Google isn't slowing down in the AI hardware race. The Alphabet-owned company has just added two new systems to its eighth-generation AI chip lineup — and they're aimed squarely at the workloads big customers keep asking for: airtight security and enormous memory.

The two new Tensor Processing Units are called the TPU 8tx and the TPU 8i Max. But here's the key detail: these aren't brand-new chip designs. They're specialist systems built around silicon Google already has, combining processing power, networking, memory, and security into packages tuned for specific jobs.

TPU 8tx: AI That Stays Encrypted Even While It's Working

The 8tx targets government agencies and businesses in regulated industries — the kind of customers handling sensitive data who can't afford shortcuts on security.

It uses the same silicon as Google's existing TPU 8t, paired with custom processors from AMD, and protects data and AI models through hardware-level encryption. The big deal? It keeps information encrypted while computers are actively processing it — closing a security gap that normally opens up when encrypted data has to be unlocked for a computing task.

Google says the 8tx also delivers up to twice the reduction in latency compared with the standard TPU 8t, with almost no performance penalty. That's a rare combination: tighter security and faster response times.

TPU 8i Max: Built for AI That Needs to Remember More

The second system solves a different headache — memory. The 8i Max lets AI models work with far larger amounts of information without running into a memory bottleneck.

How? It literally combines two existing TPU 8i chips into a single package, doubling computing power, on-chip memory, and high-bandwidth memory capacity compared with the standard setup.

Who's this for? Workloads where a model has to chew through extensive information in a single task — think coding assistants, mathematical reasoning, and drug discovery. More memory means bigger context, and bigger context means more capable AI.

Google's Eighth-Gen TPU Family Is Now Four Strong

With these two additions, Google's eighth-generation portfolio now includes four systems:

• TPU 8t — large-scale AI training

• TPU 8i — real-time reasoning

• TPU 8tx — secure, encrypted AI processing

• TPU 8i Max — high-memory, heavy-workload AI

Google says its TPUs now serve nearly 100 customers, spanning tech companies, trading firms, retailers, and universities.

The Bigger Picture: Google vs. Nvidia

This expansion is a direct shot at Nvidia, whose GPUs remain the default hardware for training and running AI models across the industry. Google is betting that specialist TPU systems — tuned for security, memory, and specific workloads — can win over customers who want alternatives to Nvidia's dominance.

And the numbers suggest it's working. TPU sales made a meaningful contribution to Alphabet's September-quarter results, and Google Cloud revenue more than doubled to $30.71 billion, according to Reuters.

What This Means for You

If you're a business running AI at scale, the message is clear: the hardware market is getting more specialized, not less. Security-first AI and memory-heavy workloads now have purpose-built options. And more competition in AI chips usually means better pricing, better performance, and more choice down the line.

Which matters more to you — secure AI processing or massive memory for bigger models? Tell us in the comments.