Google Inclines Towards Taiwanese Manufacturers For Next-Gen Tensor & AI Chips, Samsung Takes a Step Back

wccftech.com:

Google has reportedly taken an "unexpected" shift in its semiconductor sourcing strategy for next-gen Tensor & AI chips, as the firm turns its eyes towards Taiwanese manufacturers, distancing itself from Samsung.

The report comes from Taiwan Economic Daily, which discloses that Google has decided to look for an alternative semiconductor supplier for its "Tensor" lineup of SoC, and has approached Taiwanese firms, with names such as KYEC already gaining a portion of Google's orders. It is said that Google has already made investments in KYEC's facilities in Taiwan to aid the manufacturing process along with ensuring a "consistent" supply. Google is expected to initiate process testing in the middle of the year.

Previously, Samsung Foundry was the sole supplier of Google's Tensor chips, which made the lineup stand out from competitors like Qualcomm. However, things have taken a back step the past year, with reports suggesting that Google might switch to TSMC for next-gen Tensor SoCs since it is believed that Google has been held back by Samsung Foundry since not only are Tensor chips similar to Samsung's Exynos in terms of performance and power efficiencies, but both of them employ the same design as well, which means that Google has no chance to stand out, if they were to in this department.

Apart from Tensor chips, Google is also expected to utilize KYEC for their own AI chips as well, since the firm wants to mark a dominant entry in Taiwan's semiconductor industry, and instead of going the conventional route, which is opting for TSMC, Google has decided to do something unique by choosing a supplier relatively newer in the markets. Not only could this provide Google an upper hand in terms of improving chip designs and performance, but this can prove to be a turning point for KYEC, as Google has entrusted them with a responsibility that they can well capitalize on.

For now, we shouldn't expect much changes within the Tensor lineup, at least with the upcoming model,