It probably benefits Thunderbolt the most, where you do have GPU's and high speed storage that could use a doubling of bandwidth because fewer pci-e lanes are available to the adapter.
But...
4.0 is going to generate more heat than 3.0
Traces will be shorter which anyone who knows how short TB cables are now will understand is going to have a balancing impact on how useful the added bandwidth is
Given that high impact lanes are coming straight off the CPU, we won't really see 4.0 from Intel until they both have 10nm AND they have the heat from 10nm better controlled, not to mention that 4.0 requires new CPU design (probably the real reason they are behind the 8 ball on this tech)
The real world impact of SSD's going from SATA to NVME was limited already. The impact of 4.0 on SSD's real world results (not benchmarks) isn't going to night and day anyone. Shave a second off a software load or 1.5 sec from a boot?
SSD's moving that fast are going to be hot as can be until we have much more efficient controllers. Again that means you won't see the tech in anything but desktops or servers for years.
It's something Intel is definitely lagging on. But they aren't wholly incorrect on the impact to real world computing. Will we eventually need the speed? Yes surely. But the software world will continue to develop products that hit the middle of the market with some products (games usually) targeting the more advanced hardware once in a while. So even if you have the potential, for this market you won't really see results from actual products that have a real effect on your daily use for 5 years or so. That means Intel should be ready now, but if they are ready in a year or two they'll be late but not out of the game.