TikTok video #7457178428010188062

You know, if you were overwhelmed by Jensen Huang's CES keynote, everybody was. This is not Steve Jobs, this is not one more thing. This is 20 more things. I boil it down for you to three key things that you can take away here. Number one, they are sticking with their roots in consumer. They are announcing the GeForce RTX 5000, it runs on Blackwell architecture. It's a GPU for gaming. They've always done gaming. Why is this interesting? This is interesting because at the end of the day, they are moving from just being a GPU that can handle deterministic graphics for video games to a Blackwell architecture that's designed for AI from the ground up. It's going to become the most powerful AI system in a lot of consumers' homes. And their introduction to AI, even if they're suspicious of chat GPT, is going to be gaming. Because a lot of games that are going to ship are going to use generated graphics. It's vastly cheaper. They're going to have some percentage, like they'll have the main character and like 20% of great game graphics that are deterministic or sort of built into the programming. And then they're going to have like 80% that are generated, where it's one of those things where it goes to Blackwell and it generates the graphics for the game that look appropriate and you move on. That is going to be the way a lot of games ship. And if that's the case, think about it. You can't run the game if it's not on Blackwell. They've reinforced their ecosystem. Okay, number two, they are really into the infra layer that runs on top of their chips. And that's a deliberate play to lock you into the ecosystem. So with Nemotron and with Cosmos, they're building these worlds that are going to enable companies to stay locked into the NVIDIA chips. They don't want to move. So Nemotron fine-tunes Meta's llama-based language models for Enterprise. So Nano, Super, Ultra, those are all flavors of Nemotron, and they're for different Enterprise sizes. And really the takeaway there is that Enterprises sometimes don't want to roll their own. They would rather just be told this is the correct way to do this. And NVIDIA's got it, and here are the chips, and we've got the language models, and thanks Mark Zuckerberg for training that model for us. Cosmos is a world foundation model, similar to the gaming thing. It generates photorealistic environments, but this one is for training robots and autonomous vehicles, so that you can train them safely and they don't generate warehousing accidents. Again, you're not just selling the chips and the robots, you're selling the whole infra-layer. All right, third thing, this gets back to robots and automotive. Jensen clearly thinks that robots and automotive are the future. And so he launched Thor, which is a system on a chip that's designed to deliver 20x the performance of their previous system on a chip for autonomous driving, which was called Orin. It has new safety certifications, it's designed for autonomous driving, it's designed to handle pathway planning, it's designed to be able to handle all of the sensors that run on autonomous vehicles. Again, NVIDIA hardware in the car, NVIDIA software. In fact, they have a new OS, Drive OS, that goes with that. So this is all reinforcing Jensen's play. And when he talks about the Omniverse and Isaac and Cosmos together, it's basically the same thing, but it's basically looking at training and simulation environments for robots. Again, infra-layer, chip-layer, it all comes together. So when I look at this, and I look at what really matters, NVIDIA wants to knit together hardware, software, and data to remain the essential AI platform. From mainstream desktop GPUs, like the GeForce RTX, to AI foundation models like Nemotron, to robotic systems like DriveThor, these are moves that are going to define what NVIDIA is doing over the next half decade. It was a hugely impactful CES. I'm going to write a sub-stack about it that goes much more in-depth, because this TikTok is not a great sort of short-form video format that enables me to cover all the detail here. This was an enormous, enormous inflection point in the year. We got to see what NVIDIA is planning. So I hope you enjoyed this. It's a nice teaser trailer. There will be more in the sub-stack. Cheers. Bye.