🔍 Read the full analysis: Google’s AI Chip Expansion Explained: Why Nvidia Needs A Second Source on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Google announced two additional eighth-generation TPU systems on November 9: TPU 8tx for confidential computing and TPU 8i Max for workloads with very long inputs. Google says both are due in 2027, but neither has a published price or independent benchmark. The expansion gives cloud customers another potential option alongside Nvidia GPUs, though the practical cost and ease of switching remain unknown.
Google announced two new eighth-generation TPU systems on November 9, adding a confidential-computing option and a system designed for very long inputs to its AI chip lineup. The systems, TPU 8tx and TPU 8i Max, are expected in 2027, according to Google AI and infrastructure chief Amin Vahdat, giving cloud customers another potential source of AI computing alongside Nvidia GPUs.
The announcement expands Google’s eighth-generation family from two systems to four. Google introduced TPU 8t for training and TPU 8i for inference at its Cloud Next conference on April 22. The new TPU 8tx is aimed at secure, confidential inference, while TPU 8i Max is designed to handle large context windows, such as prompts involving extensive codebases or long video.
Google says TPU 8tx uses the same silicon as TPU 8t, paired with fifth-generation AMD EPYC processors, and is built to protect data and model weights while they are actively being processed. The company claims it can deliver up to two times lower latency than TPU 8t with a near-zero performance penalty. Those are vendor claims; independent tests have not been published.
TPU 8i Max combines two TPU 8i chips in one package. Google says it doubles the base system’s compute, on-chip SRAM and high-bandwidth memory bandwidth while fitting within the existing rack footprint. The company has not published absolute specifications for either new variant. Google’s TPU product page lists all four eighth-generation systems as “Coming soon”; prices and cloud regions have not been announced.
Google’s TPU Expansion Adds Options for AI Computing
Google’s new systems target confidential inference and very long inputs. They could give cloud customers another option alongside Nvidia, but price, availability and real-world performance remain unknown.
Two systems, distinct needs
Google says the additions adapt its existing silicon for specific workloads.
TPU 8tx
Designed to protect data and model weights during processing. Google claims up to two times lower latency than TPU 8t; independent tests are not published.
TPU 8i Max
Combines two TPU 8i chips in one package for workloads such as large codebases or long video.
A broader choice for buyers
More supply options could improve customer choice and negotiating leverage.
- Another source of AI computing alongside Nvidia GPUs
- Google says support includes PyTorch, Hugging Face models, vLLM and SGLang
- Oracle plans to offer Google TPUs in its cloud from calendar year 2027
- Moving workloads may require engineering, software changes and testing
- Prices, cloud regions and precise launch dates are not announced
- Cost and performance versus Nvidia alternatives remain unverified
What the lineup covers
Google split its eighth-generation systems across training, inference and specialised needs.
| System | Designed for | What is known |
|---|---|---|
| TPU 8t | Large-scale training | Up to 9,600 chips in a superpod |
| TPU 8i | Inference | Base system for TPU 8i Max |
| TPU 8tx | Confidential inference | Same silicon as TPU 8t, with security-focused configuration |
| TPU 8i Max | Very long inputs | Two TPU 8i chips in one package |
The 2027 availability test
The announcement sets Google’s direction; customer decisions will depend on practical details.
Google has announced 2027 availability, but no exact dates or cloud regions.
Pricing and capacity details will help buyers assess the cost of switching.
Independent benchmarks and workload testing can show how the systems perform in practice.
More Options Beyond Nvidia GPUs
Google’s expansion matters because access to AI computing has become a strategic concern for companies training and serving models. Nvidia remains the dominant supplier of AI processors, and demand for its chips has exceeded supply, according to Bloomberg’s reporting cited by ThorstenMeyerAI.com. A credible alternative could give cloud providers and customers more choice and potentially more negotiating leverage.
That does not mean buyers can switch without cost. Moving workloads from Nvidia GPUs to TPUs can require engineering work, software changes and testing. Google has pointed to support for PyTorch, Hugging Face models and inference engines including vLLM and SGLang as ways to make TPUs easier to use. Those software developments may lower barriers, but they do not establish that every workload will perform as well or cost less on a TPU.
There are signs of demand for Google’s chips: the source report says Google has reported nearly 100 TPU customers and that Alphabet described a supply-constrained environment. Oracle has also said it will offer Google TPUs in its cloud from calendar year 2027. The new systems could broaden the workloads those customers can run, but availability and real-world performance will determine how much they change the market.
Products Worth Considering
Apple MacBook Air 13.6-inch Laptop with M5 Chip, 16GB Memory, 512GB SSD, Midnight
Dell XPS Tower 8960 Desktop, Intel Core i7-13700, NVIDIA GeForce RTX 4060, 16GB DDR5, 512GB SSD
Apple Mac mini (M6, 12-core CPU and GPU, 24GB RAM, 1TB SSD)
As an affiliate, we earn on qualifying purchases.
How Google Split Its TPU Lineup
Google’s April launch divided its eighth-generation TPUs by task. TPU 8t is designed for large-scale model training, while TPU 8i targets inference, the process of generating responses from a trained model. Google says TPU 8t can be configured in superpods of 9,600 chips. Its published figures for the base systems include 216 GB of high-bandwidth memory for TPU 8t and 288 GB for TPU 8i, with TPU 8i also carrying more on-chip SRAM and memory bandwidth.
The November additions are described as new systems built from existing silicon rather than entirely new chip designs. Vahdat told Bloomberg the variants were reconfigured around tasks customers had requested. Google has argued that changes to packaging, memory, networking and security can be made faster than designing new silicon. The company also says its eighth-generation systems improve performance per dollar over the prior generation, but those comparisons are Google’s own estimates and are not independent benchmarks.
“This is the first time we have introduced four TPU systems in a single year.”
— Google, in its November 9 announcement
Prices, Benchmarks and Rollout
Google has not announced prices, launch regions or precise availability dates for TPU 8tx and TPU 8i Max. Vahdat told Bloomberg the variants would be available in 2027, but the company has not provided a more specific schedule. Google’s product page continues to list the four eighth-generation systems as “Coming soon.”
There are also no independent benchmarks for the new variants. Google’s performance and latency figures remain company claims, and the source material does not provide absolute specifications for the two systems. It is not yet clear how they will perform across different models and workloads, how their costs will compare with Nvidia-based alternatives, or how much engineering customers will need to move existing applications.
The 2027 Availability Test
The next milestone is Google’s planned release of TPU 8tx and TPU 8i Max in 2027. Buyers will need details on pricing, cloud regions, capacity and software support before deciding whether to use either system. Oracle’s planned offering of Google TPUs in its cloud is also scheduled to begin in calendar year 2027, giving customers another route to access the chips if that plan proceeds on schedule.
For now, the announcement establishes Google’s product direction, not an independently measured alternative to Nvidia. Customer adoption, published specifications and third-party testing will help show whether the specialised systems offer practical performance or security benefits for the workloads they target.
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. They join TPU 8t and TPU 8i in Google’s eighth-generation lineup.
When will the new TPUs be available?
Google’s Amin Vahdat told Bloomberg that the two new variants are expected in 2027. Google has not announced exact launch dates or cloud regions.
Are the performance claims independently verified?
No independent benchmarks for TPU 8tx or TPU 8i Max have been published in the source material. Google’s claims, including lower latency for TPU 8tx and doubled resources for TPU 8i Max, have not been independently tested.
Does Google’s expansion mean companies can easily replace Nvidia GPUs?
Not necessarily. TPUs offer another potential option, but moving workloads can involve software changes, engineering time and performance testing. The cost and ease of switching will depend on each customer’s applications and the systems’ eventual availability and results.
Source: ThorstenMeyerAI.com
