Cool stuff! I suspect we will now see a bunch of startups coming up focused on helping companies reduce their AI usage, similar to how we have companies focused on optimizing cloud costs.
Definitely - when you consider how varied inference workloads will be, and the different ways to minimize costs - better prompting, SLMs, different chips, batching, etc, there will be tons of opportunity