NVIDIA, every mention

AI Semiconductor Landscape feat. Dylan Patel | BG2 w/ Bill Gurley & Brad Gerstner · Bg2 Pod Dec 23, 2024 · 42 mentions

• ▶ 9:33 Dylan Patel You know, today everyone's freaking out about, or not freaking out, but like everyone's like very excited about Nvidia's Blackwell system, right? 2 times in the scene
• ▶ 13:30 Dylan Patel So, so the main thing for NVIDIA is, you know, hey, this workload is this big, right? 7 times in the scene
• ▶ 14:16 Dylan Patel Like with Blackwell, not only is it way, way, way faster, anywhere from 10 to 15 times on really large models for inference, because they've optimized it for very large language models, they've also decided, hey, we're gonna cut our… 2 times in the scene
• ▶ 15:23 Dylan Patel So, so I think a lot of people in the investor community, you know, call CUDA, which is just like one layer for all of NVIDIA software. 3 times in the scene
• ▶ 18:42 Dylan Patel NVIDIA has been pushing non AI workloads for accelerators for a very long time.
• ▶ 49:15 Dylan Patel I can group together over 256 users' requests on one server for Llama-Four of NVIDIA server, right?
• ▶ 54:34 Dylan Patel I can run Lama-seventy-b on one NVIDIA GPU, and soon enough there will be on one set of Amazon's Neutronium, right? 2 times in the scene
• ▶ 1:02:27 Dylan Patel And it, it objectively is when you look at the cost of goods sold of NVIDIA, um, their highest cost of goods sold is not TSMC, which is a, uh, 5 times in the scene
• ▶ 1:04:00 Dylan Patel It's a commodity in that sense, but because reasoning requires so much more memory, and the cost of goods sold of an H 100 to Blackwell, the percentage of cost to HBM has grown faster than the percentage of cost to leading edge silicon,…
• ▶ 1:04:00 Dylan Patel It's a commodity in that sense, but because reasoning requires so much more memory, and the cost of goods sold of an H 100 to Blackwell, the percentage of cost to HBM has grown faster than the percentage of cost to leading edge silicon,…
• ▶ 1:06:30 Dylan Patel Like, hey, we maybe can't design as well as NVIDIA, but we can put more memory on the package, right? 6 times in the scene
• ▶ 1:11:12 Dylan Patel Nvidia maybe has not focused on as much, right? 4 times in the scene
• ▶ 1:13:10 Dylan Patel But like, they'll go down because they, you know, and, and, and then you're like, oh, well, like what's the market clearing price for, uh, if I wanted an H 100 for two years or a year.
• ▶ 1:14:45 Dylan Patel And I think there's, there's a whole long story around why Apple hates NVIDIA. 4 times in the scene
• ▶ 1:22:23 Dylan Patel The problem is, nowadays, if you look at rental prices for H-one hundreds, they're tanking, right?
• ▶ 1:22:42 Dylan Patel Three-month, six-month deals at way better pricing than even the four-month or the four-year, three-year deals that you used to have for Hopper, right?