Metaverse and crypto will be a bust, but democratized AI is going to explode, the big risk for NVDA is that models and tools get small and efficient enough that they don't need $1,000+ hardware.
You will always be able to do more and better with the premium hardware. Maybe for video games the marginal differences aren't enough to sway most consumers towards the high end cards, but I expect there will be a lot of people willing to pay top dollar to be able to run the best version of X AI tool.
For workloads though, does it really matter if it completes in 20 seconds instead of 15?
Unless there's some feature difference in the higher end cards (like AI-specific subchips), not just more of the same & faster, then a lower card of the same generation shouldn't impact your ability to run something or not.
Maybe RAM size...
Or maybe it's an automated thing that runs constantly? The difference between 20 and 15 seconds is 480 additional iterations per 8-hours. How much is that worth to you?
There's also the unmeasurable benefits. What if you're losing potential good employees applying because word on the net is your company uses last-gen hardware for everything, and your competitors use the newest.
I mean, it's no different than CPUs and RAM and such. A middle-of-the-road laptop from the last year or two is good enough for most users... no company I've ever worked for gets the top-of-the-line model every single refresh for every single dev. That'd be an absurd waste of money, IMO.
Personally, I wouldn't want to work for a company that upgrades hardware all the time for single-digit performance increases. That money is better used elsewhere, whether employee salaries/benefits or donating it to the community or just saving customers some money or paying dividends to shareholders.
Even if you're running massive continuous workloads, it is probably more cost-efficient to just parallelize a bunch of midrange systems (optimize for perf/dollars and/or perf/watt) than to just chase the highest-end chips for no reason.
I love when HN netizens just throw stuff like this out without further explanation. It's as if they're monks reciting the dogma. Sigh.
You and I both could write up plausible explanations of what he might mean by that, as both those topics are already discussed to death here. There's no particular reason to think he's being dogmatic, just pithy.
I'll also note that it would be easy to make the same accusation about your own posts. E.g.: "The contemporary democratic nation state is a facade over a distributed and decentralized oligarchy." https://news.ycombinator.com/item?id=32837598
But presumably you see yourself as holding that position for non-dogmatic reasons, so I'd encourage you to do the same for others.
The Metaverse is already a bust, essentially a stillborn fetus. The market doesn't actually want it.
Although his particular vision is very weird and ugly, Zuck is on the right track: in the future, people will spend real time (not async time, a-la tiktok/instagram) with each other in virtual spaces. Facebook is already a text-based version of the metaverse, and it minted a pretty penny.
The metaverse, as Zuck imagines it, is already here though: video games. People who are hardcore about video games wake up, go to work, get home, and then plug in. Their friends and social circle exists on the internet, in the game. This is the metaverse. Video-game communities; Fortnight [1] , WoW, League, CS-Go, etc; these are the most complete so far existing metaverse(s). One exists for every major flavor of gameplay we can think of. Look on steam for people who have 10,000 hours in a game. These people live as Zuck wants us all to: in the metaverse.
Zuck's vision is to expand these virtual communities beyond the hardcore gamer group. You go to work (in a metaverse) then you come home and spend time with your friends (in a different metaverse), then you play a game (in yet another metaverse), then you sleep and do it all again the next day. It's actually a very forward-looking vision, and one that probably will come to pass over the next few decades. A metric for this is "time spent in front of a screen." It inexorably climbs year after year, generation after generation.
But the key thing is that the experience has to be good. The metaverse of Fortnight or CS-GO is worth spending time in because those games are fun and engaging in their own right; for some, more fun and engaging than reality itself. These games print money, because a small group of people legitimately spend every leisure moment in them. Zuck's vision is to expand this beyond the market for hardcore gamers and into work, socialization -- life itself.
I suspect that the market doesn't quite fully understand this part of Zuck's metaverse dreams. In part because of Meta's spectacular failure at making the metaverse look fun and compelling.
Personally, my money's on Epic to actually bring the Metaverse to fruition. They already own a Metaverse, Fortnight, and they also own Unreal, which will probably be the software foundation upon which the normie metaverse gets built.
[1] https://medium.com/@ow/the-metaverse-is-already-here-its-cal...
Inexorably? That's not a measure that even can grow infinitely. There are only so many hours in a day. Nor should it grow infinitely. Many people find that taking up hobbies away from the screen improves their lives significantly.
If they can find a way to have us on the computer while sleeping, it will be done.