🔍 Read the full analysis: Four Chips In A Year: Google’s Bid To Offer An Alternative To Nvidia 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 specialized additions to its eighth-generation TPU lineup on November 9, bringing the number of systems introduced this year to four. Google says TPU 8tx is designed for confidential computing and TPU 8i Max for long-context workloads; both are expected in 2027, and independent performance tests and prices are not available.
Google announced two new AI chip systems on November 9, expanding its eighth-generation Tensor Processing Unit lineup to four offerings in 2026 as it seeks to give cloud customers an alternative to Nvidia processors. The TPU 8tx is aimed at confidential computing, while the TPU 8i Max is designed for workloads involving very long inputs; neither is available yet, and Google expects both in 2027.
The additions build on the TPU 8t, which Google designed for training AI models, and the TPU 8i, intended for inference — the computing used to generate answers from trained models. Google describes TPU 8tx as optimized for secure, confidential inference, particularly for regulated businesses and government users. It says the system protects sensitive data and model weights while they are being processed, rather than only when stored or transmitted.
Google says TPU 8tx uses the same silicon as TPU 8t and is paired with fifth-generation AMD EPYC processors. The company claims it can provide up to two times lower latency than TPU 8t with a near-zero performance penalty. Those figures are Google’s claims; independent test results have not been published.
TPU 8i Max combines two TPU 8i chips on one package. Google says this gives it double the compute, on-chip SRAM and HBM bandwidth of TPU 8i while fitting within the existing rack footprint. It is intended for tasks involving long context windows, such as analyzing a large software codebase or extended video. Google has not released absolute performance or memory figures for either new system.
Google’s four-chip lineup seeks a Nvidia alternative
Google added two specialized systems to its eighth-generation TPU lineup in 2026. Both are expected in 2027, but prices and independent performance tests are not available.
Two systems, distinct workloads
Google tailored the new variants from existing silicon for different customer needs.
TPU 8tx
Designed for confidential inference, protecting data and model weights while in use.
TPU 8i Max
Combines two TPU 8i chips for tasks such as analyzing large codebases or extended video.
From two base chips to four
The lineup spans model training, inference and specialized variants.
What buyers still need to assess
A second source could offer more choice, but its value depends on real workloads and access.
- Confidential inference for regulated businesses and government users
- Long-context tasks involving codebases or video
- Google Cloud and planned Oracle Cloud access
- No prices or specific launch dates
- No independent benchmarks for the new systems
- Cloud regions and available capacity are unspecified
A Second Route to AI Compute
The announcement adds another option for organizations planning large AI workloads at a time when access to specialized processors is a major consideration. Nvidia remains the dominant supplier, while Google is positioning its own TPUs as systems customers can use through Google Cloud. A credible alternative could give some buyers more choice and negotiating leverage, though the announcement alone does not establish how the systems compare on price, availability or real workloads.
Google says nearly 100 customers use its TPUs, and Alphabet has described its operating environment as supply-constrained. Oracle has also said it will offer Google TPUs through its cloud beginning in calendar year 2027. These developments point to interest beyond Google’s own services, but the scale of customer deployments and the amount of capacity available have not been disclosed.
For customers, switching infrastructure is not just a matter of choosing a chip. Teams may need to adapt software, workflows and deployment practices, and the value of a second source will depend on how well it performs for a specific workload. Google is promoting software support alongside the hardware, including TorchTPU and integrations for open-source inference engines, but their effect on adoption remains to be seen.
Products Worth Considering
Apple MacBook Air 13-inch Laptop with M5 Chip, 16GB Memory, 512GB SSD, Liquid Retina Display, and 12MP Center Stage Camera (2026, Silver)
Apple Mac mini (2026), M6 12-core CPU and GPU, 24GB Memory, 1TB SSD
Apple Mac mini Desktop Computer with M2 Pro, 16GB Memory, 512GB SSD, Gigabit Ethernet
As an affiliate, we earn on qualifying purchases.
From Two Base Chips to Four
Google introduced TPU 8t and TPU 8i at its Cloud Next conference on April 22, 2026. The company said TPU 8t was built for large-scale training in systems of up to 9,600 chips, while TPU 8i was designed to serve model responses with low latency. Google reported improvements over its previous generation, Ironwood, of up to 2.7 times in performance per dollar for TPU 8t training and up to 80% for TPU 8i inference. These are vendor comparisons, not independently verified results.
The new announcement is notable for adding specialized systems without introducing wholly new chip designs. Google AI and infrastructure chief technologist Amin Vahdat told Bloomberg the variants were reconfigured from existing silicon to address customer requests. Google says changes to packaging, memory, networking and security can be made more quickly than developing a new processor design. That approach lets the company tailor systems to different needs within a single generation.
Google has published specifications for the two base systems. TPU 8t is listed with 216 GB of high-bandwidth memory and 128 MB of on-chip SRAM; TPU 8i has 288 GB of high-bandwidth memory and 384 MB of SRAM. The figures illustrate Google’s split between training and inference priorities, but they do not provide a direct performance comparison with Nvidia products or establish how the new variants will perform.
“This is the first time we have introduced four TPU systems in a single year.”
— Google, in its November 9 announcement
Prices, Tests and Capacity Await
Neither new system is available, and Google’s TPU product page lists the eighth-generation systems as coming soon. Vahdat told Bloomberg the two variants are due in 2027, but Google has not announced specific launch dates, cloud regions or prices. It has also not published absolute specifications for TPU 8i Max or TPU 8tx.
There are no independent benchmarks for the new systems in the source material. Google’s latency and capacity comparisons are company claims, and the published TPU 8t and TPU 8i figures do not settle how either system will compare with Nvidia hardware under equivalent conditions. It is also unclear how much capacity Google will make available to outside customers, particularly given the supply constraints described by Alphabet.
Watch for 2027 Cloud Availability
The next concrete milestone is the expected 2027 availability of TPU 8tx and TPU 8i Max. Google has not given dates or regional rollout details, so customers will need further announcements to know when they can access either system. Oracle’s planned offer of Google TPUs in its cloud, also beginning in calendar year 2027, could provide another route for customers, though product availability and capacity remain unspecified.
For a clearer comparison with Nvidia, buyers and analysts will need published prices, independent benchmarks and evidence from deployed workloads. Software compatibility and the practical effort of moving existing systems to TPUs will also shape whether Google’s expanded lineup becomes a usable alternative, rather than simply another announced option.
Key Questions
What did Google announce?
Google announced TPU 8tx, aimed at confidential inference, and TPU 8i Max, designed for workloads with very long context windows. They expand Google’s eighth-generation TPU lineup to four systems, including TPU 8t and TPU 8i.
When will TPU 8tx and TPU 8i Max be available?
Google’s Amin Vahdat told Bloomberg the two new variants are expected in 2027. Google has not published specific release dates or cloud regions.
Has Google shown that the new systems outperform Nvidia chips?
No independent benchmarks for TPU 8tx or TPU 8i Max are available in the source material. Google’s performance comparisons are company claims, and no direct, independently tested comparison with Nvidia systems has been reported.
What is TPU 8i Max intended to do?
Google says TPU 8i Max is for workloads with massive context windows, including work across large codebases or long videos. It combines two TPU 8i chips on one package; Google says it doubles compute, SRAM and memory bandwidth relative to TPU 8i.
Can customers buy or rent the new TPUs now?
No. Google’s TPU page lists the systems as coming soon. The company has not disclosed prices or specific availability details for the new variants.
Source: ThorstenMeyerAI.com
