Why AI Computers Cost THOUSANDS (It's Not What You Think!)

Transcript

1. 0:00If you're sitting there thinking like, wait, $4,000 for a computer or $10,000 for a computer in an era where we can
2. 0:06get a decent computer for 1,200 bucks, why would I do that? Again, because AI tokens um are very valuable things and you can
3. 0:19you can think of it less like I'm buying a machine and more like I'm I'm buying a potential employee that I can train up to do stuff for me
4. 0:29that works 24/7. The implication of that is that as AI capabilities get better and more people discover them and AI starts to
5. 0:43take over work and more people start using it at work, more people start using it at home, more people start using it everywhere the
6. 0:50demand for for AI inference, the the ability to use these things is going to go up dramatically. Now, Apple and and
7. 1:01AMD and Nvidia and all of these companies that are going to make specialized hardware to to to push inference to what they call the edge so
8. 1:12to your local computer is going to help with that, but the hunger for for training AI and for using AI is
9. 1:22going to skyrocket. So, I don't see data centers going away anytime soon. What I do see is data centers on the ocean.
10. 1:32So, I saw a proposal today uh of a company that is doing floating data centers where the wave action generates
11. 1:40the electricity and the ocean water cools cools the things. Um Elon Musk and a couple of other companies are talking about putting data
12. 1:50centers in space. Um there's likely going to be data centers in the middle of the desert with massive solar farms.
13. 1:56Um there's likely going to be data centers next to hydroelectric dams or substations where there's spare
14. 2:02electricity. There's like there's all sorts of plans going on. Um I don't see them going away at all. Why do we need all these data centers?
15. 2:12Well, AI So, if you think think steam engines and then think electrical grid and then think Interstate Highway
16. 2:27and then think internet and then think AI. Um computers, internet um AI is going to be infrastructure-level um
17. 2:41capacity for every country. So, if you don't have enough compute to do all the stuff, to run all the robots, and run all the cars, and
18. 2:50you know, the fleet of robo taxis that Tesla's going to put out and everyone's going to put out um your country will not be able to
19. 2:58compete. So, having access to high-powered inference will literally be power. Like like electrical power.
20. 3:07Right? If you're going to live without AI compute, you're not going to be able to do as much stuff. So, um yes, local inference will help. I
21. 3:14think most inference for individuals will migrate out to the edge. Like we will have phones within the next
22. 3:253 years that are so powerful and then models that are so good that are that are quantized for phones that everyone's
23. 3:32individual device like this will will likely be doing things as powerful as what I'm doing on a 64 gig MacBook Pro
24. 3:40right now. But the high-end compute stuff, the I think the the appetite for that is insatiable. Watch the full replay at
25. 3:47community.thesalon.ai.