Multimodal AI is reshaping compute demand. - Video foundation models - Real-time computer vision - AI avatars - Spatial computing These workloads don’t behave like chatbots. They require a new class of GPU infrastructure 🧵⚡