The target markets were too small (or even were shrinking) to justify the investments necessary to keep the architectures going. Keep in mind that all in all on

The target markets were too small (or even were shrinking) to justify the investments necessary to keep the architectures going. Keep in mind that all in all only around 10k machines were produced over the technology lifetime (75-92). That's not a large number. Lisp machines started with hand-made CPUs and ended with special micro-processors (TI, Symbolics, ...). The jump to Lisp-supporting newer RISC (or similar) architectures did not happen, because the main sponsors (DARPA, Military, etc.) did no longer saw them as interesting - applications could run on workstations and high-end PCs. Next-gen CPUs were in the making in end 80s / early 90s (Symbolics, Xerox, LMI, SPUR...) but they did not reach the market - money was running out. The whole thing had its last incarnation in commercial emulators: Medley (the Interlisp-D from Xerox) for Intel+SPARC and Open Genera (Symbolics) for DEC Alpha.