It is extremely unlikely that this will ever be possible.
Fundamentally, any computation can be done more efficiently on hardware that was specifically designed for that computation than on general-purpose hardware. The only question is when there is enough to gain from the (possibly enormous) costs to design and build custom hardware for the used proof-of-work function.
For Bitcoin, that happened somewhere in 2013, when the first ASICs became available for mining. Currently, Bitcoin mining ASICs are many orders of magnitude more efficient than CPUs. CPUs will continue to evolve, but as long as Bitcoin remains profitable, so will ASICs. There is no reason to assume that general-purpose CPUs will ever get close enough to be able to compete with ASICs.
Unless of course the Bitcoin proof of work function is changed. While that is theoretically possible, it is an extremely invasive change, which needs near unanymous support throughout the ecosystem. That is extremely unlikely to happen, unless the alternative was an almost certain threat to the system as a whole, and changing PoW was the solution. Even if that approach is taken, it would be a horrible precedent, hurting the premise that Bitcoin is immune to such changes. I don't see this happening.