TL;DR
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
Google announced two new systems based on its eighth-generation TPUs: TPU 8tx for confidential inference and TPU 8i Max for workloads with very long inputs. Google says both are planned for 2027, but neither has independent benchmarks, published prices or availability details.
Google announced two new AI chip systems on November 9, adding confidential inference and support for very long inputs to its eighth-generation Tensor Processing Unit lineup. The systems, called TPU 8tx and TPU 8i Max, are planned for 2027 and give cloud customers another potential option alongside Nvidia-based computing, though pricing and independent performance results are not yet available.
Google describes TPU 8tx as a system for secure, confidential inference aimed at regulated businesses and government users. The company says it protects sensitive data and model weights while they are being used, and claims up to two times lower latency than TPU 8t with a near-zero performance penalty. Those figures and security claims come from Google; external testing has not been reported in the source material.
TPU 8i Max targets workloads with very long context windows, such as AI agents processing a codebase or extended video. Google says it combines two TPU 8i chips in one package, doubling compute, on-chip SRAM and high-bandwidth memory bandwidth compared with TPU 8i, within the existing rack footprint. The company says both new systems use existing silicon in reconfigured systems rather than entirely new chip designs.
Google’s AI and infrastructure chief, Amin Vahdat, told Bloomberg that the two variants are expected to be available in 2027. Google’s TPU product page lists the eighth-generation systems as “Coming soon.” The company has not published prices or deployment regions for the new variants.
Google’s latest TPUs and Nvidia competition
Google announced two systems for confidential inference and very long inputs. They add another potential cloud option, but buyers cannot yet compare price or independently tested performance.
Two systems, distinct workloads
Google says both use existing silicon in reconfigured systems.
TPU 8tx
Designed to protect sensitive data and model weights during use, for regulated businesses and government users.
TPU 8i Max
Targets tasks such as processing a codebase or extended video, combining two TPU 8i chips in one package.
A broader TPU lineup
The new variants join systems Google introduced earlier in its eighth TPU generation.
For training AI models.
For inference, or running trained models.
Adds confidential inference.
Adds capacity for very long inputs.
What cloud customers gain
More accelerator capacity could give customers another option and more negotiating leverage when supply is tight.
- Another AI computing option alongside Nvidia-based systems.
- Oracle plans to offer Google TPUs in its cloud from calendar year 2027.
- Moving workloads from Nvidia GPUs may require engineering changes.
- Price, access, software compatibility and independent performance are not yet established.
Evidence buyers still need
Google’s performance and security statements have not been independently verified in the supplied source material.
- Independent benchmarks for both new systems.
- Published prices, launch dates and deployment regions.
- Details on customer access, including whether Oracle will offer these variants.
- Information on how customers can verify protection during active use.
A Second Route to AI Capacity
The announcement broadens Google’s effort to sell AI computing as an alternative to relying on Nvidia GPUs. For cloud customers, another source of accelerator capacity could provide more choice and negotiating leverage, particularly when supply is tight. Google says nearly 100 customers use its TPUs, while Alphabet has described its operating environment as supply-constrained.
That does not mean customers can switch easily. Moving AI workloads from Nvidia GPUs to TPUs can require engineering changes, and the source material provides no independent comparison of the new systems. The practical value will depend on access, price, workload performance and software compatibility once the systems are available.
Google is also extending TPU access beyond its own cloud: Oracle plans to offer Google TPUs in its cloud from calendar year 2027, according to the source material. That could make the chips available through another provider, although it does not establish where or at what scale the newly announced variants will be offered.
Products Worth Considering
Apple MacBook Air 13.6-inch Laptop with M5 Chip, 16GB Memory, 512GB SSD, Midnight
Bose TV Speaker Compact Soundbar with Bluetooth and HDMI-ARC, Black
LG UltraGear 27-inch QHD OLED Gaming Monitor, 240Hz
As an affiliate, we earn on qualifying purchases.
From Two Systems to Four
Google introduced the eighth TPU generation at its Cloud Next conference on April 22, 2026, with two systems: TPU 8t for training AI models and TPU 8i for inference, or running trained models. The November announcement adds 8tx and 8i Max. Google said this was the first year it had introduced four TPU systems; Bloomberg reported that the chip unit previously released a new AI processor about every two years.
Google’s published specifications for the base systems show different priorities. TPU 8t is listed with 216 GB of high-bandwidth memory and 12.6 petaflops of peak FP4 compute. TPU 8i is listed with 288 GB of high-bandwidth memory, more on-chip SRAM and 10.1 petaflops of peak FP4 compute. These are company-published specifications, not independent test results.
Google also cites software support as part of its TPU strategy, including TorchTPU for PyTorch and Hugging Face models, and TPU backends for the open-source inference engines vLLM and SGLang. Its Cloud Pathways runtime, which Google DeepMind uses behind Gemini, is generally available on Google Kubernetes Engine, according to the supplied source material.
“This is the first time we have introduced four TPU systems in a single year.”
Performance and Rollout Remain Open
Independent benchmarks are not available for TPU 8tx or TPU 8i Max, and no prices or launch regions have been published. Google’s performance and security figures are vendor claims, so buyers cannot yet compare the new systems against Nvidia GPUs or other cloud accelerators using independent results.
The exact 2027 release timing and customer access are also unclear. Google has not said in the supplied material how many systems it will offer, which customers will get access first, or whether Oracle’s planned TPU service will include these two variants. The source material also does not detail how customers will verify the claimed protection of data and model weights during active use.
Availability Expected in 2027
The next milestone is Google’s expected 2027 availability for TPU 8tx and TPU 8i Max, as stated by Vahdat to Bloomberg. Google will need to publish deployment regions, pricing and access details for customers to evaluate the systems.
Independent workload tests and customer experience will help establish whether the claimed latency, memory and security features translate into practical gains. Oracle’s planned Google TPU offering in calendar year 2027 may also clarify how broadly customers can access the chips outside Google Cloud.
Key Questions
What did Google announce?
Google announced TPU 8tx for confidential inference and TPU 8i Max for workloads that need very long context windows. They expand the company’s eighth-generation TPU lineup.
When will the new TPU systems be available?
Google’s Amin Vahdat told Bloomberg the two variants are expected in 2027. Google has not published specific launch dates or regions.
Are the performance claims independently verified?
No independent benchmarks for TPU 8tx or TPU 8i Max appear in the supplied source material. The figures currently available are claims from Google.
How could the announcement affect Nvidia?
Google’s systems add another potential AI computing option for cloud customers. The announcement alone does not show how many customers will switch from Nvidia GPUs; price, availability, software needs and independent performance comparisons remain unknown.
Source: Thorsten Meyer AI
