{"slug": "toyota-shows-off-soc-that-stores-model-weights-and-its-os-on-integrated-memory", "title": "Toyota shows off SoC that stores model weights and its OS on integrated memory", "summary": "AISIN, the automotive components maker in which Toyota is the major shareholder, has demonstrated a 16nm FinFET system-on-chip that stores AI model weights and an operating system in integrated STT-MRAM instead of external DRAM, reaching kernel startup in 65.7 milliseconds — 38 times faster than with external memory. AISIN and Tohoku University researchers said a hybrid 8-bit/16-bit multiplication architecture cut Tiny YOLOv3 weight data by 53 percent, from 17.7MB to 9.36MB, while maintaining 16-bit precision, and the chip runs the Tiny YOLOv3 object detection algorithm on-device. AISIN has not said if or when the chip will enter production, as Nvidia and Qualcomm continue courting carmakers for their silicon.", "body_md": "# Toyota shows off SoC that stores model weights and its OS on integrated memory\n\n[The Register](https://www.theregister.com)\n\nAccessing external memory puts the brakes on physical AI, so carmaker has found an economical and speedy shortcut\n\nJapanese carmaker -maker Toyota has decided to have a crack at the market for AI processors.\n\nToyota is the major shareholder in a company called AISIN, a manufacturer of automotive components. Both companies are aware that cars are increasingly reliant on tech like image recognition, and that [robotics](https://www.machinebrief.com/category/robotics) and other forms of “physical AI” will need parts that overlap with cars.\n\nAISIN thinks existing chips aren’t a good fit for such roles, because they rely on external memory to store AI models and then generate a lot of traffic between a processor and the model. That traffic, the company contends, makes image recognition and other on-device uses of AI less effective. DRAM is also currently vastly expensive.\n\nThe company’s response is a system-on-chip built on a 16nm FinFET process, an Arm Cortex A53 dual-core CPU, plus an Arm MaliG31 [GPU](https://www.machinebrief.com/glossary/gpu). The chip includes built-in STT-MRAM, which stores both model weights and an operating system.\n\nAISIN and its research partners at Tohoku University say that approach means the chip can go from power-on to kernel startup in 65.7 milliseconds – 38 times faster than is possible if using external memory – and run the Tiny YOLOv3 [object detection](https://www.machinebrief.com/glossary/object-detection) algorithm on the chip.\n\n“To effectively utilize the limited on-chip memory capacity, a hybrid multiplication architecture combining 8-bit and 16-bit multiplication was developed” according to a machine translation of Tohoku University’s [announcement](https://www.aisin.com/jp/news/uploads/20261102_1.pdf) [PDF]. “By combining 16x16, 16x8, and 8x8 bit multiplication, the amount of [weight](https://www.machinebrief.com/glossary/weight) data for the AI model was reduced by 53 percent, from 17.7MB to 9.36MB, while maintaining 16-bit precision.”\n\nAISIN is an established parts supplier and already sells to Toyota and other automakers but hasn’t said if or when it can get this chip into production.\n\nMeanwhile, the likes of [Nvidia](https://www.machinebrief.com/glossary/nvidia) and Qualcomm very much want to sell their silicon to carmakers, and position themselves as sources of tech and AI smarts that existing automotive supply chain participants can’t match. ® \n\nGet AI news in your inbox\n\nDaily digest of what matters in AI.\n\n## Key Terms Explained\n\n[GPU](https://www.machinebrief.com/glossary/gpu)\n\nGraphics Processing Unit.\n\n[NVIDIA](https://www.machinebrief.com/glossary/nvidia)\n\nThe dominant provider of AI hardware.\n\n[Object Detection](https://www.machinebrief.com/glossary/object-detection)\n\nA computer vision task that identifies and locates objects within an image, drawing bounding boxes around each one.\n\n[Weight](https://www.machinebrief.com/glossary/weight)\n\nA numerical value in a neural network that determines the strength of the connection between neurons.", "url": "https://wpnews.pro/news/toyota-shows-off-soc-that-stores-model-weights-and-its-os-on-integrated-memory", "canonical_source": "https://www.machinebrief.com/news/toyota-shows-off-soc-that-stores-model-weights-and-its-os-on-1uce", "published_at": "2026-11-03 03:25:32+00:00", "updated_at": "2026-11-03 06:20:58.921177+00:00", "lang": "en", "topics": ["ai-chips", "ai-infrastructure", "computer-vision", "robotics", "autonomous-vehicles"], "entities": ["Toyota", "AISIN", "Tohoku University", "Arm Cortex A53", "Arm Mali-G31", "Tiny YOLOv3", "Nvidia", "Qualcomm"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/toyota-shows-off-soc-that-stores-model-weights-and-its-os-on-integrated-memory", "markdown": "https://wpnews.pro/news/toyota-shows-off-soc-that-stores-model-weights-and-its-os-on-integrated-memory.md", "text": "https://wpnews.pro/news/toyota-shows-off-soc-that-stores-model-weights-and-its-os-on-integrated-memory.txt", "jsonld": "https://wpnews.pro/news/toyota-shows-off-soc-that-stores-model-weights-and-its-os-on-integrated-memory.jsonld"}}