I don’t think the effects on the hardware market of open weight models triumphing has been reflected upon enough. It’s not clear that it will be less impactful if every enterprise decides to build for predominantly on-premise inference. In fact, before the question is ultimately decided, we’re probably heading towards a few years where hobbyists, enterprises and ‘hyper scalers’ are all competing for certain parts in common. Ram booked through 2027 sounds about right.

The stock market crash caused by AI companies halting their planned build outs may in practice cause RAM to remain unaffordable.

Last thing i heard was that large companies are starting to seriously spend on ai infrastructure (think nvidia clusters rather than tokens from openai or anthropic) to fine-tune and run custom models.

This might become the driver for memory and chips demand in the near future.

But i don’t know, of course, we’ll i guess.

Fable would have a much bigger userbase if Anthropic didn’t remove a zero data retention option because muh safeguards