CS-4 just doubled wafer performance and said inference can run up to 30x faster than GPUs.
Here's why this matters:
Speed is where lots of demand is heading. Coding agents make hundreds of model calls in a loop so faster decode -> a job finishing quicker.
GPT 5.6 for example will have a premium fast mode which suits customers like:
\-> Quant funds: For them time is money.
\-> Pharma and research agents: $GSK has talked about intelligent research agents on $CBRS.
\-> Voice and real time support
$CBRS unlocks a new level of speed, and in AI...speed is what people are happy to pay a premium for.