In general, ball bearings are applicable in higher speeds because of the smaller contact area and smaller friction force, but roller bearings can carry greater loads because the load applied to the contact area produces smaller contact stresses.
-- Principles of Mechanical Engineering Design, Liubomir Dimitrov

added: The premises of the question are quite wrong. Be sure to read through the answers before forming any opinions.

In bicycle applications there are neither high speeds nor high loads (compared to cars or industrial machines). However, I would assume that the "smaller contact stresses" would result in significantly longer service life of the bearing, even if water and contaminants penetrate the bearing. And still the only time I have heard of bicycle roller(pin) bearings is in the hyped Lefty.

Why are ball bearings used exlusively in bikes e.g. in the headset where some minor friction would not be a problem?

  • 6
    Ball bearings are much simpler to construct and maintain. They're also lighter and more compact. Sep 29, 2015 at 12:21
  • 2
    "It's all ball bearings nowadays" - Fletch
    – user25511
    Apr 15, 2016 at 17:54

7 Answers 7


I think several of your assumptions are incorrect:

  1. no high speeds: rim dynamos spin quite fast, up to 10,000rpm, and that does require effort on the part of the designers. Typically the drive wheel on those will have a radius of about 1cm, or a circumference of 3 1/3rd cm and 30 revolutions per metre. At 10m/s (30kph) that means 300rps or 18000 rpm. This is why they make such a high pitched whining noise when in use.
  2. no high torque: a 100kg cyclist standing on a 165mm crank is exerting about 1kN at 0.165m, or 165Nm. That's comparable torque to a mid-range car. I repeatedly snapped a 8mm high tensile bolt in the drive system of my four wheel load bike, and replacing that with industrial freewheels meant buying two of the highest rated units I could find (German conveyer belt parts rated to 250Nm)
  3. No high loads: a touring cyclist with 50kg on the rear rack, plus half their 100kg body weight is putting a load of 100kg or 1kN on a 10mm diameter hollow rear axle. Those axles and the dropouts they bear on must be made of hard steel to stand the load, where the load bearing parts of a car or truck are typically made of much softer steel.
  4. Only ball bearings are used: modern sealed bottom brackets commonly use roller or needle bearings, as do many headsets. Internally geared hubs have used needle bearings for many years. Sheldon Brown describes roller bearings in his glossary, for example.

The answer is that various types of bearings have been used in bicycles and will likely continue to be. Again, Sheldon discusses needle bearings in Shimano hub gears from some years ago and they're still mentioned in the marketing.

One problem not unique to bicycles but especially relevant to them, is that more precise bearings are more vulnerable to grit and also to damage. So especially needle bearings need to be carefully protected from dust, sand, and forces outside their design range. Unfortunately doing that inside a suspension unit is hard, and last time I looked Cannondale had not solved the problem (my HeadShock is unmaintainable after only about 10Mm, I've seen multiple lefty shocks fail during a race because of light rain and fine grit causing them to abrade to uselessness very quickly)

But inside a hub or headset, for example, they can be amazing. I've talked to people who've ridden 20Mm on an Alfine 8, for example, with no problems.


Firstly I disagree with the comment about high speed. The highest velocity bearings are plain: crankshaft bearings in a full size car engine are plain, whereas in a radio controlled model, ball bearings are used. Where velocity (rpm x diameter) is high the oil wedge effect tends to centre the shaft (hydrodynamic lubrication). This can be further enhanced if the oil is pumped.

Straight roller bearings are great at supporting radial loads, but useless at supporting axial loads. Where there is no axial load (for example a generator driven by an engine, either directly through a sliding spline or via belt drive) they are a good solution. In most bicycle applications, axial forces are present, therefore a total of 3 bearings must be considered: roller, plus axial thrust in each of two directions. (in some applications you might find a plain bearing for small axial thrust, but I can't think of one on a bike.)

Ball bearings are much simpler to make, and two ball bearings forms a great all-purpose joint. On a bike wheel, the two bearings are at the ends of the axle, as far apart as possible, which is exactly where we need them to resist non-axial torque applied to the wheel. In this respect, the central part of a roller bearing would be doing nothing, so you might want two roller bearings, one at each end, plus some additional bearings for the axial thrust (getting complicated isn't it?) The better solution is to have a single ball bearing at each end of the axle, with a conical form, which enables it to resist both radial and axial forces. Tapered roller bearings do exist, but they need a cage to help guide the rollers, which avoids the need for four bearings on a wheel, but adds other complexity.

Finally, headsets definitely do need some kind of antifriction bearing. There is an enormous nonaxial torque: your weight is trying to push the forks up and forward, except when you apply the brakes, when they will be pushed back. To be able to steer properly, you need a smooth movement there. Anyone who has tried to ride a bike with a damaged headset will know that it is possible, but uncomfortable and dangerous.

  • WRT "you might find a plain bearing for small axial thrust" I have seen one made by drilling a hole in the end of the shaft and inserting a ball bearing used in sidewall dynamos. I suspect they're not much needed, but since the end cap/thrust bearing was also the seal, removing it would have likely destroyed the dynamo.
    – Móż
    Nov 15, 2015 at 3:08

The current answers give lots of technical reasons for the use of ball bearings while noting that other types of bearings are sometimes use on bike.

However they miss the most important factor. Ball bearings are cheap!


You will see sealed cartridge bearings in many bottom brackets, which use roller bearings. Also common in head sets. Ball bearings were more widley used up to around 10 years ago. But still common in pedals and hubs.

  • That's interesting. I never opened a sealed BB. Why BBs but not the much slower rotated headsets? Why only in the past?
    – Vorac
    Sep 29, 2015 at 10:12
  • 1
    Some headsets do use roller bearings, as said in the answer. As to the reasons why: probably becasue of tradition, possibly economical reasons, ...
    – stijn
    Sep 29, 2015 at 10:17
  • 5
    @Vorac - Conventional roller bearings do not take thrust very well. Sep 29, 2015 at 22:41

I think you are grossly mistaken in the fact that bicycles do not have heavy loads.

Consider me on my bike. I have measured that I can pull up with the rear foot 25 kg using clipless pedals. I have also measured that I can pull up from the handlebar at a force of 30 kg. I weigh 110 kg. If I pedal uphill on a high gear accelerating hard, my force on rear pedal is 25 kg, and my force at front pedal is 110+25+30 kg = 165 kg. That's a total of 190 kg or 1864 N at a crank arm of 175 mm, which is a torque of 326 Nm.

That's more torque than in my 2.5 liter car engine which produces only about 210 Nm! Now of course, the car engine being a 4-cylinder isn't completely smooth so maybe it has higher peaks than 210 Nm around a full rotation, but still I think my bike wins on torque.

What bicycles are: very very high torque/force, very very low speed devices. Like steam locomotives.

Bikes also carry lot of load for the weight. A car of 1500 kg maybe carries 500 kg load. That's third of its weight. A bike of 13 kg carries a load of 130 kg. That's ten times its weight!

Also bikes, having no suspension, have to absorb huge road imperfection impacts, while carrying ten times their weight as load, whereas a car carrying third of its weight as load, has very good suspension that reduces peak forces caused by road imperfections.

Electric motors and internal combustion engines achieve what they can do, not by huge torque (their torque is usually insignificant compared to the torque of bicycles), but by huge speed. Power is torque times speed. Maybe some low-speed two-stroke marine diesel engines used in large container ships are the exception to the rule that internal combustion engines are high-speed low-torque devices.

As for why roller bearings aren't used, many bike bearings need to handle side loads. Thus, they are angular contact bearings. Consider what happens to an angular contact roller bearing: the contact surface is conical, so its radius is larger at one end than at the other. At the high radius side, the rollers would have to roll a larger distance than at the low radius side. This is impossible without skating. So the roller bearing actually would function mostly as a plain bearing! It would require very good lubrication at all times, and a failure of lubrication would immediately destroy the bearing. Bike bearings aren't lubricated as well as car engines where plain bearings function well due to excellent lubrication. Furthermore, the roller skating would dissipate energy as heat, which is unacceptable on bikes where all energy loss mechanisms need to be minimized because there's so little human power available.

Actually, roller bearings where tried on bikes when headset ball bearings would often develop a condition called "indexed steering": https://www.sheldonbrown.com/brandt/indexed-steering.html

Roller bearings of various designs have been tried, and it appears that they were possibly the ones that finally made obvious that fore and aft motion was the culprit all along; a motion that roller bearings were less capable of absorbing than balls. This recognition lead to using spherical alignment seats under the rollers. Although this stopped dimpling, these bearings worked poorly because the needle complement tended to shift off center, skewing the needles and causing large bearing friction as the rollers skated.

Roller bearings didn't help there. Later, it was recognised that a cartridge bearing system where the cartridge sits on a 45 degree conical surface of the bearing cup helped. In this case, actually, the headset bearing is a compound bearing: a ball bearing takes rotation motion, while the 45 degree conical surface functions as a plain bearing, taking the impacts that would destroy earlier rigid headset designs with the "indexed steering" condition. This compound bearing works well as long as the plain bearing remains lubricated. Today, nearly all headsets are these cartridge bearing designs, that have the compound bearing system, and indexed steering is a thing of the past.

So you are also incorrect in that only ball bearings are used on bikes. Actually, the modern headset is a compound bearing: a ball bearing for rotation, and a plain bearing for taking fore-to-aft impacts.

(Note the sheldonbrown.com link mentions spherical seat -- today, the seat is usually conical, and it functions as well as a spherical seat)

  • 1
    Two remarks: 1) The weight of the vehicle in relation to the transported weight is irrelevant. The important figure is how much total weight you've got. And how much of that weight is suspended. The body of the cyclist counts as suspended, the weight on the luggage rack does not. Add in the fact that bikes may use over 5 bar in tires while cars are usually below 3 bar, and you realize how easy life is for a car's bearings... 2) If you angle a roller bearing, the rollers should of course be of conical shape as well as the cups. No skating required. Dec 5, 2022 at 16:56

Roller bearings and ball bearings, are efficient. If you take plain bearings for the wheels, e.g. PTFE-bushings, the bicycle will work, but becomes a bit heavy to pedal, not so fun... Plastic bearings also makes the bicycle lesser stiff, which can feel uncomfortable, it can easily get into self-resonance and start wobbling if too low egen-frequency, but that part depends a lot of the detail design. You can use metal bronze bearings, but they need some lubrication.

For the headset ball bearings are also normally used. I just replaced the trust ball bearings for the headset on my old bicycle with home-made delrin (Acetal plastic) bearings I made myself on the lathe, because the old one were rusty and I found new one a bit expensive... Result: Yes, delrin plain bearings works fine for the headset, it will be heavier to turn it if you wish same rigidly as with ball bearings, but not a problem... It can be a bit of stick-slip with plain bearings in the head-set, this is probably the main reason why ball bearings are usually used - it is stiff and it turns smooth - and you will feel control.


As usual, the answer is almost certainly money. Ball bearings are simple, manufactured in gazillions and cheap!

Bike manufacturers don't make bikes to be the best they can, they make them to make money. Anything which is wearable makes more money and the faster it wears, they make even more money.

The sub answer is that most bikes probably never get used enough to need anything replacing, so cheap ball bearings are good enough. For the bikes that do, the spare parts cost are dirt cheap and the profit can be huge.

Tapered roller bearings would make more sense for headset bearings to stop or reduce brummeling. Given that they have been used in car wheel bearings for decades and they can last hundreds of thousands of miles, they would probably last forever as bike headset bearings, with no signs of brummeling. A different taper angle would be needed than car wheel bearings, as most of the load is vertical and side loads are less.

Tapered roller bearings could also make much more sense for the bottom bracket bearings although the forces are not generally axial, the differential radial forces can be big, particularly when cleats or clips are used. I've lost count of the number of bottom bracket bearings I've replaced on my 50kg semi recumbent cargo trike. I often forget to change down at junctions, then pull away in a higher gear, so put massive force on the crank arms (I think my crank might actually be bent now!) and stretch the chains!

  • 2
    That is not really how chains "stretch". The second paragraph also sounds a bit too much like proof by conspiracy to be credible.
    – ojs
    Dec 4, 2022 at 13:47
  • @ojs Not always. Just search for how the litigious bastards who used to run Speedplay before they sold out went after anyone who tried to sell the bog-standard bearings used in Speedplay pedals. Dec 4, 2022 at 13:58

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.