One problem for internet vs AI is the internet boom laid down a bunch of fiber, which has a usable life of decades. The AI boom is 10% datacenter, 90% GPU, and the GPUs are consumables on a timeline of ~3-5 years, and even that's being optimistic compared to rumored real-world degradation rates. 2000 put a couple trillion in the ground that we got to dig up later, 2026 is setting a couple trillion on fire.

The usable life of a model is infinite. You can always fine tune it for current events and technology, or distill into a new one with a more efficient architecture.

If Anthropic or OpenAI go bankrupt they will have to auction off all their IP including code and weights. Claude won't disappear just change its name if someone else buys the trademark.

> the GPUs are consumables on a timeline of ~3-5 years

When newer GPUs with clearly better perf/$ are released at a yearly cadence, then older GPUs might indeed become _obsolete_ at this pace, meaning that it makes more economic sense to swap them out because datacenter occupancy is extremely limited. Many of these GPUs might themselves still be functional.

The infra and its lifespan isn't what created the value that came out of the Internet. The question I think is how much value will people create with these investments