Google Adds TPU 8tx and 8i Max to Its AI Chip Lineup

0

Google's latest custom chip systems target confidential AI inference and large-context workloads, widening its eighth-generation TPU lineup to four

AI-generated illustration of Google TPU 8tx and 8i Max chips
AI-generated illustration showing Google's TPU 8tx and TPU 8i Max chip systems.

Google is giving its AI hardware business two new specialized options. On November 9, 2026, the company announced TPU 8tx and TPU 8i Max, bringing its eighth-generation Tensor Processing Unit portfolio to four systems and sharpening its push to win enterprise customers in a market Nvidia dominates.

Google says both systems are built around existing eighth-generation silicon rather than entirely new chip architectures. Each is aimed at a different problem: one at protecting sensitive data during AI processing, the other at workloads that need very large amounts of memory and context.

TPU 8tx: Built for Secure, Confidential AI

TPU 8tx is designed for confidential AI inference, with government agencies and regulated businesses that handle sensitive information as its stated audience. It uses the same silicon as TPU 8t and is paired with fifth-generation AMD EPYC CPU hosts.

According to Google, the system provides hardware-level protection for sensitive data and model weights while they are actively being processed. It runs on Google's low-latency Boardfly network, and Google claims up to two times lower latency than TPU 8t, with a near-zero performance penalty.

That figure is specific to TPU 8tx compared with TPU 8t, and it should not be read as twice the speed in every application. Beyond confidential inference, Google says the system supports small-scale training, private-data fine-tuning and reinforcement learning.

AMD's Role

Google's announcement confirms that TPU 8tx relies on AMD's fifth-generation EPYC processors as host CPUs. The accelerator itself is Google's own TPU silicon, so AMD's part in this system is on the host side.

TPU 8i Max: More Compute in the Same Rack

TPU 8i Max takes a different approach. It combines two TPU 8i chips in a single package, which Google says doubles compute, SRAM and HBM bandwidth compared with TPU 8i.

The system is aimed at workloads that depend on large context windows and heavy memory bandwidth, including work with large software codebases, coding assistants, mathematical reasoning and drug discovery. Google also says it increases computing density within the existing rack footprint.

The Four-System Eighth-Generation Lineup

With the new additions, Google's eighth-generation portfolio now looks like this:

  • TPU 8t: large-scale AI training
  • TPU 8i: low-latency AI reasoning
  • TPU 8tx: secure and confidential computing
  • TPU 8i Max: large-context workloads and higher computing density

Google introduced TPU 8t and TPU 8i earlier in 2026 at Google Cloud Next. The company says announcing four TPU systems in a single year is a first for it.

From Internal Tool to Cloud Business

Google began developing TPUs more than a decade ago to meet its own computing needs, including voice recognition and language translation. The chips later supported Google's AI models, including Gemini, along with its internal services and Google Cloud customers.

Google says its TPU business now serves nearly 100 customers, among them frontier AI laboratories, trading firms, retailers and universities. The company is expanding access through its cloud business and software ecosystem.

What Alphabet's Earnings Show

The hardware push comes as Google Cloud grows quickly. On Alphabet's October 28, 2026, third-quarter earnings call, the company reported Google Cloud revenue of about $30.7 billion, up 103% from a year earlier. It said TPU system sales contributed meaningfully to the quarter.

Alphabet expects TPU system sales to keep scaling in the fourth quarter of 2026 and anticipates a meaningful revenue ramp from them in 2027. Those are reported results and company guidance, not independent forecasts.

Competing With Nvidia

The new systems add to Google's effort to compete with Nvidia, which remains a major force in AI GPUs. Google is focusing on specialized needs such as confidential computing, large-context applications and enterprise buyers rather than a single general-purpose product.

There is no confirmation that Google has overtaken Nvidia or gained market share as a result of these announcements.

Outside Signals on Demand

Several outside data points show how much attention TPUs are drawing, though none of them confirms demand for the two new systems specifically.

  • Anthropic: In a November 2, 2026, SEC filing, Anthropic described additional TPU computing capacity expected between 2028 and 2031, with estimated future payments of roughly $125 billion to $150 billion. The filing describes possible delivery arrangements involving Google or third-party TPU providers, so the figure is not a confirmed purchase of Google's newly announced systems.
  • Bloomberg Intelligence: The firm reportedly estimates that TPU-related sales, including associated hardware and software, could grow from about $12 billion in 2026 to more than $135 billion by 2028. That is an analyst forecast, not confirmed revenue.
  • Tenspresso: Bloomberg reported on November 4 and 5, 2026, that the startup, founded by former Google AI chip engineers, was in talks to raise financing at a valuation of about $500 million. Those discussions had not been confirmed as a completed investment.

What Remains Unclear

Google's announcement does not establish exact pricing or availability dates for either TPU 8tx or TPU 8i Max. It also does not name customers for the new systems.

Some secondary publications have cited an analyst estimate that Google TPUs cost roughly 30% less than Nvidia GPUs. That figure has not been established as a like-for-like pricing comparison.

For now, the announcement confirms the technical direction: Google is splitting its TPU line into more specialized systems, and the market will determine how well each one sells.

Post a Comment

0 Comments
* Please Don't Spam Here. All the Comments are Reviewed by Admin.
Post a Comment (0)
To Top