Google Divides TPUv8 Strategy Into Two Chips for Broadcom Training and MediaTek Inference Responsibilities

Google is set to launch two innovative chips under the TPUv8 series, aimed at enhancing both AI training and inference tasks. While early reports hinted at a third chip, the main focus is now clearly on two distinct models tailored for specific functionalities.

Anticipation Grows for Google’s TPUv8 Lineup, Set to Focus on AI Inference & Training while Strengthening Semiconductor Supply Chains

According to sources, Google is actively developing a new generation of Tensor Processing Units (TPUs) expected to debut at the upcoming Google Cloud Next event from April 22–24, 2026. This new lineup, replacing the existing TPUv7 “Ironwood”series, will consist of two major variants: the TPUv8t (codenamed “Sunfish”) and the TPUv8i (codenamed “Zebrafish”).

Google’s next-generation TPU v8 chips are expected to feature prominently at its Google Cloud Next event this week (April 22–24, 2026), according to media and supply-chain reports. The v8 series includes two main variants: • TPUv8t ‘Sunfish’ — for Training: the high-performance…

— Dan Nystedt (@dnystedt) April 20, 2026

The focus of these new TPU variants is straightforward. The TPUv8i will serve as a cost-effective inference accelerator, while the TPUv8t will be engineered for high-performance training tasks. Notably, the development of the TPUv8i will be undertaken by MediaTek, and the TPUv8t will be crafted by Broadcom, showcasing a robust collaboration between leading semiconductor firms.

Interestingly, Marvell, a potential partner for Google’s TPU series development, does not seem to play a role in this particular project. It’s plausible that any future collaborations could lead to custom TPUs or advanced post-TPUv8 solutions.

Google Axion is the company's first fully custom ARM-based chipset

Both the TPUv8 series chips will be closely integrated with Google’s Axion Arm CPUs, which utilize the Neoverse N3 Armv9.2 core architecture. Since their deployment in 2024, Axion processors have positioned themselves as the backbone for this upcoming TPU generation, ensuring optimized performance in AI applications.

Google’s latest v8 architecture Tensor Processing Unit (TPU) is expected to be unveiled this week, which will not only boost the semiconductor and assembly supply chain markets, but also create a series of upgrade opportunities for peripheral components, including OCS all-optical switches, liquid cooling, power supplies, and optical communications companies.

via UDN

The imminent launch of Google’s TPUv8 chips is projected to have significant repercussions for the semiconductor market, potentially leading to tighter supply chains as demand surges to meet the needs of Google’s expansive server infrastructure and AI frameworks. This development signals an exciting time for advancements in AI technology and the broader semiconductor landscape.

Source & Images