The selling point of an LED bulb is that it runs cool, and compared with a filament lamp that is true: far more of the electricity leaves as light and far less as infrared radiation. But "cool" is not "no heat". A light-emitting diode is a semiconductor junction, and the heat it does produce is generated in a very small volume with no convenient way out. It has to be conducted, layer by layer, from the chip to the air in the room. Trace that path and you understand almost everything about why one bulb lasts for years and another dims in a season.
It also explains a fact that surprises people who take a dead bulb apart. The little luminous squares are usually fine. What has failed is the driver: the compact circuit in the neck that turns mains voltage into the low, steady current the diode needs. The chip is the visible part of the product; the driver is the part with a hard life.
Heat travels out, not away with the light
A filament lamp radiates its heat away as infrared through the glass. An LED cannot do that. The heat leaves by conduction: chip, solder joint, metal-core printed board, a flat interface, an aluminium or thermally conductive plastic body, and finally the surrounding air by convection. Every interface in that chain is a chance for a gap, an air pocket, or a badly applied compound, and each one adds thermal resistance. A design can have an excellent chip and still run hot because of a tenth of a millimetre of poor contact two steps down the chain.
This is why the body of a bulb is not merely a housing. The fins and the slight taper on the lower half are a heat sink with a shape constraint: it must also fit a lamp socket and not block the light. Thermal design here is the art of conducting heat downward and outward through a part that is supposed to look like a light bulb.
The driver sits in the worst place
Mains voltage has to be converted to a controlled direct current, and that conversion needs components — including capacitors — that are far more temperature-sensitive than the diode. Those components are packed into the neck of the bulb, immediately above the screw cap and immediately below the heat sink. It is the hottest, tightest real estate in the whole object.
There is a well-known rule of thumb in electronics: for electrolytic capacitors, every sustained rise of about ten degrees Celsius roughly halves the expected life. The exact number depends on the part and the conditions, but the direction is reliable. So a small improvement in how quickly heat leaves the base can have a much larger effect on service life than a similar improvement in the chip itself. The bulb fails at its weakest thermal link, which is usually the component that was never mentioned on the box.
Dimming is a symptom, not a mode
As a diode runs hot over time, its light output gradually falls and its colour can drift. People notice this as a bulb that "isn't as bright as it used to be". That slow decline is a different phenomenon from a bulb that flickers or refuses to start, which tends to point back to the driver or to a dimmer circuit that was never designed to work together.
Fixtures matter as much as bulbs. A sealed decorative shade traps the air the heat sink was counting on; a recessed downlight can bake the neck; a socket pointed upward puts the driver at the top where the warm air collects. A bulb rated for open air may have a very different life inside a closed glass globe. The rating describes a test condition, not every cupboard the bulb will end up in.
Why it still looks like an old lamp
The familiar pear shape and screw base are inherited from an object designed around a glowing wire in a vacuum. That legacy fixes the diameter of the neck, the position of the electrical contact, and the volume left for the driver and heat sink. Modern electronics have to be fitted into a silhouette chosen more than a century ago for a completely different thermal problem. In the lighting districts around Zhongshan's Guzhen, a great deal of design work is essentially negotiation with that constraint.
The diffuser is part of the same inheritance. The chips are tiny, intensely bright points, and nobody wants a room lit by visible dots. A milky polycarbonate or glass dome spreads that light with scattering particles, which costs a little output and adds one more plastic part to the thermal story. The white colour you see when the bulb is off is an optical component, not paint.
Colour temperature is also a surface decision. A warm-white and a cool-white bulb often use essentially the same emitter, with a different phosphor formulation converting some of the blue light into longer wavelengths. So "warm" is not a different kind of light source; it is a different coating on top of one. This is why two bulbs with the same printed specification can still look slightly different side by side — the coating is a manufactured chemical layer with its own tolerances.
How to read a bulb in a shop
Ignore the wattage first. Watts describe what goes in, not what comes out; lumens describe the light. Then look at the lower half of the body: is there real metal, or a thin shell that flexes under a thumb? Check whether the diffuser is a separate, evenly fitted piece, and whether any vent or seam is placed where a closed shade would cover it. If the box claims a very long life, notice that the claim is about the whole assembly — and the whole assembly is only as good as the small circuit hidden in the neck.
An LED bulb is a new device wearing an old shape, and its failure mode is thermal even though it feels cool to the touch. The light is the easy part. Moving a few watts of heat out through a screw base is the hard one.