In the interview he runs down the issues they encountered going down the pure AI accelerator path. It sounds like they've decided the opportunity wasn't there (i.e. too hard) so they've pivoted.