Bevin Slattery (NEXTDC founder), has written up some thoughts on where the demand cycle goes from here:

Why growth in Digital Infra has only just begun...

His conclusion:

If the effective size of the prize for having the best LLM on the planet is $10T in market capitalisation over the next 5 years, then what’s the right level of investment to get your piece of that pie? 5%? 10%? 20%? Well that’s the arbitrage. Right now, NVIDIA sold US$60B of chipsets last year. That is less than 1% of the AI prize. Let that sink in. If the world needs 50GW of data centres over the next 3-5 years that will cost US$300B to build. That’s about 3% of the prize.

No one is going to complain about GPU or data centre pricing for the next 3 years because the cost of not investing will be a 10-50x destruction of your market capitalisation. If you don’t keep up, you will be left behind. Just ask Intel…

I don't know if his numbers are right, but I put some weight in the broader point. There's a big prize to be claimed and investors will back investment or risk losing it.