@SullyOmarr Don’t think it will ever be too cheap to meter. Cost to serve won’t go to zero and volumes will be so large that even fractions of what it costs today will be huge revenue opportunities. Impressive nonetheless
@saranormous@NoPriorsPod Leading indicators of breakout potential in crowded, early application spaces (eg coding, customer support, etc.)
Everybody knows they’ll be big, but products & traction still very early across the board.
@BG2Pod@altcap@bgurley You end up with a barbell effect where only the best/largest/most stable companies go public (confidence in trajectory, high float, etc) or the worst companies go public (can’t get access to private market capital at attractive terms)
@BG2Pod@altcap@bgurley The mark to market impact of underperforming in the public markets can be very negative for employee morale. Add the extra overhead and headache of being public, for many companies it may seem more attractive to right the ship in the private markets
@ID_AA_Carmack Both true. Unified, low latency memory across 100s of GPUs is a step function improvement. But CSP-level models require 4 digit+ clusters, so interconnect / networking is still (and becoming more) critical