Nvidia, every mention
25 scenes (said by Dylan Patel), the whole family · ← back to Nvidia
tap a year for its mentions
every year anyone Dylan Patel 93Anjney Midha 21David George 20Gavin Baker 16Erik Torenberg 15Marc Andreessen 14Guido Appenzeller 14Jeff Schmidt 13Dwarkesh Patel 12Martin Casado 10
Verbatim, from the transcripts: the passages where Nvidia comes up
Dylan Patel on the AI Chip Race - NVIDIA, Intel & the US Government vs. China
- ▶ 0:21 Dylan Patel It's kind of poetic that everything's gone full circle and Intel's sort of crawling to NVIDIA.
- ▶ 2:53 Dylan Patel Five billion NVIDIA, two billion SoftBank, um, US government was 10. 3 times in the scene
- ▶ 8:31 Dylan Patel The H-twenty got banned in the beginning of the year. 2 times in the scene
- ▶ 20:23 Dylan Patel Um, we'll see how long, and we'll see how fast the rest of the, you know, the, the NVIDIA technology advances past what Huawei's capable of, right?
- ▶ 33:38 Dylan Patel Whether it be, you know, the H-Twenty, which is more regulatory, but, like, other cases they've ordered and had to cancel.
- ▶ 45:25 Dylan Patel There's a story about how Volta, 2 times in the scene
- ▶ 45:35 Dylan Patel Um, and they decided we should go all in on AI.
- ▶ 47:08 Dylan Patel It's, it's, it's really a question of, and this is like, I think every time I've talked to, uh, you know, some executives at NVIDIA have asked this question, because I really want to know, and, you know, they won't answer it, obviously,… 16 times in the scene
- ▶ 50:48 Dylan Patel Um, CoreWeave, right, like, yeah, everyone's, like, throwing a fuss about it, but it's like, he invested a couple hundred million plus, you know, early on, plus, um, you know, rented a cluster from them for internal development purposes… 4 times in the scene
- ▶ 58:01 Dylan Patel Uh, that's sort of the name of the game today with NVIDIA's hardware. 3 times in the scene
- ▶ 1:04:02 Dylan Patel Like, you aren't using QDNN, which is NVIDIA's, like, library that's, like, super easy to generate your, you know, to generate kernels and stuff, right? 2 times in the scene
- ▶ 1:07:28 Dylan Patel Was, was the 20, 23 Q one NVIDIA not bigger? 2 times in the scene
- ▶ 1:08:10 Dylan Patel They'll deploy NVIDIA networking, um, InfiniBand or Spectrum X, um, and they have really good network engineers.
- ▶ 1:10:44 Dylan Patel If I'm talking about, um, a GV 200, right?
- ▶ 1:24:20 Dylan Patel You know, there's also B. 200. 4 times in the scene
- ▶ 1:24:46 Dylan Patel Um, but, you know, when, when you have a GB, when you have a, uh, H 100, right, box, or H 208 GPUs, one of them fails, you take the entire server offline, you have to fix it, right?
- ▶ 1:34:38 Dylan Patel So, or, or like, you know, if NVIDIA takes the same margin, then, then the cost of this pre-fill chip is much, much lower, and now the whole process is way cheaper and more efficient.
- ▶ 1:36:58 Dylan Patel It turned out that, um, multiple major neoclouds had sold out of hopper capacity. 3 times in the scene
Dylan Patel on GPT-5’s Router Moment, GPUs vs TPUs, Monetization
- ▶ 0:00 Dylan Patel NVIDIA's gonna have better networking than you. 8 times in the scene
- ▶ 22:04 Dylan Patel So, so I think, I think that's the biggest threat to NVIDIA is that people figure out how to use Custom Silicon more broadly. 6 times in the scene
- ▶ 27:28 Dylan Patel There's, there's a lot of capital being invested to dispel sort of Nvidia's top dollar or top position. 16 times in the scene
- ▶ 35:40 Dylan Patel Um, China, there's like provinces in China that have like rules that say the H-Twenty is not efficient enough to be deployed, which is like super bizarre because it's clearly the best AI chip China has.
- ▶ 36:21 Dylan Patel Um, they would, they like, you know, Jensen keeps saying he couldn't give away H-Twenty in America for free, but I've literally like heard companies like say, like now say like, yeah, no, I mean, I wouldn't because like, I only have this… 6 times in the scene
- ▶ 38:48 Dylan Patel We, what we see is that like, even with H-Twenty being sold to China, into China, um, and, and, and future versions of the chip, H-Twenty-E and other chips, um, we still see like Chinese companies like Alibaba, uh, renting GPUs outside of…
- ▶ 56:37 Dylan Patel It's like, well, NVIDIA's gonna spend tons and tons of cash, um, or, uh, they're gonna spend, like, you know, tens of billions of dollars of taxes, 6 times in the scene