🔍 Read the full analysis: Why A Second Source For Nvidia Matters As Google Adds AI Chips on ThorstenMeyerAI.com
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
TL;DR
Google announced two additions to its eighth-generation TPU lineup: TPU 8tx for confidential inference and TPU 8i Max for workloads with very long inputs. The systems are expected in 2027, and their performance figures are vendor claims; prices, cloud regions and independent test results have not been published.
Google announced two new systems for its eighth-generation tensor processing units on November 9: TPU 8tx, designed for confidential computing, and TPU 8i Max, intended for AI workloads with very long inputs. The additions give cloud customers more specialized options alongside Nvidia GPUs, but both are due in 2027 and have not been independently tested.
The new systems build on TPU 8t, Google’s training processor, and TPU 8i, designed to run trained models, or perform inference. Google says TPU 8tx uses the same silicon as TPU 8t, paired with fifth-generation AMD EPYC processors, and is aimed at regulated businesses and government customers. The company says it protects data and model weights while they are being processed, and claims up to two times lower latency than TPU 8t with a near-zero performance penalty.
TPU 8i Max combines two TPU 8i chips in one package. Google says it doubles compute, on-chip SRAM and high-bandwidth memory bandwidth compared with TPU 8i, while fitting within the existing rack footprint. The company describes it as suitable for tasks involving large context windows, such as processing an entire codebase or hours of video. Google has not published absolute specifications for either new system.
These announcements follow Google’s April 22 introduction of TPU 8t and TPU 8i. Google said this was the first year it had introduced four TPU systems in a single year. Its TPU product page lists all four as “Coming soon.” Amin Vahdat, Google’s chief technologist for AI and infrastructure, told Bloomberg the new variants are expected in 2027. Google has published no prices or availability by cloud region.
Why a second source for Nvidia matters as Google adds AI chips
Google’s new TPUs offer specialized options for confidential inference and very long inputs. They are expected in 2027, with pricing and independent results still unavailable.
Two systems, distinct workloads
The additions build on Google’s existing TPU 8t and TPU 8i systems.
TPU 8tx
Designed to protect data and model weights during processing for regulated businesses and government customers.
TPU 8i Max
Combines two TPU 8i chips in one package for workloads such as processing a codebase or hours of video.
What buyers could gain
A broader supplier choice may matter when AI computing capacity is constrained.
- Another source of AI computing capacity
- More options for sensitive or long-input workloads
- Potentially stronger negotiating position
- Prices and cloud regions
- Performance against Nvidia GPUs
- Effort needed to move existing software
Claims still need validation
Google describes the systems’ capabilities, but has not published absolute specifications for either variant.
Watch for pricing, launch regions, specifications and customer access terms.
Independent benchmarks and customer experience can show whether the claimed gains translate in practice.
Oracle has said it plans to offer Google TPUs through its cloud in calendar 2027.
More Options in AI Computing
For companies buying AI capacity, Google’s expanded lineup could provide another source of computing supply in a market where demand for Nvidia processors remains high. Google says nearly 100 customers use its TPUs, and Alphabet has described an environment constrained by supply. Oracle has also said it will offer Google TPUs through its cloud starting in calendar 2027, potentially making them available beyond Google’s own cloud.
A second supplier can give buyers more choice when capacity is scarce and may improve their negotiating position. But the announcement does not establish that the new TPUs will be cheaper or outperform Nvidia products. Google has released no prices or independent comparisons for these systems. Moving workloads from GPUs to TPUs can also require engineering work, so availability alone may not make switching practical for every organization.
The specialized designs address distinct needs: confidential processing for organizations handling sensitive information, and additional memory capacity for models or applications using long inputs. Whether those features translate into better results for customers will depend on real-world performance, cost, access and software support—not just Google’s specifications.
Products Worth Considering
Logitech MX Master 4 Ergonomic Wireless Mouse with Haptic Feedback, Graphite
Amazon Kindle Scribe (3rd Generation), 64GB, 11-inch Display, Graphite, with Premium Pen
Amazon Kindle Paperwhite Signature Edition (12th Generation, 2024), 32 GB, Metallic Raspberry
As an affiliate, we earn on qualifying purchases.
Google’s TPU Lineup Splits by Workload
Google introduced TPU 8t for training and TPU 8i for inference in April 2026. Its published figures for those base systems include 216 gigabytes of high-bandwidth memory on TPU 8t and 288 gigabytes on TPU 8i. Google says TPU 8i has more on-chip memory and memory bandwidth, a configuration intended to help with inference workloads that must move substantial amounts of data. These figures describe the base systems, not the newly announced variants.
Google’s April performance comparisons are also company claims: it reported up to 2.7 times better performance per dollar for TPU 8t than its previous generation for large-scale training, and up to an 80% improvement for TPU 8i in inference. They are not independent tests of TPU 8tx or TPU 8i Max. Google has also been working on software access, including TorchTPU for running PyTorch code and Hugging Face models, and support for open-source inference engines vLLM and SGLang.
Rather than presenting the two additions as wholly new chip designs, Google describes them as systems reconfigured around existing silicon. Vahdat told Bloomberg the variants were developed in response to customer requests. That approach can let a supplier tailor packaging, memory, networking or security without waiting for a new processor design, though the specific implementation and performance still need external validation.
“This is the first time we have introduced four TPU systems in a single year.”
— Google, in its November 9 announcement
Pricing, Tests and Access Remain Unknown
The systems are not yet available, and Google has not announced prices, launch regions or specific customer access terms. The expected 2027 timing comes from Vahdat’s comments to Bloomberg; the company has not provided a more precise launch date in the supplied material.
All performance and security figures for TPU 8tx and TPU 8i Max come from Google. No independent benchmarks or external security evaluations are available in the announcement material. Google has published ratios for the new systems but not their absolute specifications, making comparisons with other processors difficult. It is also unclear how much work customers will need to move existing GPU-based software to the TPU platform, despite Google’s stated software-support efforts.
Watch for 2027 Launch Details
The next milestones are Google’s release of pricing, regions, specifications and customer access details, followed by the systems’ expected availability in 2027. Independent performance testing and customer experience will help determine whether the claimed gains hold for actual workloads.
Oracle’s planned calendar-year 2027 offering could provide another route to Google TPU capacity. Buyers will be able to compare that option with Google’s own cloud and other available processors once commercial terms and capacity are clearer. Until then, the announcement signals a broader product range, not a currently available alternative.
Key Questions
What did Google announce?
Google announced TPU 8tx, a confidential-computing system, and TPU 8i Max, a system designed for workloads with very long inputs. Both extend its eighth-generation TPU lineup.
When will the new TPUs be available?
Google’s Amin Vahdat told Bloomberg the two new variants are expected in 2027. Exact launch dates and cloud regions have not been announced.
Are the performance claims independently verified?
No independent benchmarks for TPU 8tx or TPU 8i Max were provided. Their reported performance figures are Google’s claims, and the new systems have not yet been released.
How could this affect Nvidia customers?
The expanded TPU range could give some buyers another supplier to evaluate, especially for confidential inference or long-context workloads. It does not yet show that the systems are cheaper or faster than Nvidia GPUs, and switching platforms can require engineering work.
Source: ThorstenMeyerAI.com
