🔍 Read the full analysis: Google’s AI Chips Could Give Businesses Another Option Beyond 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 new systems in its eighth-generation TPU lineup: TPU 8tx, designed for confidential inference, and TPU 8i Max, designed for workloads with very long inputs. Google says they will be available in 2027, but has not published prices, deployment regions or independent performance tests. The additions give cloud customers another potential option alongside Nvidia GPUs, though availability and switching costs remain important questions.
Google announced two new AI chip systems on November 9, expanding its eighth-generation tensor processing unit (TPU) lineup with hardware for confidential computing and workloads involving very long inputs. The systems are expected in 2027, giving businesses another possible cloud-computing option beyond Nvidia GPUs, but neither is available to buy or independently tested yet.
The additions are TPU 8tx, aimed at confidential inference for regulated businesses and government users, and TPU 8i Max, intended for applications that process large amounts of context, such as AI agents working across a software codebase or lengthy video. They build on the TPU 8t training system and TPU 8i inference system that Google introduced in April.
Google says TPU 8tx uses the same silicon as TPU 8t, paired with fifth-generation AMD EPYC processors, and is designed to keep sensitive data and model weights protected while in active use. The company claims it can deliver up to twice the lower latency of TPU 8t with a near-zero performance penalty. Google describes TPU 8i Max as two low-latency TPU 8i chips on one package, with double the compute, on-chip SRAM and high-bandwidth memory (HBM) bandwidth of TPU 8i.
Those performance figures are Google’s claims, not independent test results. The company has not released absolute specifications for either new system, or announced prices or cloud regions. Its TPU product page lists all four eighth-generation systems as “Coming soon.” Amin Vahdat, Google’s chief technologist for AI and infrastructure, told Bloomberg the two new variants would be available in 2027.
Google’s AI Chips Expand Business Options
Google announced two specialised TPU systems for 2027. They could give cloud customers another option alongside Nvidia GPUs, though key buying details remain unknown.
Two systems, distinct workloads
Each system targets a different business need.
TPU 8tx
Designed for confidential inference, with sensitive data and model weights protected during processing.
TPU 8i Max
Designed for applications handling extensive context, such as AI agents working across a software codebase or lengthy video.
What buyers still need to know
The announcement describes potential capabilities, but leaves practical comparisons open.
| Buyer question | What is known |
|---|---|
| Pricing | Not announced |
| Cloud regions | Not announced |
| Performance evidence | Google claims; no independent tests yet |
| Availability | Expected in 2027; no specific date |
A possible route beyond Nvidia
More chip options could give businesses greater choice, if supply and software support meet their needs.
- Another cloud AI compute option
- Specialised systems for sensitive data or long inputs
- Google says PyTorch, Hugging Face, vLLM and SGLang are supported
- Compatibility with existing workloads
- Performance and costs in customer testing
- Capacity, launch access and switching effort
What happens next
Google expects TPU 8tx and TPU 8i Max in 2027.
Oracle has said it plans to offer Google TPUs in its cloud during 2027.
Workload tests, pricing and availability will show whether the systems are practical alternatives.
A Second Route to AI Compute
The announcement matters because demand for AI processors has left many companies reliant on a limited set of suppliers and cloud offerings. A credible alternative could give customers more choice and bargaining leverage, particularly if Google can supply the capacity and performance businesses need. Oracle has said it will offer Google TPUs through its own cloud starting in calendar year 2027, potentially making the hardware available beyond Google Cloud.
But a second chip option does not automatically mean an easy switch. Companies may need to adapt software, workflows and infrastructure to a different processor. Google is promoting TorchTPU, which it says allows PyTorch code and Hugging Face models to run natively on TPUs, along with TPU support for the open-source inference engines vLLM and SGLang. Such software support could reduce friction, but buyers will still need to test compatibility, performance and costs for their own workloads.
For businesses with sensitive data, TPU 8tx’s stated focus on keeping data protected during processing could be relevant to evaluating cloud-based AI. For teams running applications with large context windows, the memory and bandwidth claims for TPU 8i Max address a different constraint. Whether either system delivers a practical advantage will depend on independent results, availability and pricing once customers can try them.
Popular right now
Samsung 48 4K OLED Smart TV (2026) with WireLogic 8ft Ultra High Speed HDMI Cable
KitchenAid Artisan Mini Tilt-Head Stand Mixer, 3.5 Qt, Contour Silver
Samsung 43-Inch M70H Series 4K Mini LED Smart TV (2026)
As an affiliate, we earn on qualifying purchases.
Google’s Eighth-Generation TPU Lineup
Google introduced TPU 8t and TPU 8i at its Cloud Next conference on April 22, 2026. The company positioned TPU 8t for large-scale model training and TPU 8i for inference, the process of generating answers from trained models. Its April specifications list TPU 8t at 216 GB of HBM and 6,528 GB/s of memory bandwidth, while TPU 8i has 288 GB of HBM and 8,601 GB/s of bandwidth. Those are vendor-published figures; they do not establish how the chips will perform across different customer workloads.
Google’s April claims included up to 2.7 times the performance per dollar of the previous Ironwood generation for TPU 8t training, and up to 80% better performance per dollar for TPU 8i inference. The November announcement extends the lineup through changes to systems and packaging, rather than presenting the two new variants as entirely new silicon designs. Vahdat told Bloomberg they were reconfigured around tasks customers had requested.
The company is also working to broaden access and software support. Google says nearly 100 customers use TPUs, while Alphabet has described an environment constrained by supply. The November 10 report by Thorsten Meyer at ThorstenMeyerAI.com also disclosed that the article was drafted with help from Anthropic’s Claude, and that Anthropic is a Google TPU customer, according to Alphabet’s October earnings call. The report said neither Google nor Anthropic reviewed or sponsored it.
“This is the first time we have introduced four TPU systems in a single year.”
— Google, in its November 9 announcement
Performance, Price and Access
Several details needed for a buyer comparison are still missing. Google has not disclosed prices, deployment regions or absolute specifications for TPU 8tx and TPU 8i Max, and the systems are not yet available for outside testing. The company’s latency, compute, memory and bandwidth statements remain vendor claims until customers or independent testers can measure them.
It is also not yet clear how widely the systems will be offered at launch, how much capacity Google can provide, or whether their performance will justify changes to existing infrastructure. The source report said Google’s TPU page listed all four eighth-generation systems as coming soon. Google and Oracle have described future access plans, but specific service dates and customer terms were not provided.
Availability Expected in 2027
The next milestone is the planned 2027 availability of TPU 8tx and TPU 8i Max. Google has not supplied a more precise launch date, pricing or regional rollout details. Oracle has said it plans to offer Google TPUs in its cloud during calendar year 2027, although the specific systems and launch timing were not detailed in the source material.
Once the hardware becomes accessible, customers will be able to assess performance against their own workloads and compare costs with other options, including Nvidia-based systems. Evidence from real deployments, alongside clearer information on software support and supply, will show whether Google’s specialised TPU systems can become a practical alternative for more businesses.
Key Questions
What did Google announce?
Google announced TPU 8tx, for confidential inference, and TPU 8i Max, for workloads involving very long inputs. They join TPU 8t and TPU 8i in Google’s eighth-generation lineup.
When will the new Google TPU systems be available?
Google’s Amin Vahdat told Bloomberg that the two new variants are expected in 2027. The company has not given a specific release date or listed prices and service regions.
Are Google’s performance claims independently verified?
No independent test results were provided in the source material. Claims about latency, compute, memory and bandwidth come from Google, and the new systems have not yet been made available for outside testing.
Will Google’s TPUs replace Nvidia GPUs for businesses?
The announcement does not establish that. TPUs could offer another option for some cloud workloads, but adoption will depend on availability, pricing, performance and the engineering required to adapt software and systems.
Source: ThorstenMeyerAI.com
