🔍 Read the full analysis: Google’s New AI Chips Put More Competition Into The Nvidia Conversation 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 eighth-generation TPU systems on November 9: TPU 8tx, designed for confidential inference, and TPU 8i Max, built for workloads with very long inputs. Google says both will be available in 2027, but has not released prices or independent performance results. The variants broaden buyers’ potential alternatives to Nvidia GPUs, though availability and real-world comparisons remain open questions.
Google announced two new eighth-generation TPU systems on November 9, adding a confidential-computing option and a high-memory system for very long inputs to its AI chip lineup. The company says TPU 8tx and TPU 8i Max will be available in 2027, giving cloud customers another potential source of AI computing alongside Nvidia GPUs, though neither system is available to buy or independently tested.
The new systems extend the TPU 8 generation Google introduced in April. TPU 8tx is designed for confidential inference, or running models while protecting data and model weights during active use. Google is targeting regulated businesses and government customers. The system uses the same silicon as TPU 8t, paired with fifth-generation AMD EPYC processors; it is a new configuration, not a newly designed processor.
TPU 8i Max targets workloads with massive context windows, such as agents processing an entire codebase or long video. Google says it combines two low-latency TPU 8i chips in one package, doubling compute, on-chip SRAM and high-bandwidth memory bandwidth relative to TPU 8i. The company also says it fits within the existing rack footprint.
Google’s November announcement gives relative performance claims, not independently verified results. It says TPU 8tx offers up to twice the latency reduction compared with TPU 8t, with a near-zero performance penalty, while TPU 8i Max doubles several specifications. The company has not published absolute specifications for either new system, list prices or deployment regions. Its product page lists the eighth-generation TPUs as “Coming soon.”
Google’s New AI Chips Broaden the Nvidia Conversation
Google announced two specialized eighth-generation TPU systems. They could give cloud buyers more options, but availability, price and real-world performance remain unconfirmed.
Two systems, distinct needs
The new configurations target sensitive workloads and applications with very long inputs.
TPU 8tx
Designed to protect data and model weights during active use, including for regulated businesses and government customers.
TPU 8i Max
Combines two TPU 8i chips for workloads such as processing an entire codebase or long video.
A wider TPU lineup
Google’s eighth-generation family now has four systems, each aimed at different AI workloads.
| System | Designed for |
|---|---|
| TPU 8t | Training large models |
| TPU 8i | Inference |
| TPU 8tx | Confidential inference |
| TPU 8i Max | Very long inputs |
What buyers need to weigh
A new option matters only if customers can access it and run their workloads effectively.
- More choice alongside Nvidia GPUs
- Potentially more negotiating leverage
- Oracle is expected to offer Google TPUs through its cloud in 2027
- No prices or deployment regions announced
- No independent benchmarks for the new systems
- Switching workloads may require engineering work
Claims still need real-world checks
Google has shared relative performance claims, but not absolute specifications for either new system.
Watch for final specifications, pricing and cloud regions.
Google expects both variants to be available in 2027.
Independent tests and customer reports can show how they perform and what switching requires.
More Options for AI Buyers
The announcement matters because access to AI computing is a strategic issue for companies building or operating large models. Nvidia remains the dominant supplier, and demand for its processors has exceeded available supply, according to Bloomberg’s reporting cited in the source material. A credible alternative could give cloud providers and customers more choice and negotiating leverage.
But an announced product is not yet a practical substitute. Buyers will need to weigh actual availability, cost, performance and software compatibility. Moving workloads from GPUs to TPUs can require engineering work, and results depend on how well applications run on Google’s platform. Pricing and independent benchmarks are not yet available, so the announcement alone does not establish how the systems compare with Nvidia hardware.
Google says nearly 100 customers use its TPUs, while Alphabet has described its AI infrastructure as operating in a supply-constrained environment. Oracle is also expected to offer Google TPUs through its cloud starting in calendar year 2027, according to the source material. If delivered, access through another cloud provider could broaden how customers obtain TPU capacity, though the scale and terms of that offering have not been disclosed.
Products Worth Considering
Apple AirPods Pro (2nd Generation) Wireless Earbuds with USB‑C Charging, Up to 2X More Active Noise Cancelling, Transparency Mode, Adaptive
Apple AirPods Pro (1st Generation) with MagSafe Charging Case
Bose Lifestyle Ultra Compact Wireless Smart Home Speaker Bundle, Black
As an affiliate, we earn on qualifying purchases.
From Two Base TPUs to Four Systems
Google introduced the base eighth-generation systems at its Cloud Next conference on April 22, 2026. TPU 8t is aimed at training large models, while TPU 8i is designed for inference, the process of generating answers from a trained model. Google says TPU 8t can be deployed in superpods of 9,600 chips.
For those base systems, Google published specifications including 216 gigabytes of high-bandwidth memory for TPU 8t and 288 gigabytes for TPU 8i. TPU 8i also has more on-chip SRAM and memory bandwidth, reflecting a design emphasis on moving data quickly for inference. Google claims performance-per-dollar improvements over the prior Ironwood generation, but those figures are the company’s own and are not independent comparisons.
Google’s November post adds the 8tx and 8i Max configurations. Google chief technologist for AI and infrastructure Amin Vahdat told Bloomberg they are not completely new silicon designs, but systems reconfigured around customer needs. Google has also pointed to software work, including TorchTPU and support for open-source inference engines, as part of its effort to make TPUs more usable. The source material says TorchTPU was open-sourced the previous month.
“This is the first time we have introduced four TPU systems in a single year.”
— Google, in its November 9 announcement
Availability and Performance Still Open
Neither new system is available yet. Vahdat told Bloomberg that the variants are expected in 2027, but Google has not announced exact launch dates, cloud regions or prices. The company’s product page still marks the generation as “Coming soon.”
There are also no independent benchmarks for TPU 8tx or TPU 8i Max, and Google has not provided their absolute specifications in the cited material. The latency and doubling figures are Google’s claims. It remains unclear how the systems will perform on customer workloads, what they will cost relative to Nvidia-based services, and how much engineering customers will need to move or adapt applications.
The source material identifies Oracle’s planned 2027 TPU offering but does not give details about capacity, pricing or the customers and regions it will serve. Google’s claim that nearly 100 customers use TPUs is also not accompanied here by a breakdown of usage or workload scale. Those details will affect how much the new systems change buyers’ practical options.
Watch for 2027 Launch Details
The next milestones are Google’s release of pricing, deployment regions and final specifications, followed by the systems’ expected 2027 availability. Independent tests and customer reports will be needed to assess Google’s performance claims against real workloads and competing GPU services.
Customers and cloud buyers will also be watching for details on Oracle’s planned TPU offering and for evidence that Google’s software support can make switching or running workloads across hardware platforms practical. Until those details emerge, the announcement signals a broader TPU lineup, not a confirmed change in Nvidia’s market position.
Key Questions
What did Google announce?
Google announced TPU 8tx, aimed at confidential inference, and TPU 8i Max, designed for workloads with very long inputs. Both are variants in its eighth-generation TPU lineup.
When will the new TPUs be available?
Google’s Amin Vahdat told Bloomberg that the two new systems are expected in 2027. Google has not given exact launch dates or cloud regions, and its TPU page lists the systems as “Coming soon.”
Have the performance claims been independently tested?
No independent benchmarks for TPU 8tx or TPU 8i Max are included in the source material. The claims about latency and doubled compute, SRAM and memory bandwidth come from Google.
Do the new systems replace Nvidia GPUs?
The announcement does not show that Google’s TPUs replace Nvidia GPUs. They offer another potential option for AI workloads, but comparisons will depend on availability, price, performance and software needs.
What is the difference between TPU 8tx and TPU 8i Max?
TPU 8tx focuses on confidential computing for inference, while TPU 8i Max is designed to provide more compute and memory capacity for workloads with massive context windows. Google says both use existing silicon in new system configurations.
Source: ThorstenMeyerAI.com
