I suppose the main reason is the market requirement for the laptop. Previously, the consumers’ needs for laptops is function-intensive, like massive applications, display, and amusement performance. Then the processor development targets are more general. Cost and power consumption can definitely become better when the corresponding module is embedded into the core chip, with a higher manufacturing process.
FPGA steps into our life because of its flexibility of being reconfigured to any logic circuits, a.k.a. Accelerator. This kind of logic is tailed for the specific application, which can be highly pipelined and achieves much greater throughput. Compared with FPGA, the ASIC is mostly fixed that only works for a few applications.
The essential technical obstacle to widely using FPGA is the design cost. Even if the FPGA showed up in the 1980s, the specific circuits should still be developed fine-grandly, rather than writing behavior codes like for CPU or GPU.
In recent years, high-level synthesis technology is fast developed that aims to transform high-level code (like C, and Python) into low-level logic. This tech is already applied in the data center which has a computation-intensive requirement. However, it’s still in the preliminary stage.
