I'll answer more generally on the subject of lifespan rather than fatigue life, which Nathan has argued is a concept derived from metallurgy that doesn't operate the same way in carbon fiber.
First, remember that carbon fiber composites are made from cut-up bits of carbon fiber sheet (usually pre-impregnated with resin, aka pre-preg carbon). Those are laid up onto a mandrel. The number, shape, orientation, and position of the sheets can be altered to alter the properties of the tube. When people talk about the carbon fiber "layup", that concept parallels tube selection in metal frames, where framebuilders might choose a tube with a certain thickness (or thicknesses, if butted tubes) and shapes (e.g. ovalized at the end, or they might bend their own tubes if a custom builder or do something like hydroforming if a mass manufacturer). However, there are more dimensions of freedom to vary carbon fiber construction in.
If a carbon frame is perfectly constructed and is perfectly handled, then per this Cyclingtips article, bicycle industry insiders think it would outperform metal frames in laboratory fatigue tests. That involves putting frames on a jig and running them through load cycles until they fail. One quote is:
“Composites do not behave like metals,” explained Chuck Texiera. “In fact, they don’t actually fatigue like metals in the same classic sense of the word. The fatigue life of the fibre itself is just about infinite.” ... In the absence of any impact, the matrix can deteriorate with use, but it takes an extremely long time. “The epoxy matrix will at some point start to form little cracks,” explained Chuck Texiera, “and then over time it will just have the connectivity of the fibre. So really what’s happening, over really extended periods of time, you can expect the stiffness of the frame to change ever so slightly but it’s such a small number. We can measure it but I really wouldn’t think it would be perceivable. But it takes hundreds of thousands of cycles to even get to that. Two years would be far too short for that to occur with any kind of typical age group racer.”
Carbon is not perfectly made
The issue is that all manufacturing processes have variation. At present, carbon fiber manufacturing involves a lot of hand labor, and hand methods inherently have more variation than machine-based methods. Moreover, the layup process is more complex than with metal tubes - a metal frame has 8 metal tubes, a bottom bracket shell, and other items, but in CF, each tube is composed of many segments of carbon fiber. The more steps there are in the process, the more opportunities there are for some error.
One manufacturing process error in carbon is that you can get voids in the cured structure - those are literally air bubbles, and more layers of fabric or a more complex structure probably make them more likely. Voids weaken the structure. A void in the wrong spot makes the structure more likely to fail over time.
(NB: in metal frames, there's no exact parallel. However, improper welds, especially with earlier titanium frames, have some similarities. Titanium must be welded under inert gas, and a contaminated weld is particularly prone to eventual failure.)
This is more of a design rather than a process issue, but when a when a carbon structure flexes against a metal part, that can cause failures as well. If the metal part has an edge, even if it isn't necessarily a 90 degree one, that edge can cut into the fibers over time. Raoul Luescher has been documenting some issues on a recent recall on the Specialized Tarmac SL7 fork, and the link goes to his most recent YouTube video on the problem. I'd characterize this as design-related, but it still does affect potential lifespan. Again, metal frames aren't inherently immune from design-related failures, e.g. ultralight metal bikes have really thin tubing. However, designers aren't trying to push the design limits of metal bikes.
Last, I mentioned handling. Imperfect handling can cause invisible delaminations (a technical term, in this context like a fracture) inside the carbon. Over time and repeated loading, the delaminations can propagate and eventually cause the whole structure to fail. In principle, a pretty light impact can cause a delamination, particularly if the carbon struck something hard.
What should riders take away from this? Carbon fiber's lifespan is potentially very long. It's not infinite, but arguably no frame's lifespan can be expected to be infinite under regular conditions (i.e. you're riding the frame and it takes a few knocks here and there and your maintenance isn't perfect). Carbon fiber is a newer material, and the designs are more complex than with metal tubes, so the manufacturers are still perfecting their manufacturing processes. Additionally, more cutting edge frames are inherently more likely to have manufacturing or design flaws because they are a) pushing the limits and b) because the layup is more complex, there are more steps in manufacturing, and thus there are more chances for an error to occur. Additionally, the lighter the carbon (or any other material) frame, the more of a risk handling mishaps pose.
NB: I have no engineering or cycling industry background. The statements above are drawn from industry interviews that I have listened to, plus my own inferences.