Google adds TPU 8tx and 8i Max as AI chip business nears 100 customers, Forbes reports

Forbes reported on November 10, 2026, that Google had added TPU 8tx and 8i Max and that its AI chip business was nearing 100 customers.

The reported approach toward 100 customers is an indicator of the reach of Google’s AI chip business. However, customer numbers alone do not confirm adoption for physical AI training or real-world robot operations, so the expansion of the TPU business should be assessed separately from robot deployment.

The report focuses on the addition of TPU products and the business-scale indicator that the number of customers for Google’s AI chip business is nearing 100. Customer numbers offer a clue to the reach of the chips, but do not directly reflect purchase volumes, continued use, actual applications or performance. The specifications and supply scale of TPU 8tx and 8i Max also cannot be confirmed from the information provided in the report, and the product names and customer count alone are not enough to assess how they differ from existing products or how competitive they are. The connection to physical AI also needs to be considered separately. Google DeepMind announces virtual world generation AI “Genie 4,” with support for audio and conversation discussed Google DeepMind’s AI for generating virtual worlds and noted that it could be used for robot training, while also reporting that objects’ appearance or position could change during long generation runs. That is an issue for research and training in virtual environments; it does not indicate adoption of the TPUs reported on here or real-world robot operating results. Expansion of a computing chip business and practical deployment of robots that perform tasks in the field should be viewed as separate stages. Related coverage from this newspaper includes Business Journal’s article on xAI’s “Colossus”, which covers the development of AI computing infrastructure, and stockanalysis.com’s publication of the transcript of Arm’s Q2 2027 earnings call, which distinguishes the publication of a call transcript from specific remarks and figures. As with those cases, the information in this report does not justify directly linking expanded computing resources or business reach to robots’ processing capabilities or deployment results. Key points to verify include the workloads, performance and power requirements of both products; how the customer count is calculated; and whether the chips are used for robot training or inference. Separate public information showing real-world operation and safety results would be needed.