Google is partnering with Taiwan’s MediaTek to develop the next version of its Tensor Processing Unit (TPU), an AI chip set up for production next year.
Despite this new collaboration, Google has not ended its relationship with Broadcom, a chip designer that has worked with for several years. Broadcom employees have confirmed the ongoing partnership.
This shift coincides with a wider industry trend as companies like Amazon, Meta and Microsoft push to design their own AI chips. the goal? It reduces Nvidia’s reliance on GPUs and ASICs, which currently dominates the AI chip market.
MediaTek stood out against Google for two main reasons. It has reported a close relationship with chip manufacturing giant TSMC and a more competitive pricing compared to Broadcom.
Google’s investment in TPU is substantial. Research firm Omdia estimates it spent between $6 billion and $9 billion on TPU last year, based on Broadcom’s AI semiconductor revenue targets.
The partnership illustrates Google’s broader strategy to further strengthen AI hardware while reducing costs and reducing reliance on external suppliers such as NVIDIA.
Tech companies are building their own AI chips to reduce their dependence on Nvidia
With this partnership, Google is joining the growing list of high-tech giants working to reduce their dependence on NVIDIA. On February 5, 2025, Openai revealed that Openai is nearing completion of its first custom AI chip, with the aim of reducing its dependence on Nvidia. This was not a side hustle. It was a clear strategy to promote next-generation models without relying on Nvidia hardware. For a company with huge AI ambitions, this was a big move. If Openai doesn’t need Nvidia, who will do it?
“Openai is planning to reduce its reliance on Nvidia by creating its own AI chip. The company will complete its first custom chip design in the coming months and send it to Taiwan Semiconductor Manufacturing Co (TSMC) for production.”
On March 11, Meta began testing the first in-house AI chip designed for both training and inference, reducing the sky high costs associated with NVIDIA GPUs. Meta’s goal? Complete control over AI operations without passing billions to external suppliers. This is a calculated movement that is part of a larger shift. This is Hyperscalers that cut loose from Nvidia’s trajectory. In Meta, this tip is a ticket.
This shift is not limited to high-tech giants. Amazon is quietly moving forward with training chips supported by its investment in humanity. This is an AI company that already uses Google’s TPU for most reasoning. Startups like Celebras have also stepped up, with specialized chips offering power models such as Perplexity sonar and Mistral chatbots. This trend is heading towards efficiency rather than brute force, a threat to Nvidia’s premium margin. If Google can run Gemma 3 on one chip, why should others pay Nvidia’s hardware fleet?
Founded in 1997, MediaTek is a Taiwanese semiconductor company that designs chips for smartphones, smart devices and automotive systems. MediaTek, known for its Dimenity series, offers processors to brands such as Xiaomi, Oppo, and Vivo. The company is also working closely with Taiwan Semiconductor Manufacturing Company (TSMC) to produce chips. It has become a major player in the global chip market due to its focus on cost-effective solutions.
Source link
