Google Is Making CPUs! (Tensor)

Google is expanding hardware ambitions beyond software by introducing its Tensor chip, a mobile system on a chip intended to tightly couple Google's software with its own hardware. The video explains that Tensor is designed as a mobile focused SOC, combining a CPU with other components like RAM, modem, and audio processing to boost efficiency for Android devices, particularly in the Pixel lineup. It highlights that while Tensor shares the goal of silicon-level integration with Apple and Samsung, Google is taking a more in-house approach to chip design, aiming to optimize on-device AI workloads for features such as image processing, improved low-light photography, and smarter voice input without always relying on cloud processing. The presenter notes that Tensor is expected to emphasize AI workloads, enabling multiple AI algorithms to run in parallel on Pixel devices to enhance photo quality, reduce motion blur, and provide better captions and object recognition. The discussion also covers how Tensor could reduce dependence on cloud AI, potentially freeing up bandwidth and enabling some apps to work offline or with weaker network connections. The video briefly touches on the Pixel 6 as the first device to feature Tensor and anticipates a fall 2021 release window, while suggesting broader device compatibility in the future. A sponsored plug for Grammarly is included, describing how the writing assistant can save time on professional communications. Overall, the segment positions Tensor as Google’s move toward more vertically integrated hardware to optimize Android experiences, especially around imaging, AI, and on-device processing, while leaving the door open for broader hardware adoption later on. The takeaway is that Google aims to close the gap with Apple and others by delivering a tightly integrated, AI-optimized mobile chip that could reshape how its devices handle photography, language processing, and real-time AI tasks.

Topics · technology · mobile_devices · artificial_intelligence · semiconductors · consumer_electronics