The Problems
All of this sounds great but it is pretty much vaporware until we can build it. Reconfigurable computing still faces many issues before its true potential can be realized. The major stumbling block is creating software that can dynamically reconfigure the FPGA in real-time.
The two predominant ways to configure an FPGA is during compilation (static) or during execution (runtime). For the FPGA to reconfigure during runtime, it has to interpret the instructions it receives and decide on the optimal configuration. If the FPGA isn't configured properly, it will lose its performance edge over general purpose processors.
Even static reconfiguration isn't the solution. It would take some real effort to get software companies to recompile their code.
Another impediment to the FPGA's acceptance is its high cost. A FPGA development kit can cost anywhere from $1,000 to $3,000! Imagine if you have to buy not just one; but multiple FPGAs! With even high-end PCs costing less than that, it certainly doesn't encourage people to invest in FPGAs.
Companies would look at it from another perspective. Is there potential profit in the market worth the risk and effort of developing FPGA-based computers? Most high-end systems cater to gamers trying to push that extra 2 FPS for bragging rights. The large investment in research and development may not result in large enough profits to be commercially viable for most companies.
Still, there will always be niche groups who would love to have the heavy computing power that FPGAs can provide at any cost.