If you mean they're sandbagging themselves then that is something that can't be done over time, because in that case you'd be just holding progress back. If they did sandbag themselves at one point so the next generation of cpus would seem more impressive then after that second generation you'd be back up to using the best there is again. If they alternate between sandbagging and then releasing something really impressive then for every two families of cpus released there'd be no change from if they weren't doing it at all. In the long run (a period of a couple years) everything evens out and it turns out they really do give you the best there is. That is if they were doing such a thing at all to begin with.

Another issue is that for something like that to even be feasible to do you would have to hide the fact that you can produce something better, but choose not to. Now considering that 99%+ of hardware innovations in cpus, gpus, ram, storage, anything really, is a result of collaboration between thousands of individuals and that academic research on these topics are available for anyone to read, it's a little difficult to hide the fact that it's possible to {insert generic manufacturing technique for drastically increasing chip density here} when everyone else who have read up on the technique from the same paper possess the knowledge already.

Detailed technical specs on all modern desktop cpus are widely available for everyone to read, including the competitors. If Intel were holding themselves back then AMD would see where and how Intel did it (or vice versa) and the result would be that Intel just handed AMD a clear and free advantage they could make use of by simply building their next processor without such limitations. It would make absolutely no sense as a business strategy.

The problem with conspiracy theories is that most of them require the cooperation of thousands of individuals to keep the conspiracy itself a secret. People in general aren't very good at keeping secrets. Thousands of people as a whole, even less so. This is particularly true in a field like computing and science where their entire field of study is based upon sharing of information and the benefits derived from it.