Imagine you magically had a supercomputer from 5 or 10 years in the future. It's only a few dozen racks and it scores one exaflop on the LINPACK benchmark. What would you use it for?
Sell it. (Probably in pieces to a few different interested buyers) :-)
Aside from making jobs for people in the US supercomputing industry, I don't see any application for faster versions of these same CPU-style supercomputers beyond low-rent stuff like climate modeling and physics research.
If the time traveling supercomputer's manufacturers start heading in the GPU direction, mining Bitcoin would be a pretty good way to monetize the appliance, if selling it wasn't an option.
I would use it to render huge fractal images, do interesting and complicated path tracing (ie physically-based raytracing in mantra or luxrender), and explore reaction-diffuson simulations. All assuming, of course, that completeley parallel and compatible code is available for all of these.
Edit: happy to write such code myself too.
Rent it out to people who want the compute power. Sell it to the same. Nothing I'm doing needs that much ability, so the profit from selling it could easily fund the purchase and use of a more moderate system, as well as fund other projects I'm interested in seeing done.
It's hard enough in this day and age to make something using its patent as a guide, even though by definition that's what a patent is supposed to show you how to do.
The reverse is next to impossible. If I had an Apple A8 chip from 2019, I might get some good ideas on what it's made out of, but not what processes Apple used to make it. Furthermore, many of the patents used in 5-10 years have already been applied for.
Playing time-traveling patent troll is also fraught with other problems- many innovations come precisely because they aren't already patented by someone else. Rather than giving into the extortion they may just push their research in a different direction and still achieve their performance goals, making your patents somewhat worthless.
Licensing a patent for free is silly- just abandon it when your 7.5 or 11.5 fees come due.
The one thing that would have significant resale value would be the supercomputer's software (assuming it wasn't something really esoteric), since 5-10 years' worth of hard work on the part of thousands of programmers could be skipped. That's assuming that was included with the supercomputer and you don't have to send for it (Doc Brown style).
Aside from making jobs for people in the US supercomputing industry, I don't see any application for faster versions of these same CPU-style supercomputers beyond low-rent stuff like climate modeling and physics research.
If the time traveling supercomputer's manufacturers start heading in the GPU direction, mining Bitcoin would be a pretty good way to monetize the appliance, if selling it wasn't an option.