🔍 Read the full analysis: Google’s Latest AI Chips Spotlight The Value Of A Second Nvidia Source on ThorstenMeyerAI.com
Get smart everyday buys 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 specialized systems in its eighth-generation TPU lineup: TPU 8tx for confidential inference and TPU 8i Max for very long inputs. Google says both are due in 2027, but has not published prices or independent performance tests. The expansion gives cloud customers another potential AI-computing source alongside Nvidia, though availability and the cost of switching remain open questions.
Google announced two new systems for its eighth-generation tensor processing units on November 9: TPU 8tx, designed for confidential inference, and TPU 8i Max, designed to handle very long inputs. The additions expand Google’s alternative to Nvidia-based AI computing, but both systems are listed as coming soon and are expected in 2027.
The new systems build on the TPU 8t and TPU 8i, which Google introduced at its Cloud Next conference in April. The company positions TPU 8t for training AI models and TPU 8i for inference, the process of generating outputs from a trained model. Google says TPU 8tx targets organizations such as regulated businesses and governments that need to protect data while it is being processed.
Google says TPU 8tx uses the same silicon as TPU 8t, paired with fifth-generation AMD EPYC processors. The company claims it can deliver up to twice the lower latency of TPU 8t with a near-zero performance penalty. TPU 8i Max combines two low-latency 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.
Those performance figures are Google’s claims, not independent test results. Google has not published absolute specifications for either new system, list prices, or deployment regions. Its TPU product page lists the systems as coming soon. Google chief technologist for AI and infrastructure Amin Vahdat told Bloomberg the new variants are expected to be available in 2027.
Google expands its TPU lineup giving cloud buyers another route
Two specialized systems target confidential inference and very long inputs. Their value alongside Nvidia depends on availability, cost and real-world testing.
Two systems, distinct needs
Google adds specialized options to its eighth-generation TPU range.
TPU 8tx
Designed to protect data while it is processed, for organizations such as regulated businesses and governments.
TPU 8i Max
Combines two TPU 8i chips to increase compute and memory resources for long-context workloads.
A second route to AI compute
More cloud options could help buyers evaluate alternatives as demand for Nvidia processors exceeds available supply.
- Another source of cloud AI compute
- More options when choosing infrastructure
- Specialized features for sensitive data and long inputs
- Workloads may need engineering and software changes
- Performance depends on practical testing
- Full service costs are not yet available
What remains to be learned
The announcement sets out product plans, while key buying details remain open.
Google expects availability in 2027; exact launch dates, regions and capacity have not been announced.
Google has not published prices or full specifications for either new system.
Performance comparisons are Google’s claims; independent benchmarks for the new systems are not available.
A Second Route to AI Compute
The announcement matters because demand for AI computing has outstripped available supply, and Nvidia remains the dominant supplier of AI GPUs. Bloomberg has reported that demand for Nvidia processors exceeds available supply. Google’s expanding TPU range gives cloud customers another option to evaluate, and could strengthen their negotiating position when choosing infrastructure.
That does not mean buyers can switch without cost. Moving workloads from Nvidia GPUs to TPUs can require engineering work, software changes and performance testing. Google points to support for PyTorch and Hugging Face models through TorchTPU, as well as integrations with the open-source inference systems vLLM and SGLang. These tools may lower barriers, but their practical value will depend on how well customer workloads run and what the full service costs.
The specialised systems also target distinct needs. Confidential processing is relevant to organizations handling sensitive information, while larger memory capacity may help applications that must process long documents, codebases or video. Whether those features deliver a useful advantage for buyers remains to be tested against real workloads and competing hardware.
Popular right now
De'Longhi Rivelia Automatic Espresso Machine with Grinder and Milk Frother, Beige
Philips 3300 Series Fully Automatic Espresso Machine with Milk Frother, Grinder, Hot and Iced Coffee Presets, SilentBrew, 15-Bar Pressure (E
Samsung 48 4K OLED Smart TV (2026) with WireLogic 8ft Ultra High Speed HDMI Cable
As an affiliate, we earn on qualifying purchases.
Google’s Eighth-Generation TPU Lineup
Google first announced its eighth-generation TPU systems on April 22, 2026, splitting the generation into two base offerings. TPU 8t is aimed at large-scale training, with Google describing superpods containing 9,600 chips. TPU 8i is built for inference, where memory capacity and the speed of moving data can affect how quickly models respond.
Google’s published figures for the base systems show that trade-off. TPU 8t has 216 GB of high-bandwidth memory, 128 MB of on-chip SRAM, and peak FP4 compute of 12.6 petaflops. TPU 8i has lower stated peak compute at 10.1 petaflops, but more memory: 288 GB of high-bandwidth memory and 384 MB of SRAM. Google also lists higher memory bandwidth for TPU 8i. These are vendor specifications; independent comparisons of the new variants are not yet available.
Google says it introduced four TPU systems in a single year, a faster pace than its earlier pattern of releasing a new AI processor about every two years, Bloomberg reported. Rather than designing entirely new silicon for each use, the company says it can adapt systems through packaging, memory, networking and security changes. Oracle has also said it will offer Google TPUs in its cloud beginning in calendar 2027, potentially extending access beyond Google’s own cloud platform.
“This is the first time we have introduced four TPU systems in a single year.”
— Google, in its November 9 announcement
Availability, Pricing and Tested Performance
Key buying details remain unavailable. Google has not announced prices, which cloud regions will offer the systems, or exact launch dates within 2027. It has also not published full specifications for TPU 8tx or TPU 8i Max. The performance comparisons are company claims, and no independent benchmarks for the new systems are cited in the source material.
It is also unclear how much of Google’s stated performance will carry over to customer workloads, or how easily organizations can move existing software from Nvidia GPUs. Google says its software support is expanding, but buyers will need practical testing to compare performance, reliability and total costs. Google has not disclosed how much capacity will be available when the systems launch.
The 2027 Cloud Rollout
Google says TPU 8tx and TPU 8i Max are expected to become available in 2027. The next useful milestones for prospective customers will be Google’s announcements on pricing, cloud regions, capacity and launch timing, followed by independent testing once systems can be evaluated.
Oracle has separately said it plans to offer Google TPUs in its cloud starting in calendar 2027. That rollout could give customers another way to access the chips, but details about the service and its availability have not been provided here. Until the systems are offered and tested, the announcement establishes Google’s product plans—not a demonstrated performance or cost advantage over Nvidia.
Key Questions
What did Google announce?
Google added TPU 8tx, for confidential inference, and TPU 8i Max, for very long inputs, to its eighth-generation TPU systems.
When will the new TPUs be available?
Google’s Amin Vahdat told Bloomberg the two systems are expected in 2027. Google has not given exact launch dates or cloud regions.
Has Google published prices or independent benchmarks?
No prices or independent benchmarks for the new systems are available in the source material. Google’s published performance comparisons are vendor claims.
How do TPU 8tx and TPU 8i Max differ?
TPU 8tx is aimed at keeping data and model weights protected during processing. TPU 8i Max combines two TPU 8i chips in one package to increase compute and memory resources for long-context workloads, according to Google.
Why does a Google TPU alternative matter to AI customers?
Nvidia remains a leading supplier of AI processors, and demand has been reported to exceed supply. Google’s TPUs could provide another source of cloud AI compute, though switching workloads may require engineering effort and the new systems have not yet been independently tested.
Source: ThorstenMeyerAI.com
