There is more to these than processing power
Having done quite a lot of work on helmet mounted and head mounted displays, it is clear that these people do not understand the system level latency problem.
There are helmet mounted displays in a number of applications for military aircraft (Typhoon and F-35 come to mind), and the architecture to achieve a smooth transition when the pilot moves their head around is anything but easy (but rater necessary as there is conformal data such as is displayed on a HUD like heading, aircraft attitude and altitude to name a few). This issue has to be solved at system level and virtually no amount of processing grunt will save them from this issue if the data paths have not been carefully constructed.
Another issue is pre-distortion of the imagery because displaying on a curved surface requires the image to be manipulated to appear to be flat and that adds latency as well quite apart from getting the video from sensors in the first place.
Seems like they need to go back to a drawing board and analyse their various path delays.