“Compute will get cheaper” is a risky assumption to build your margins around.
An agent can make dozens of model calls to complete one task. Add retries and failed attempts, and the bill grows quickly.
I’d want to know the cost per completed task before pricing the service.
Renting Nvidia’s H100 costs 45% more than it did in December.
It went on sale in 2022, and two newer generations have come out since.
Yet, demand for compute is so far ahead of supply that older GPUs are getting more expensive.
@MGMurray1 I’ve been exploring a related theme in my newsletter, The Trust Layer, about how agents can connect information across systems while keeping people in control, glad to see this being explored in healthcare too
@gauraavvii People are going to pay for it. More people will start their own ventures and they will pay for subscriptions. More like how we pay for milk.
In 1988, an eight-page letter from my maternal uncle changed how I saw myself and the world around me.
Decades later, its lessons continue to shape my thinking on identity, trust and the humanisation of technology.
Read the full story in my latest LinkedIn post: https://t.co/LApydzZJF9