TL;DR
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
Google announced two eighth-generation TPU systems on November 9: TPU 8tx, designed for confidential inference, and TPU 8i Max, designed for long-context workloads. Google says both will be available in 2027; pricing, regions and independent performance results have not been published.
Google announced two new eighth-generation Tensor Processing Unit systems on November 9, adding confidential inference and long-context workloads to a lineup that now has four systems introduced this year. The new TPU 8tx and TPU 8i Max are expected in 2027, according to Google AI and infrastructure chief Amin Vahdat; neither has a published price or independent benchmark.
TPU 8tx is designed for confidential inference, where organizations run AI models on sensitive data. Google says it protects data and model weights while they are actively being processed, and claims up to 2x lower latency than TPU 8t with a near-zero performance penalty. Google describes it as using the same silicon as TPU 8t, paired with fifth-generation AMD EPYC processors. These performance and security descriptions come from Google; independent testing has not been published.
TPU 8i Max is aimed at workloads that need to handle very long context windows, such as analyzing a large codebase or extended video. Google says the system combines two low-latency TPU 8i chips on one package, doubling compute, on-chip SRAM and HBM bandwidth compared with TPU 8i while fitting within the existing rack footprint. Those are vendor specifications, not independently verified results.
The two additions build on Google’s April 22 launch of TPU 8t for model training and TPU 8i for inference. The company said this is the first year it has introduced four TPU systems. Vahdat told Bloomberg the new variants are reconfigured systems rather than entirely new silicon designs, reflecting customer requests. Google’s product page lists all four eighth-generation systems as “Coming soon.”
Google expands its TPU lineup to four systems
Two new eighth-generation TPUs target confidential inference and very long-context workloads. Their performance and availability remain unverified.
Two systems, distinct workloads
The additions give businesses more options for matching TPU hardware to specific AI tasks.
TPU 8tx
Designed to protect sensitive data and model weights while they are processed.
TPU 8i Max
Targets very long inputs, such as large codebases or extended video.
How the lineup has grown
Introduced for model training.
Introduced for inference.
Add confidential inference and long-context workloads.
What buyers still need to know
- Google’s intended workloads and hardware descriptions
- Expected availability in 2027
- Prices, regions and exact launch dates
- Independent performance benchmarks
More Options for AI Workloads
Google is expanding its offer to businesses that need AI computing, with systems tailored to confidential data processing and workloads involving unusually long inputs. If the systems deliver as described, customers may be able to match hardware more closely to a task rather than use one accelerator configuration for every workload.
The expansion also matters to organizations weighing alternatives to Nvidia GPUs. Google says nearly 100 customers use TPUs, and Alphabet has described a supply-constrained environment. Oracle plans to offer Google TPUs through its cloud starting in calendar year 2027, according to the source material. Wider access could give some buyers another supplier to consider, though moving workloads from GPUs to TPUs can require engineering work. The new systems’ availability and actual performance will determine whether they change purchasing decisions.
Products Worth Considering
Logitech MX Master 4 for Mac Wireless Ergonomic Mouse, White Silver
Samsung Galaxy Tab S7 11-inch Android Tablet 128GB Wi‑Fi with S Pen Mystic Black
Jackery Explorer 1000 v2 Portable Power Station, 1070Wh, 1500W
As an affiliate, we earn on qualifying purchases.
From Two TPU Systems to Four
Google introduced its eighth-generation TPU family at Cloud Next on April 22, 2026, dividing it into TPU 8t for training and TPU 8i for inference. Google’s April materials list TPU 8t with 216 GB of high-bandwidth memory, 128 MB of on-chip SRAM, 12.6 petaflops of peak FP4 compute and 6,528 GB/s of memory bandwidth. TPU 8i is listed with 288 GB of HBM, 384 MB of SRAM, 10.1 petaflops and 8,601 GB/s of bandwidth.
Google also claimed up to a 2.7x performance-per-dollar improvement for TPU 8t over its prior Ironwood generation in large-scale training, and up to an 80% improvement for TPU 8i in inference. Those are Google’s comparisons, not independent benchmarks of the new systems. The November announcement extends the family through system-level changes to packaging, memory, networking and security. Google has also pointed to software support including TorchTPU, vLLM and SGLang as part of making its hardware usable by developers.
“This is the first time we have introduced four TPU systems in a single year.”
— Google, in its November 9 announcement
Pricing and Performance Remain Unpublished
Google has not announced prices, launch regions or specific availability dates for TPU 8tx or TPU 8i Max. Its TPU page lists the four eighth-generation systems as “Coming soon,” while Vahdat told Bloomberg the two new variants are expected in 2027.
There are also no independent benchmarks for the new systems. Google’s latency, compute, memory and bandwidth figures remain company claims, and the source material does not provide test conditions or a direct comparison from an outside evaluator. It is not yet clear which customers will get access first or how the confidential-computing protections will be independently assessed.
Availability Expected in 2027
The next expected milestone is the planned 2027 availability of TPU 8tx and TPU 8i Max. Google has not provided a more specific launch schedule, pricing or cloud regions. Oracle’s planned TPU offering in calendar year 2027 could also broaden access beyond Google Cloud, though details on configurations and rollout were not included in the source material.
Once the systems become available, buyers and independent evaluators will be able to assess performance, cost and software compatibility under real workloads. Until then, purchasing comparisons will rely on Google’s published claims and results for the earlier TPU generation.
Key Questions
What are TPU 8tx and TPU 8i Max designed to do?
TPU 8tx is designed for confidential inference on sensitive data. TPU 8i Max targets workloads that need very long context windows, such as processing large codebases or extended video.
When will the two new TPU systems be available?
Google’s Amin Vahdat told Bloomberg they are expected in 2027. Google has not announced exact dates or launch regions.
Are Google’s performance claims independently verified?
No independent benchmarks for TPU 8tx or TPU 8i Max are included in the source material. The stated latency and hardware comparisons are Google’s claims.
How many eighth-generation TPU systems has Google announced?
There are four systems: TPU 8t for training, TPU 8i for inference, TPU 8tx for confidential inference and TPU 8i Max for long-context workloads.
Source: Thorsten Meyer AI
