The TensorG3 used in the Pixel 8 and Pixel 8 Pro has so far been disappointing compared to its competitors, forcing Google to turn to a completely custom solution to compete with Apple's A series and the upcoming Snapdragon 8 Gen 4. Unfortunately, it will take some time for the rumored TensorG5 to be available, but according to the latest news, it will be mass-produced using TSMC’s N3E process, which is also used by Qualcomm and MediaTek’s next-generation SoCs.


There are rumors that TensorG4 will be mass-produced using TSMC’s 4nm process, but Revegnus believes that Google will stick with Samsung again. First of all, Google’s chip orders are limited because its Pixel series shipments are far inferior to Apple’s iPhone15 series. In addition, due to some issues, TensorG4 will be a slight upgrade of TensorG3, which is also an issue we have discussed before. The chip is codenamed "ZumaPro".

On TensorG5, Google may intend to abandon Samsung's foundry and chipset design and focus entirely on custom solutions, possibly leveraging TSMC's N3E process. It is rumored that in 2024, both Snapdragon 8 Gen4 and Dimensity 9400 will be produced using TSMC's 3nm process, which is a whole generation ahead of TensorG5. It is said that TensorG5 will be available as early as 2025. You know, in addition to customizing the CPU, it is rumored that Google will also develop an internal GPU for TensorG5 to improve its overall performance.

This means that the first custom SoC solutions will be launched on the Pixel 10 and Pixel 10 Pro, while the Pixel 9 and Pixel 9 Pro will be regarded as minor upgrades from their predecessors, at least when talking about chip generations. Much of the information Revegnus talked about has appeared in the past, but which TSMC manufacturing process Google is leaning towards has not yet been confirmed. Considering that N3E is said to have a higher yield rate and a more suitable price than N3B, it makes sense for other companies to wait and place orders.

Hopefully, Google will end its embarrassment of constantly launching slower Tensor chipsets and focus on products that not only surpass the competition in raw CPU and GPU performance, but also in energy efficiency and AI capabilities.