Cost, size, weight, and power are definitely things that can be improved on. The atomic clocks you can buy for a lab are big 3U boxes but shrinking it down to something portable is the future.

I actually work with one of those chip scale atomic clocks. The goal is to put it in space. If you need precise timing in space, the current state is to lock to a GPS reference clock (just for timing, don't care about position), train an OCXO, and let it drift for a day or two before you require GPS again. OCXOs (and crystal oscillators in general) can be very accurate but the stability is easily affected by temperature swings. So as your satellite goes in and out of the sunlight or as components inside turn on and off, the surrounding temperature changes and can make the clock drift. This isn't a problem on earth since you can always have a GPS receiver running but on a power constrained satellite, this is not ideal. The CSAC gets 100x better thermal stability than even some of the best space-qualified OCXOs. This means you could potentially run for 100x longer before you need to reacquire GPS, assuming the same environment and performance requirements. The CSAC also has a brain inside so potentially you can do all of your characterization on the ground and not even need a GPS receiver onboard. It also requires much less time to become stable after powering on than a normal crystal oscillator.

You get the same benefits on the ground, super precise timing for the processors and modems without necessarily needing a GPS receiver. The modern assumption (and pretty much proven in Ukraine) is that GPS will be completely unavailable on a battlefield. Anything portable or battery-powered will likely not be able to make use of newer anti-jam features of GPS (very power hungry) so you need something small, precise, and quick to come online.

Modern waveforms often require very tight timing to properly demodulate so having a pocket sized atomic clock solves a lot of problems.

Where are you getting GPS receivers that work in space?