A glimmer of hope
from Google regarding the Spectre problem (variant 2). "Retpoline" effectively disables speculative execution by isolating branch target prediction. The patch works on the binary level and not at the OS level. Google claims that the performance loss is almost zero, close to the old mispredicted branch.
Of course this does not address Meltdown (variant3) so KPTI is still needed for the Intel CPUs with possibly large performance hits.