China still very far behind on memory bandwidth speeds and GPU production, they won't save us anytime soon
Of course prices will eventually fall but it could take 5+ years
China’s persistent lag in hbm stacking yields,advanced euv lithography capacity,and high bandwidth interconnect ip means domestic gpus remain bandwidth constrained relative to the H100/B200 class, rendering near term supply relief improbable
Price normalization will require not merely incremental process node gains but fullstack ecosystem maturity foundry capacity,packaging,& software stacks that historically takes half a decade or longer to achieve at scale
Price normalization will require not merely incremental process node gains but fullstack ecosystem maturity foundry capacity,packaging,& software stacks that historically takes half a decade or longer to achieve at scale