What a 100,000 Hour LED Rating Actually Promises

This page contains affiliate links. If you buy through one we may earn a commission, at no cost to you.

Because an LED almost never burns out, so "lifetime" is not a thing that happens to it. It fades, continuously, from the first hour, and never stops. That means someone has to choose a fraction of original output and call that the end, and the quoted number is entirely a consequence of which fraction they chose. The same lamp on the same decay curve is good for 29,500 hours or 100,000 or 194,300 depending only on where the line is drawn, a spread of 6.6 times.

Which makes a lifetime figure with no grade attached unfalsifiable. We carry exactly one product in 1,412 that publishes an LED life in hours. It says 100,000 hours, it does not say to what output level, and on the same page it publishes a 36 month warranty. At eight hours a day, 36 months is 8,760 hours of use, so the claim is 11.4 times what the maker will stand behind.

What the number means when it is stated properly

An LED's output declines because of degradation in the phosphor, in the die attach, in the encapsulant, and in the package materials that yellow and scatter. None of it stops the device working. It just produces less light for the same current, which is lumen depreciation.

So life is quoted as an L-figure: the number of hours to reach a stated percentage of initial output. L70 is the time to fall to 70 percent, which is the conventional figure for general lighting. L90 is the time to fall to 90 percent, used where colour and level matter, such as retail and studio work. L50 is the time to half, used for decorative and signalling applications where nobody is reading by it.

Because the decay is close to exponential, converting between them is straightforward. If output follows a decaying exponential, then the ratio of two L-times is the ratio of the natural logarithms of the two output fractions. So L90 divided by L70 is the log of 0.90 over the log of 0.70, which is minus 0.1054 over minus 0.3567, which is 0.295. And L50 divided by L70 is the log of 0.50 over the log of 0.70, which is 1.943.

Run those on a 100,000 hour L70 claim. L90 is 0.295 times 100,000, which is 29,500 hours. L50 is 1.943 times 100,000, which is 194,300 hours. Every one of those three is a defensible answer to "how long does this lamp last", they describe the same physical object, and they differ by a factor of 6.6 from end to end. A figure quoted without its grade has had the only information that made it meaningful removed, in exactly the way a lumen figure has had the beam angle removed.

Why 70 percent, and what it admits

The choice of 70 is not arbitrary and it is worth knowing what it encodes, because it is more honest than it first appears and less reassuring.

Perceived brightness does not track luminance linearly. It follows roughly a power law with an exponent near one third, so apparent brightness goes as something close to the cube root of the light actually arriving. Take 0.70 to the power of one third and you get 0.888. A 30 percent reduction in light output therefore reads as about an 11 percent reduction in apparent brightness.

Now add the second thing: it happens over years, with no reference. You never see the new lamp and the old lamp side by side, and your visual system adapts continuously. An 11 percent change delivered over a decade is invisible.

So the L70 convention is a statement about perception rather than about function. It does not mean the lamp is still good at 100,000 hours. It means that at 100,000 hours it will be noticeably dimmer than new to an instrument and not to you. Which makes a published L70 figure a promise that you will not notice the degradation, dressed as a promise that there will not be any.

Where the number comes from, and why it is six times longer than anyone watched

There is a second problem, and it is that these figures are projections rather than measurements.

The measurement standard is LM-80, which specifies how to measure lumen maintenance of LED packages: run them at controlled case temperatures and drive currents and measure output periodically, for at least 6,000 hours and preferably 10,000. Nobody runs a lamp for 100,000 hours to see, because 100,000 hours is 11.4 years of continuous operation and the product would be obsolete several times over before the test finished.

The projection standard is TM-21, which takes LM-80 data and extrapolates it. And TM-21 contains a deliberate guard rail: the reported projection is limited to six times the total test duration. The standards body put that cap there precisely because extrapolating an exponential fit far past your data is how you get numbers that mean nothing.

So work backwards. A 50,000 hour claim requires 8,333 hours of testing, which is eleven and a half months. A 100,000 hour claim requires 16,667 hours, which is one year and eleven months of continuous measurement, on a statistically adequate population, before the number can be reported at all.

That arithmetic is the useful part, because it tells you what a 100,000 hour claim on an inexpensive fixture almost certainly is. It is not a TM-21 projection of that fixture. It is the LED chip maker's datasheet figure for the bare emitter at a stated junction temperature and drive current, lifted onto the finished product. Which is a specification for a component, under conditions the finished product very likely does not provide, and it says nothing at all about the lamp.

The one lamp we carry that publishes a figure

A LitONES LED desk lamp with swing arm is that product, and the two figures sitting next to each other in its record are the whole argument. Its pros say "84 LED beads rated for 100,000 hours with a stated 36-month warranty". Its specifications say "LEDs: 84 beads, rated 100,000 hrs" and "Warranty: Stated 36 months".

Put them in the same units. One hundred thousand hours at eight hours a day is 12,500 days, which is 34.2 years. At twelve hours a day it is 22.8 years. Run continuously it is 11.4 years. The warranty is 36 months, and at eight hours a day that is 8,760 hours of actual use, which is 8.8 percent of the claimed life. So the lamp is sold on 34 years and underwritten for three.

Notice also what the phrase "84 LED beads rated for 100,000 hours" is actually saying. It rates the beads, not the lamp. That is the component-specification transfer described above, stated almost explicitly, and to the listing's credit it is more candid than most: it tells you the figure belongs to the emitters. The lamp contains a driver, a dimming circuit, a colour temperature control, a memory function, a timer and a multi-joint arm with internal wiring, and none of those is covered by a bead rating.

Its record is properly careful elsewhere, noting that "Stated figures are the maker's specifications, not independently measured". That caveat is the right one and it applies to the 100,000 more than to anything else on the page. Details in our LitONES LED desk lamp review. For the colour side of that same lamp's specification, which is a separate argument with its own arithmetic, see why a CRI 95 light still renders red badly.

The component that will actually fail

Here is the part the Quora question is really getting at, and it is the reason real-world LED failures cluster at a few thousand hours rather than at a hundred thousand.

An LED runs on direct current at a low, tightly controlled level. Mains supply is alternating current at a high voltage. Something has to sit between them, and that driver is an electronic circuit with electrolytic capacitors in it, because smoothing a rectified supply needs capacitance and electrolytics are the only economic way to get enough of it in a small space.

Electrolytic capacitors dry out. The electrolyte diffuses out through the seal, capacitance falls and series resistance rises until the circuit stops regulating. The process is chemical and therefore follows an Arrhenius relationship, which in this field is conventionally handled with a doubling rule: life doubles for every 10 degrees Celsius the capacitor runs below its rated temperature.

Work a realistic case. A common part is rated for 2,000 hours at 105 degrees Celsius. Inside a lamp enclosure near the heat of the emitters, 65 degrees is a reasonable operating temperature. That is 40 degrees of margin, so four doublings, which is 2 to the fourth power, or 16. Sixteen times 2,000 is 32,000 hours.

And the sensitivity is brutal, which is the real lesson. Run the same capacitor 10 degrees hotter, at 75, and you get 16,000 hours. Ten degrees hotter again and you get 8,000. A single design decision about ventilation, or a single installation decision about putting a lamp in an enclosed fitting, halves the life of the part that determines the life of the product.

So the honest ranking for that desk lamp is: emitters claimed at 100,000 hours, the same emitters at L90 around 29,500, a plausible driver life somewhere between 8,000 and 32,000 depending entirely on a temperature nobody publishes, and a warranty of 8,760. The component most likely to end the product's life is the one with no specification anywhere on the listing, and that is true of every LED product we carry and, as far as the listings show, of the category.

Binning, and why a replacement rarely matches

The second thing a lifetime figure cannot tell you is whether two units of the same model look the same, and that is a manufacturing matter rather than a degradation one.

LED dies come off a wafer with a real spread in brightness, forward voltage and colour, because the process cannot be controlled tightly enough to make them identical. So they are measured after manufacture and sorted into bins. How narrow those bins are is a commercial decision: tight binning means discarding or down-selling more of the wafer.

For white light the relevant scale is the MacAdam ellipse, a contour of equal perceptual colour difference. A one-step ellipse is a just-noticeable difference under ideal comparison; three steps is roughly where about half of observers see a difference in normal viewing. Premium fixtures specify two or three step binning. Consumer product frequently specifies only the ANSI colour quadrangle for the nominal correlated colour temperature, which is substantially larger than a three step ellipse. Two fixtures of the same model, same nominal colour temperature, can therefore differ visibly straight out of the box.

A string of 48FT color-changing outdoor string lights with 15 LED bulbs is where this would bite hardest in our catalogue, and it is also the product that sidesteps it. Its record publishes 15 E26 colour-changing bulbs, 16 colours and 8 modes, a stated 1200 lumens, IP65, end-to-end linkable strings, and a temperature range of minus 22 to 140 degrees Fahrenheit. Fifteen emitters hung in a line, all in view at once, is the worst possible arrangement for colour mismatch, because side by side comparison is exactly the condition under which small differences become visible.

What saves it is that these are addressable colour-changing bulbs rather than fixed white ones. A unit that synthesises its colour from separate red, green and blue emitters is set by its controller rather than by its phosphor bin, so all fifteen receive the same instruction and land in the same place. The honest flip side is that RGB mixing has its own colour accuracy problems, which are a different subject, and that the 140 degree Fahrenheit upper limit is 60 degrees Celsius of ambient, which on the capacitor arithmetic above is a very hot place for a driver to live. Its own cons field says plainly that "Durability figures are maker-stated only". Our 48FT color-changing outdoor string lights review has the rest. And for what a bi-colour design costs you in exchange for the adjustment, see what a bi-color LED costs you for the dial, which works through that trade with its own figures.

Two cases where the LED is not the limit at all

It is worth finishing on the two products where the lifetime question has a different answer entirely, because they show that the published figure is often not even the right thing to ask about.

A set of AGPTEK solar driveway lights publishes six LEDs, three colours across five modes, a five hour full sun charge, a stated 30 hour runtime, an IP67 sealed aluminium alloy body and a rating to two tons of pressure. On everything above, the emitters in that product will outlive the owner. What will not is the battery, because a lithium cell in an outdoor light is cycled once every single day and loses capacity by cycle count and by calendar ageing, with both accelerated by the temperature swings a driveway provides. A few hundred to a thousand full cycles is the realistic range for a small cell, which is one to three years of daily use, and the stated 30 hour runtime is a new-battery figure that will fall steadily.

One genuinely good thing about that design deserves pointing out, because it is the opposite of the enclosure problem described earlier. A sealed aluminium alloy body has no air cooling but excellent conduction and a lot of thermal mass, and it is in contact with the ground. For a six-LED accent light at low power, that is a better thermal path than a vented plastic housing, which helps both the emitters and whatever small driver circuit is in there. Its record is honest that "Runtime and 2 ton rating are maker-stated" and that "Brightness in lumens not listed". See our AGPTEK solar driveway lights review.

And a GreenSun RGB LED aquarium light is the thermal best case in the whole catalogue, publishing an IP68 submersible rating, 5050 RGB beads, 16 colours, a 24-key remote and a transparent acrylic body on suction cups. An emitter running underwater is being cooled by a fluid with roughly four thousand times the volumetric heat capacity of air, so its junction temperature sits far lower than the same emitter in a room would, and junction temperature is the single biggest lever on lumen depreciation. On the degradation physics, submersion is a large and unadvertised advantage.

What replaces it as the limiting factor is water ingress, which is a seal problem rather than a semiconductor one, and the honest point is that IP68 is a rating for continuous immersion under stated conditions rather than a guarantee against a seal degrading over years in warm water. Its record is appropriately restrained, noting that it is "Not a specialist plant-growth light", that "No PAR figures stated by maker" and that "Colour effects suit display over function". Our GreenSun RGB LED aquarium light review covers it.

What our own records say, and the gap

This is worth stating plainly because it was the surprise of writing this page.

Across 1,412 published products, zero publish an L70, L80 or L90 figure. Zero publish the phrase lumen maintenance. Zero publish a 50,000 hour claim. Zero publish a driver life, a capacitor rating, a junction temperature, a case temperature, or a MacAdam step. Exactly one publishes an LED life in hours, and it attaches no L-grade.

Across 2,527 published pages, lumen maintenance, LM-80, TM-21 and lumen depreciation appear zero times. The two occurrences of the word binning on this site are both the British verb for throwing something away. So this is not a case of our listings quoting a standard badly. The standard does not appear in our catalogue at all, and the one lifetime claim that does appear is a component rating.

That is a genuine finding about the category rather than a failing of any one product. A 100,000 hour bead rating is probably accurate as a bead rating. It is being asked to carry a question about a lamp, and nothing else on the listing can help it.

The honest limit, and what to do instead

There is no specification available to you that predicts how long an LED product will last. The figure that would is the driver's expected life at the operating temperature in that enclosure, and it is not published by anyone, for anything, as far as our listings show. So the arithmetic on this page is computed from published hours and warranty figures plus standard capacitor and perception relationships, and the capacitor numbers in particular rest on an assumed operating temperature that could reasonably be twenty degrees either side, which moves the answer by a factor of four.

Which leaves one specification worth reading and several free habits worth more than any of it.

The specification is the warranty, and it is the only number on the page that costs the maker something to be wrong about. A lifetime claim is a projection from someone else's datasheet; a warranty is a commercial commitment with a cost attached. Where the two disagree by a factor of eleven, believe the smaller one. For the desk lamp above, read 36 months as the real statement and 100,000 hours as marketing, and note that the stated 36 months is itself longer than much of the category offers.

Then the habits, all of which act on junction and driver temperature, which is the variable that actually decides the outcome. Do not put a lamp in a fully enclosed fitting unless the maker says it is rated for one, because trapping the heat is the single fastest way to halve its life twice over. Leave air around a fixture rather than boxing it in behind a shade. Prefer a product with metal in its housing over one that is entirely plastic, since the metal is a thermal path and the plastic is an insulator. And where a product offers dimming, use it: running an emitter below full current drops its junction temperature, and because the capacitor rule doubles per ten degrees, a modest reduction in output can buy a large multiple in life.

And the last one, which is the free technique that beats the purchase. If you are buying several fixtures that will be seen together, buy them at the same time and keep them together, because that is the only thing you can do about binning. Units bought in one batch are very likely from one production run and one bin. Replacing a single unit from a set two years later is the case where mismatch is most visible and least fixable, and the reason it happens is not that the old ones faded. It is that the new one came out of a different box.