During the I/O 2022 keynote, Google didn’t reveal too many details related to the Pixel 7/7 Pro smartphones. But one thing is certain, the new flagship will use the second-generation Tensor SoC. Typically, manufacturers are tight-lipped about important details ahead of an official release. However, a recent new report claims that Google will once again choose Samsung’s 4nm process for mass production of next-generation chips.
Even though Qualcomm previously switched to the same-generation process of TSMC with better reputation, after the first-generation Tensor SoC chose Samsung’s 5nm process for manufacturing, Google chose Samsung’s 4nm process for its second-generation custom chips. In addition to technical upgrades, Korean media also expects the second-generation Google Tensor SoC to start mass production in June.
As for when the Pixel 7 / 7 Pro will be officially released, the Korean media did not mention it further. For reference, Google released the Pixel 6/6 Pro in October 2021, so a new generation of models is also expected to arrive at the same time this year. There may be a number of factors involved as to why Samsung is the sole foundry for custom chips for the Pixel 7.
The reason why Qualcomm chose Samsung was that a report said the deal was more cost-effective, while TSMC may be asking for a higher price. It now appears that Samsung may have made an equally attractive offer to Google to completely disrupt the possibility of a switch to a competitor. On the other hand, as TSMC’s largest customer, Apple has won most of the company’s production capacity, while Google, which is relatively small, naturally lacks the cost of negotiating prices and competing for production capacity.
It should be pointed out that Qualcomm has experienced Samsung’s 4nm overheating and performance issues on the Snapdragon 8 Gen 1. As for whether Google’s second-generation Tensor custom chips will repeat the same mistakes, it remains to be seen. Of course, even if it can’t keep up with the competition in hardware, Google is still pretty good at building confidence with optimizations in software, such as machine learning, imaging algorithms, speech recognition, and more.