Your logic would indicate that "All MX's should fail before 20 years", which isn't the same as Mean Time Before Failure. Assumedly MTBF would be calculated from a certain number of MX's and tracking when they fail. Some may be as low as 1 year, some may be as high as 40 (for example). That spread would then be averaged to 20 years. In the real world, this is probably extrapolated data rather than real lab devices sitting for 20+ years but the theory is still the same. The other thing to note is that devices don't have a linear failure rate. They usually adhere to some sort of non-linear graph that typically follows a high rate of immediate failures (factory defects), then a very low rate of failures for a certain period of time that eventually begins to spike at the MTBF. ... View more