Google Ships Four TPU Systems in 2026 as TPU 8Tx and 8i Max Arrive

Google announced two new TPU systems on November 9, 2026, making 2026 the first year the company has introduced four TPU systems in a single year. The new TPU 8tx targets confidential inference and small-scale training; TPU 8i Max is built for massive context windows and higher rack density.

Google put its TPU customer count at nearly 100, spanning frontier labs, trading firms, retailers and universities. The broader TPU business covers both Google Cloud infrastructure and deployments inside customers' own data centers.

What does TPU 8tx protect?

TPU 8tx serves regulated enterprises and governments that run frontier or proprietary models and autonomous agents alongside sensitive data. It fills a gap in conventional encryption: data can be encrypted at rest and in transit, but must be decrypted in memory while computation runs. Google says the hardware keeps sensitive data and model weights isolated even during active processing.

The system uses the same silicon as the training-focused TPU 8t, but pairs it with 5th Generation AMD EPYC CPU hosts. It also adopts the low-latency Boardfly network developed for real-time reasoning. Google reports up to 2x lower latency than TPU 8t, with almost no offsetting performance loss.

That combination lets organizations run and customize models on a single system. Teams can fine-tune models on private data, distill larger models into smaller ones, and train agents through reinforcement learning. Private information stays on that one system for all three tasks instead of moving between separate clusters.

The takeaway: TPU 8tx turns one box into the whole pipeline for sensitive workloads.

Why does memory bandwidth matter for TPU 8i Max?

TPU 8i Max targets workloads where dozens of agents work across an entire software codebase, or hours of video, simultaneously — and memory bandwidth becomes the bottleneck. Google's answer is a single package containing two low-latency TPU 8i chips.

The numbers are straightforward:

• Double the compute of TPU 8i
• Double the on-chip SRAM
• Double the high-bandwidth memory (HBM) bandwidth
• The package fits within the existing rack footprint

That last point matters most for data centers: it means more computing power per square foot and per watt for context-heavy workloads, with no new floor space.

The underlying TPU 8i and TPU 8t debuted at Google Cloud Next on April 22, 2026 — the 8i for real-time reasoning, the 8t for large-scale training. The 8i design pairs 288 GB of high-bandwidth memory with 384 MB of on-chip SRAM, providing the baseline that 8i Max now doubles.

How does Google ship new TPUs without new chips?

Google draws a sharp line between these launches and new base silicon. A new chip requires billions of transistors laid out through a multi-year design and manufacturing cycle. A system, by contrast, determines how compute, networking, storage, packaging and security are assembled around that silicon.

Its modular AI Hypercomputer approach treats TPUs as building blocks whose memory, optical networking, packaging and security can be adjusted. Google said engineers can deliver specialized supercomputers in months rather than the years required for new silicon.

The two new systems apply that method differently: 8tx changes the hosts, network and security around training silicon, while 8i Max changes the package around inference silicon.

Google has long co-designed TPUs with Google DeepMind, but the company said external organizations are increasingly shaping requirements — specifically requests for massive context windows and specialized encryption for regulated workloads. Those requests map directly onto the two new systems.

In short: system design, not silicon design, is now Google's fastest lever for specialization. Confidential inference and massive-context agents present different hardware constraints, but Google is addressing both on existing eighth-generation silicon — segmenting its infrastructure offerings without waiting for another base-chip generation.