1. Unless we are significantly better than DRAM then it's just...

1. Unless we are significantly better than DRAM then it's just too hard and costly for the industry to change. From engineering, to purchasing to fab production changeover to awareness. This would be a multi year exercise just to replace dram.
2. We can substantially be better than DRAM in generative architecture with changes on the motherboard and clock speed.
3. The scale of generative AI means you could be storing 1TB of data in DRAM for a basic query, using processing power to refresh instead of running more queries. (this is a high level view and if someone can go deeper by all means dig in). In the mean time, new queries are banking up. (more gpu/racks required) Power savings are multidimensional, save power on memory refresh, save power on cooling, we become more efficient within the cpu/gpu, meaning less racks of server per installation. In data centres this will equate to millions of savings per mega installation.
4. I have been around for Y2K, dot com boom and watched crypto. (don't like it), and now AI, and this is not a bubble, this will continue. I have a few AI investments. One we use automation and robots to do the heavy lifting of the sales process. 450,000 interactions with customers for the month. Sales performance went up by 7x. They can only talk to a consultant when ready to buy. That is not generative but this AI thing will keep going. Meaning more data centres, more power, demand for space and power will be a premium.
5-6. I just try to read between the lines
7. I'd personally be looking to develop AI relationships now
8. truly amazing, you can repurpose memory as you see fit
13. This is true

Let the new guy do his stuff, he comes highly recommended.