Alphabet-owned company "Google" announced the launch of two new systems within the eighth generation of AI accelerators, namely "TPU 8tx" and "TPU 8i Max", in a move aimed at enhancing computing capabilities, meeting the growing needs of institutions, and competing with "Nvidia" in the AI processor market.
According to "Bloomberg", the two systems represent specialized versions of Tensor Processing Units (TPUs) developed by "Google" to train and run AI models, targeting growing needs related to data security and processing massive amounts of information. The first system, "TPU 8tx", focuses on providing a secure computing environment that allows processing sensitive data while maintaining its encryption and isolation during operation, enabling government institutions and companies to train models and run AI agents without exposing their private information to risk.
For his part, Amin Vahdat, head of AI Infrastructure at "Google", clarified that the four systems were designed to meet different needs, noting that the company's direction toward developing specialized processors aligns with its vision for the future of computing, where the benefits of customizing processors according to the nature of tasks remain significant. He added that the second system, "TPU 8i Max", enhances the processing capabilities of large models and long contexts by integrating two "TPU 8i" chips into a single package, thereby doubling computational power, memory capacity, and bandwidth, contributing to accelerating complex tasks and reducing data transfer bottlenecks.
Data indicated that with the launch of the two systems, the number of eighth-generation systems at "Google" rises to four, including "TPU 8t" dedicated to model training, "TPU 8i" for running and inference, along with the two new systems dedicated to secure computing and enhancing processing capabilities. These expansions come amid the increasing use of AI agents in various tasks, including software development, analysis of long video clips, processing large amounts of information, and drug discovery.
"Google" seeks to expand its customer base for these processors beyond internal use, after developing them over a decade ago to meet its growing computing needs, and began making them available to customers via "Google Cloud" services. The customer base for these processors includes approximately 100 entities, including technology companies, financial institutions, and universities, while estimates from "Bloomberg Intelligence" suggest that sales of Tensor Processing Units may exceed $135 billion by 2028, compared to an estimated $12 billion in 2026.
These estimates include sales of related hardware and software, and do not account for revenue from customers using these processors via "Google Cloud" services, according to the institution's data. This step reflects "Google"'s efforts to strengthen its position in the AI processor market, dominated by "Nvidia", by offering specialized solutions that combine processing efficiency, data security, and the ability to handle growing volumes of information.
Source: Syrian Arab News Agency (SANA)
Comments (0)