Alphabet's Google is reportedly planning to collaborate with Taiwan's MediaTek on its upcoming Tensor Processing Units (TPUs), which are set to launch next year. Despite this new partnership, Google remains committed to its existing relationship with Broadcom, a company it has exclusively worked with on AI chip development in recent years.
Neither Google, MediaTek, nor Broadcom have responded to requests for comment regarding this development. Google, much like NVIDIA, designs its own AI server chips for internal research and cloud computing clients. This strategy helps Google reduce its reliance on NVIDIA, gaining a competitive edge in the AI race, even as competitors like Microsoft-backed OpenAI and Meta Platforms increase their demand for NVIDIA chips.
Last year, Google introduced its sixth-generation TPUs, aiming to provide an alternative to NVIDIA's highly popular processors for itself and its cloud clients. Google's choice to partner with MediaTek is partly due to the Taiwanese company's strong collaboration with TSMC and the lower per-chip costs compared to Broadcom.
According to research firm Omdia, Google's expenditure on TPUs last year was estimated between $6 billion and $9 billion, based on Broadcom's revenue targets for its AI semiconductor business during the same period.
