A friend of mine tested fatigue failure of carbon composite materials for aircraft design and came up with 400,000 flexes before failure, and that number seemed similar to what I've seen for bicycle carbon fiber. Fatigue limit for aluminum is on the order of ten million cyclic stresses if I recall correctly.

That's all more vague than I'm comfortable with, and it's not really a hard limit of stresses before failure but more like a probability of failure that increases as the number of cyclic stresses increase. Does anyone know of more specific and reliable numbers for fatigue failure, comparing carbon fiber, aluminum and steel? You won't hit the limit with aluminum with the stresses from normal riding but the 400,000 for carbon fiber seems approachable given enough time.