AI chips may move beyond the GPU architecture

Transcript

Transcribed by machine (Whisper) from the episode’s audio, so a word may be off here and there.

1. 21:02

   Yep. And the startups are also starting to go after new chip design as well. Yeah, well, that's the other thing is, yeah, you have these disruptive startups. And actually that, just for a moment on the chips, we're not really big investors in chips because it's kind of a big company thing. But it's a little bit of historical happenstance that AI is running on, quote-unquote, GPUs, you know, which in GPU stands for graphical processing unit.

2. 21:22

   So, and basically, just for people who haven't tracked this, there were basically two kinds of chips that made the personal computer happen. The so-called CPU, central processing unit, which classically was the Intel x86 chip. It's kind of the brain of the computer. And then there was this other kind of chip called the GPU or graphical processing unit that was the sort of second chip in every PC that does all the graphics.

3. 21:42

   And, you know, and this is graphics like, you know, 3D graphics for gaming or for CAD CAM or for, you know, anything else, you know, Photoshop or for anything that involves, you know, lots of visuals. And so the kind of canonical architecture for a personal computer was a CPU and a GPU. By the way, same thing for smartphones. By the way, and over time, you know, these have kind of merged. And so like a lot of CPUs now have GPU capability built in.

4. 22:03

   Actually, a lot of GPUs now have CPU capability built in. So, you know, this has gotten fuzzy over time, but like that was like the classic breakdown. But the fact that that was the classic breakdown, you know, kind of meant that while Intel had a, you know, monopoly for a long time on CPUs, there was this other market of GPUs, which NVIDIA, you know, basically fought the GPU wars for 30 years and came out the winner, like what was the best company in the space.

5. 22:24

   But it was like a hyper-competitive market for graphics processors. It was actually not that high margin. It was actually not that big. And then basically, it just, it turned out that there were two other forms of computation that were incredibly valuable that happened to be massively parallel in how they operate, which happened to be very good fits for the GPU architecture. And those two basically highly lucrative additional applications were cryptocurrency starting about, you know, 15 years ago, and then AI starting about, you know, whatever, four years ago.

6. 22:53

   And so, and NVIDIA, like, I would say, very cleverly set itself up with an architecture that works very well for this, but it's also just a little bit of a twist of fate that it just turns out that if AI is the killer app, it just turns out that the GPU architecture is the best legacy architecture is devoted to. And I go through that to say, like, if you were designing AI chips from scratch today, you wouldn't build a full GPU.

7. 23:13

   You would build dedicated AI chips that were much more specifically adapted to AI and would have, I think, would just be much more economically efficient. And, you know, Jen, to your point, there are startups that are actually building entirely new kinds of chips oriented specifically for AI. And, you know, we'll have to see what happens there. You know, it's hard to build a new chip company from scratch.

8. 23:33

   You know, it's possible that one or more of those startups makes it on their own. And some of them are, you know, doing very well. It's also possible, of course, that they get bought, you know, by big companies that have the ability to scale them. And so, you know, we'll see exactly how that unfolds. And, of course, we'll also, by the way, see, you know, the Koreans are going to play here, for sure.

9. 23:53

   The Japanese are going to play. And then, you know, the Chinese in a major way as well. And, you know, they have their own, you know, native chip ecosystem that they're building up. And so there are going to be many choices of AI chips in the future. And it's going to be, you know, that'll be a giant battle.