Google has introduced two new artificial intelligence chips as it expands its competition with Nvidia in the growing AI hardware market.
Designed to protect sensitive data during processing and handle AI applications with larger computing demands, the TPU 8tx and TPU 8i Max are expected to become available in 2027.
The announcement brings Google’s total to four new TPU systems in 2026, a considerable increase from its previous release schedule as demand for AI computing continues to grow.
Why It Matters: Google’s growing competition with Nvidia could change the cost of running enterprise AI as organizations move toward larger deployments. Specialized processors may offer a more economical way to support certain workloads, although adapting existing applications to different hardware can require additional investment. With computing demand continuing to grow, infrastructure costs could play a greater role in determining which AI projects are financially viable.
- Protecting Sensitive Data: The TPU 8tx is designed for government agencies, financial institutions and other organizations with strict data protection requirements. By pairing Google’s chip technology with AMD processors, it keeps sensitive information encrypted during processing, reducing the risk of data exposure in memory. Google says the new design delivers up to twice the speed of its previous version with virtually no performance trade-off.
- Supporting Larger AI Workloads: Combining two chips into a single package, the TPU 8i Max provides additional memory and processing power for applications that handle large amounts of information. Its design supports longer context windows, allowing AI models to process more information within a single request. Potential uses include coding assistants, mathematical reasoning and drug discovery, where applications often work with extensive datasets and technical information.
- Expanding Its Chip Lineup: Google introduced two other TPU systems earlier this year, bringing its 2026 total to four. Previously, the company released a new processor approximately every two years. The latest additions are modified versions of existing designs, allowing Google to develop specialized chips for different customer needs without creating entirely new processors.
- Reaching More Cloud Customers: Major technology companies, including Apple, Meta and Oracle, already use Google’s TPU technology, while Anthropic has secured rights to purchase significant capacity in the coming years. Google is also expanding access through a partnership with Oracle that will bring TPUs to Oracle Cloud Infrastructure in 2027. This will mark the first time the processors are available through a major cloud provider outside Google Cloud, giving enterprises another option for accessing the hardware.
- Keeping Up With Demand: Google’s existing TPU inventory is already committed to customers, while shortages of memory and manufacturing capacity have made it difficult to increase production. Industry estimates project TPU sales could rise from approximately $12 billion in 2026 to more than $135 billion by 2028, excluding chips rented through Google Cloud.
Google’s chip business is on a faster track than Nvidia’s own AI boom – Yahoo!Finance
Google unveils new AI chip variants as cloud demand surges – Reuters
