One: laser ring gyros instead of mechanical accelerometer or visual head tracking systems
two: Render a frame larger than FOV and digitally move that before the next frame is rendered.
The first costs on materials, power and research, and the second requires loads of processing power. Cheap compared to the cost of, say, an F35, but expensive for consumer electronics (which I gather is the point for this article).
For your edification as an armchair specialist on laser ring gyroscopes, framebuffers and all things VR, may I suggest that you at least bother to multiply the number of bits per pixel times the number of pixels times the FPS, to at least get a handle on how much data needs to be processed and understand the problem before you post.