4DS - Anything but Charting, page-28002

There is a huge hole in the AI generative space. AI generative will run queries, then sub queries then more queries, then a lot return out of sequence and this need to be stored in DRAM. The AI motherboards are insanely expensive to normal server motherboards. This memory needs to be constantly refreshed tying up cpu and power cycles. Depending on our retention, I don't know if it's in seconds or mins but the net effect will be less servers required. Each server consumes power and also consumes very expensive $ per sqm in data centres. For an AI installation you could be saving in the millions, so its power, its cooling, its realestate, is physically less equipment you need for the same function. This is assuming we are similar to DRAM read and write speeds. On a side note I'm happy to be fact checked on the above, I'm a bit rusty on my server and memory knowledge. If you want to see generative AI in its simplest form then download https://apps.apple.com/us/app/wonder-ai-art-generator/id1621278575