MOUNTAIN VIEW, United States, November 11, 2026, Google has introduced two new versions of its in house Tensor Processing Units (TPUs), expanding its artificial intelligence hardware portfolio as competition intensifies among technology companies seeking to provide computing infrastructure for advanced AI applications.
The new variants are designed for different workloads and customer requirements. One is aimed at government agencies and regulated industries that need additional controls when processing sensitive information, while the other is designed for demanding AI applications that require substantial computing capacity and memory.
The announcement, reported by Reuters on November 10, comes as Google seeks to strengthen its position in the AI infrastructure market and attract more enterprise customers to its cloud computing services.
Two variants target different computing needs
The first new version, identified as the 8tx, is designed for government agencies and organisations operating in regulated industries. It combines Google’s TPU technology with AMD processors and hardware level encryption, with the aim of supporting workloads involving sensitive information.
Google said the 8tx can deliver up to twice the speed of its predecessor without a significant performance trade off. The design reflects the importance of data protection and computing performance for customers handling confidential information.
The second variant, the 8i Max, is designed for AI workloads that require large amounts of information to be processed at once. Its architecture combines two chips in one package to provide additional memory and computing capacity.
Potential applications include coding assistants, mathematical reasoning and drug discovery, where AI systems may need to process complex information and carry out extended computing tasks.
Google expands its AI infrastructure strategy
The introduction of additional TPU variants is part of Google’s broader strategy to compete in the rapidly expanding AI hardware market. The company develops its own chips and offers access to them through Google Cloud, giving enterprise customers another option for running demanding AI workloads.
The market has traditionally relied heavily on specialised processors supplied by external chipmakers. However, major cloud providers have increasingly invested in custom hardware to manage costs, improve performance and reduce dependence on external suppliers.
Google’s approach combines chip development with its cloud infrastructure, software services and AI models. This integrated strategy could help the company attract customers looking for computing resources across multiple stages of AI development.
Cloud revenue supports the expansion
Google’s AI hardware announcement follows strong growth in its cloud business. Reuters reported that Google Cloud revenue more than doubled to US$30.71 billion in the September quarter, while chip sales made a meaningful contribution to the company’s results.
The expansion of cloud computing has increased demand for specialised hardware capable of supporting AI model training, inference and other data intensive applications. Companies are investing in processors, networking systems and data centre capacity to meet these requirements.
Competition in AI chips intensifies
Google’s latest hardware developments place it in closer competition with established AI chip suppliers and other technology companies developing custom processors. The ability to offer competitive performance, reliability and access to computing capacity will be important in attracting enterprise customers.
The two new variants also demonstrate that AI hardware requirements are becoming more specialised. Government and regulated organisations may prioritise security, while research and software development applications may require greater memory and processing capacity.
Google’s ability to combine custom chips with cloud services will be a key factor in determining how effectively it can expand its position in the AI infrastructure market.
This website uses cookies.