And you are starting to see this really kind of become more distinct when you look at China and then the rest. So the way we see, and we are already very deeply engaged with the Chinese is that AI is coming in, in 2 ways, right? So end-to-end ADAS are AI-driven. Today, most of the ADAS uses AI, but it's not end-to-end AI. So that's one big change. The second is AI as a smart assistant for the cockpit. That's the one that we are initially more focused on, and the two wins that we have talked about, one with Zeekr, the other is now with Cherry are for this SmartCore HPC that will bring this AI-based smart assistant for the cockpit. And if you may remember, we talked about our cognitoAI. It was also featured in our CES earlier this year, is the first framework, software framework of its kind that enables you to run AI -- Gen AI models in the car, not in the cloud. And you see many references today to having whether it is Gemini, Google, Gemini AI or something else, but these are essentially cloud applications that limit how extensive that AI-driven functionality can be offered. But in China, we see that already happening. In fact, the 2 launches that I mentioned earlier, next year will feature these AI models, probably from DeepSeek initially. And then on the -- with respect to the rest of the regions, we are seeing a lot of interest in Europe, as you can imagine. But for Europe, the difference that we're seeing is sort of redoing the whole cockpit system to be able to be built on top of AI, they're thinking of AI as an accelerator. So imagine an ECU that you bring in with minimal changes to existing cockpit domain controllers, that would enable them to offer some level of AI-enabled features. So it's not as expensive as what the Chinese would be able to offer, but it is still better than nothing. And that is seen as a stepping stone towards a full-blown AI-driven cockpit. So we are really focused on both those opportunities. One minor thing that is still very relevant that I would like to highlight, it's interesting to note that when you look at the entire cockpit and what the programs in China are doing, they tend to use Qualcomm silicon. But for the AI box as an accelerator that tends to be more NVIDIA. And so we are in the process of really developing solutions for both those architectures, which is kind of unique. I do not expect much activity in that regard from all competitors, especially outside of China, and that should position us well to take advantage of this emerging trend.