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 additions to its eighth-generation TPU line on November 9: TPU 8tx for confidential inference and TPU 8i Max for workloads with very long inputs. Google says both will be available in 2027, but has not published prices or independent performance results, so their effect on Nvidia’s position remains uncertain.
The new systems extend the TPU 8t and TPU 8i, which Google introduced at its Cloud Next conference on April 22. Google positions **TPU 8t** for training large models and **TPU 8i** for inference, the stage when a trained model generates responses. The company said it had introduced four TPU systems in one year, a faster release pace than its previous pattern, Bloomberg reported.
Google says **TPU 8tx** is designed for secure, confidential inference, aimed at regulated businesses and government customers. The company says it protects sensitive data and model weights while they are being used. It is built on the same silicon as TPU 8t and paired with fifth-generation AMD EPYC processors. Google claims up to **2x lower latency** than TPU 8t with a near-zero performance penalty; that figure has not been independently tested.
Google describes **TPU 8i Max** as two low-latency TPU 8i chips on one package. It says the system doubles TPU 8i’s compute, on-chip SRAM and high-bandwidth memory bandwidth within the same rack footprint, targeting applications that process large context windows, such as agents analyzing codebases or long video. Google’s TPU product page lists all four eighth-generation systems as “Coming soon.”
Could Google’s New AI Chips Reduce Reliance on Nvidia?
Google announced two specialized TPU systems for confidential inference and very long inputs. Their impact on Nvidia’s position remains uncertain until customers can test them.
Two chips, distinct workloads
The new systems extend Google’s TPU 8t training and TPU 8i inference lineup.
TPU 8tx
Designed to protect sensitive data and model weights during use, for regulated businesses and government customers.
TPU 8i Max
Combines two TPU 8i chips on one package for applications such as analyzing codebases or long video.
What could change for buyers
Another supplier option could affect availability, pricing and negotiating power.
- More choice alongside Nvidia GPUs
- Oracle plans to offer Google TPUs in its cloud in 2027
- Google cites TorchTPU, vLLM, SGLang and Cloud Pathways as software support
- Prices, launch regions and customer access plans
- Independent benchmarks and real workload performance
- Software adaptation and supply for the new systems
From two systems to four
Google introduced TPU 8t and TPU 8i in April 2026, then announced the two specialized variants on November 9.
Evidence buyers still need
The announcement describes potential alternatives, but does not establish a shift away from Nvidia.
Google has not published prices, launch regions or detailed access plans.
No independent benchmarks for either new variant were provided.
Software changes, supply and performance on real workloads remain open questions.
A Second Route to AI Compute
For companies buying substantial AI computing capacity, more supplier options could affect **availability, pricing and negotiating power**. Alphabet described its environment as constrained by supply, while Google says nearly 100 customers use TPUs. Oracle has also said it will offer Google TPUs in its cloud starting in calendar year 2027, potentially making them accessible beyond Google Cloud.
That does not establish that the new systems will displace Nvidia hardware. Buyers would need to weigh their performance and cost against alternatives, as well as the work required to adapt software and operations. Google points to **TorchTPU**, support for inference engines vLLM and SGLang, and Cloud Pathways on Google Kubernetes Engine as parts of its software offering. Those steps may ease adoption, but their practical effect for customers has not yet been demonstrated for these new variants.
The announcement also reflects a way to expand product choice without designing entirely new silicon for every workload. Google says changes to packaging, memory, networking and security can produce specialized systems more quickly than a new chip design. That approach may let it respond to customer needs sooner, though the systems still have to reach market and perform as claimed.
Popular right now
Breville Bambino Espresso Machine, Brushed Stainless Steel
Ninja Foodi Air Fryer 8 Qt DualZone 6-in-1 Grey DZ201
DREAME A3 AWD 1000 LiDAR Robot Lawn Mower
As an affiliate, we earn on qualifying purchases.
From Two TPUs to Four
Google introduced the base eighth-generation systems in April 2026. Its published specifications list TPU 8t with **216 GB of high-bandwidth memory**, 128 MB of SRAM, peak FP4 compute of 12.6 petaflops and memory bandwidth of 6,528 GB/s. TPU 8i has 288 GB of high-bandwidth memory, 384 MB of SRAM, 10.1 petaflops of peak FP4 compute and bandwidth of 8,601 GB/s.
The difference reflects Google’s stated division of tasks: TPU 8t is oriented toward model training, while TPU 8i puts more emphasis on memory capacity and bandwidth for inference. Google claims up to 2.7 times better performance per dollar for TPU 8t versus its prior-generation Ironwood system in large-scale training, and up to an 80% improvement for TPU 8i in inference. These are **Google’s comparisons**, not independent benchmark results, and they concern the base systems rather than the two variants announced in November.
Google’s AI chip chief, Amin Vahdat, told Bloomberg that TPU 8tx and TPU 8i Max are reconfigured systems rather than completely new silicon designs. The company says it developed the specialized versions in response to customer requests. It has not released list prices, launch regions or independent benchmarks for either new system.
““This is the first time we have introduced four TPU systems in a single year.””
— Google, in its November 9 announcement
Performance and Rollout Remain Open
Neither new TPU variant is available yet, and **independent benchmarks are not available**. Google has not disclosed prices, launch regions or detailed customer access plans. Its latency and capacity figures are vendor claims, so buyers cannot yet compare the new systems on a common, independently tested basis with Nvidia GPUs or other accelerators.
It is also unclear how easily customers will move existing workloads to the TPUs, what software adjustments they may need, or whether the systems will be available in sufficient quantities. The nearly 100-customer figure and supply constraints describe Google’s broader TPU business, not confirmed orders or delivery volumes for the new variants.
Availability Expected in 2027
Google’s Vahdat told Bloomberg the two new variants are expected to be available in **2027**. Google has not specified exact release dates. Oracle has said it plans to offer Google TPUs in its cloud during calendar year 2027, while Google’s product page currently lists the four eighth-generation systems as coming soon.
Customers and competitors will have more concrete evidence to assess once Google publishes pricing, availability and system-level benchmark results, and customers can test the products on their workloads. Until then, the announcement signals an expanded TPU lineup rather than a demonstrated shift away from Nvidia.
Key Questions
What did Google announce?
Google announced TPU 8tx, for confidential inference, and TPU 8i Max, for workloads with very long inputs. They extend the TPU 8t and TPU 8i systems introduced in April.
When will the new TPUs be available?
Google’s Amin Vahdat told Bloomberg the two variants are expected in 2027. Google has not announced exact dates or launch regions.
Could the new TPUs reduce reliance on Nvidia?
They may give some cloud customers another AI-compute option, but the announcement does not show that customers are switching from Nvidia. Prices, independent benchmarks, supply and the effort needed to adapt software remain unknown.
Has Google independently verified its performance claims?
No independent benchmarks for TPU 8tx or TPU 8i Max were provided in the source material. Google’s claims, including lower latency for TPU 8tx and doubled resources for TPU 8i Max, are vendor figures.
Source: Thorsten Meyer AI
