So that's just a hype train?
Big-name tech companies will together unveil yet another interconnect, this one dubbed Compute Express Link or CXL, which is aimed at plugging data-center CPUs into accelerator chips. The technology is built on fifth-generation PCIe protocols and electrical connections, meaning it can top out at 32 billion data transfers per …
Your software has to run on something, it's not just magical pixies or demons that do it.
An FPGA I think is really just a software-defined CPU (or any other sort of xPU). In fact, a FPGA is a more general processor than a CPU, as it can be reprogrammed on the fly via software configuration to replicate any other processor you have enough gates for.
If you have an FPGA with enough gates, and the license to do so, you could program your FPGA (via software on the fly) to be an Intel CPU, or an AMD GPU, or ARM, or whatever.
A GPU is just a hugely parallel processor from this perspective, it's not a 'Graphics'PU. Graphics is just a subset of what can be run on a massively parallel processor like what is traditionally called a GPU.