Its easy to answer that, in terms of Gbps/Km/$ that fiber occupies a very special niche and that given certain assumptions it's very hard to beat. Other posters have already made that point, and it's a good one. However there's also the point that what actually "wins" in the long run probably needs to break some of those assumptions that fiber requires. So in the long run it's unlikely that most links go to fiber. Wireless today serves a lot more people than fiber does, and that balance is likely to shift more in favor of wireless over time. Fiber's achilles heel is that you have to run a fiber to every individual location where you want access, and digging trenches is really slow, expensive, and complex. The thus connected devices also need to have physical link attached to them. There are some places where it makes sense to do that, and over time that list of places will grow, but it will always be a small subset of all the places that people want to have access. In the meantime wireless will continue to improve by leaps and bounds because the interface to the overwhelming majority of leaf devices is wireless and it's going to stay that way. Fiber is a niche technology serving a small subset of all connected devices - and it's going to stay that way. The reasons for that are both technological and financial.
The theoretical data carrying capacity of an optical fiber is ridiculously high, but largely inaccessible to near future technologies. The nonlinearity of direct modulated lasers, when combined with the signaling rate make modulation schemes beyond NRZ prohibitive in the near term. Getting to higher data rates per fiber requires increasing the high frequency corner (very expensive) or going to multiple colors (even more expensive). And for building to building distribution you still have to dig that trench (really, really, really expensive). Nothing on the roadmap today is going to change any of these variables in the next 20 years. In principle things can be done about this, but nobody is spending significant amounts of money on those technologies. Wireless, OTOH, is seeing a lot of investment.
The theoretical bandwidth of a point to point directed RF link constrained to several GHz of spectrum is quite a bit smaller than what fiber offers, but still vastly in excess of what we use today and it's full potential is much more accessible to near future technologies. If we were really up against the limits of RF capacity today then, yeah, fiber would be a good bet. But we're not - wireless speeds and network capacities can and will grow by orders of magnitude before we start to bump up against the theoretical limitations. It's cost effective, from a network architecture standpoint, to get in-home wireless links that are much faster than the available ISP speeds almost everywhere today, though the business models of cell companies are not yet aligned with that service model. LTE can get up to 1Gbps for fixed links and routinely hits 100Mbps for mobile links and 5G is likely to be 100x faster on both fronts. Increasing capacity in those networks is a capital expenditure decision, it is not being limited in any meaningful way by physics. Wireless companies in the U.S. are trying hard to keep their (extremely) high margin business, so their capital outlays are not aligned with providing lots of cheap bits. But providers in other parts of the world show that the economics of being a wireless pipe service are not actually all that bad.
Fiber does and will continue to own long haul, back haul, core networks, data centers, and a lot of fixed point services. And it will grow. But wireless is going to grow much faster both in terms of capability and in terms of device connections.