair of goods. Official Journal of the European Union.
https://eur-lex.europa.eu/eli/dir/2024/1799/oj

[16] European Commission (2023). Commission Regulation (EU) 2023/1670 laying down ecodesign requirements for smartphones, mobile phones other than smartphones, cordless phones and slate tablets. Official Journal of the European Union.
https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=celex%3A32023R1670

[17] European Parliament and Council (2023). Regulation (EU) 2023/1542 concerning batteries and waste batteries, Article 11. Official Journal of the European Union.
https://eur-lex.europa.eu/eli/reg/2023/1542/oj

[18] California Legislature (2023). SB-244, Right to Repair Act.
https://leginfo.legislature.ca.gov/faces/billTextClient.xhtml?bill_id=202320240SB244