Google’s new Tensor G5 chip, fabricated on TSMC’s 3 nm node, represents the company’s single most significant jump in processor performance to date.

Compared to its predecessor, the Tensor G4, it has 60% faster TPU for AI tasks, and 34% more CPU.

As well as raw speed, G5 boasts many impressive new components, such as a fresh ISP, GPU, modem, and power controllers.

Early benchmarks from the Pixel 10 Pro Fold, which was unveiled earlier this week, show up to 36% performance gains over the G4, even besting rival chips like Snapdragon 8 Gen 3.

The move away from sourcing from Samsung and toward in‑house design has given Google much more control over efficiency, camera processing, and AI workloads.

That noted, Google isn’t designing every element and has used an externally sourced GPU and modem.

Also, there’s always plenty of risk around the first generation of any new architecture.

A powerful processor often results in a popular device

But G5 lives up to the hype, it could be Google’s silver bullet, finally giving Pixel devices the performance and efficiency to justify their AI-first promises.

That’s because the Tensor G5 is reportedly the first mobile chip to run Google’s Gemini Nano model fully on-device. That means devices like the Pixel 10 Pro Fold can now handle generative AI features without relying on the cloud.

Better on-device transformers, smarter camera tricks, longer battery life, more frictionless access to AI – all flow from the Tensor G5.

Google has long aspired to stand toe-to-toe with Apple and Qualcomm on silicon.

Industry observers will be watching closely to see whether Google’s (largely) in-house-built new processor succeeds or fails.

If it’s the latter, the new Pixel range will presumably continue to struggle against industry-leading Samsung and Apple devices.