Especially for nvidia, who have been badly burned by crypto bubbles going from 100->0 very rapidly in the past and ending up with massive oversupply of both used and new GPUs.
Especially given the AI people are buying dedicated business GPUs and not gaming ones like the crypto miners were, that's probably good news for gamers that nvidia haven't gone full bore AI only. If they were, it wouldn't make sense to spend fab time on a new series of gaming GPUs as they have leaked they're doing.
Continuing to also produce gaming products give a few advantages, besides simply to serve as something to fall back to if the AI boom turns out to be a bubble.
For instance: - Having consumer GPU's out there has marketing effects and can also increase the population of developers who experiment with AI development.
- Consumer GPU's can be built with cheaper memory types (no HBM needed) and less efficient process technologies (like when the 30-series used Samsung's).
- Having an installed base of consumer GPU's out there allow AI companies to deploy AI products (inference) to edge devices.
- Related to the previous: In the future, AI technology, VR, gaming and TV/film may merge into a single entertainment class, with everything happening or being told in an interactive VR environment, complete with fully AI agents acting within. Having much of the processing power in the edge device may increase the quality of such experiences, and may cause consumers to be willing to significantly increase spending on such products.
But that is what happens. The issue with GPUs is that a) there are only a small number of companies that make them b) fab capacity has to be booked in advance, and c) they know it's a bubble, so they're not going to commit to buying a large amount of future fab capacity when it could pop at any time. So then they don't actually increase supply, and prices don't go down.
> Maybe increasing the supply only increases the supply of luxury condos, which if they are all consumed by wealthy individuals might push housing prices upwards etc.
Only in the sense that the average cost might increase because the average unit is now larger, not in the sense that the existing smaller units would cost more rather than less. After all, their occupants no longer have to outbid the wealthy individuals who have put their money into the new luxury units instead.
> Such activity could, similar to bitcoin, actually stimulate enough economic activity in an area itself to sustain upwards trajectory on pricing
The premise here is that if you make an area more attractive then more people may want to live there. But that's fine. Suppose building 10 units increases demand by 5 units. So if you need 10 more units, build 20 more units.
In some kind of hypothetical edge case or rare circumstance, building 10 units might increase demand by 11 units, but that is obviously not sustainable -- if you built 50 million units, there aren't 50 million people in the region to live in the city, so at some point it stops being true, if it even ever was.
> Markets are weird and inefficient in all kinds of exciting ways!
This isn't markets being weird. If prices are high then construction companies want that money, which they get by building new housing. Weirdness only occurs when regulations interfere with their natural market incentive.