Google’s forthcoming Tensor G5 chipset may represent a significant transformation in the company’s strategy regarding mobile silicon.  

Having depended on Samsung’s Exynos architecture and fabrication processes since the introduction of the inaugural Tensor, Google is anticipated to finally produce a fully in-house designed chip, manufactured by TSMC.  

Considering the customary annual release timeline for Pixel devices, Google is expected to unveil the Pixel 10 around October 2025.  
 
This timeframe corresponds with reports indicating that Google has commenced testing early samples of the Tensor G5, which were recently uncovered through shipping documentation.  

Which phones will utilize the Google Tensor G5?  

The Tensor G5 is set to debut in Google’s Pixel 10 lineup, succeeding the existing Pixel 9 series, which was initiated in August 2024.  

The Pixel 10 series is forecasted to encompass the standard Pixel 10, the Pixel 10 Pro, the Pixel 10 Pro XL, and the Pixel 10 Pro Fold.  

With the Tensor G5, Google seeks to address criticisms directed at prior Tensor models, particularly concerning thermals, battery efficiency, and overall performance.  

In recent months, Android Authority has published two comprehensive reports (first report, second report), one of which originates directly from Google’s gChips division, providing insights into the forthcoming features of Google’s chipsets.  

Below is a summary of the most anticipated specifications and characteristics of the Tensor G5.  

TSMC’s 3nm N3E fabrication process  
One of the most notable modifications for the Tensor G5, codenamed ‘laguna,’ is its fabrication at TSMC’s 3nm N3E node.  

This advanced node manufacturing process is the same utilized in Apple’s A18 Pro chip, which powers the iPhone 16 Pro and the iPhone 16 Pro Max, indicating a significant advancement in efficiency and performance.  
 
Previously, Tensor chips encountered performance challenges associated with Samsung’s process nodes, which were inferior to TSMC’s offerings.  
This configuration is expected to comprise one Arm Cortex-X4 prime core, a high-performance core designated for demanding tasks.  

The mid-tier configuration is anticipated to consist of five Cortex-A725 cores, which will enhance multi-core performance, particularly beneficial for multitasking.  
Finally, two efficiency-centric Cortex-A520 cores may be allocated for low-power tasks, thereby aiding battery conservation when the device is not under heavy load.  

This new configuration, highlighting a larger mid-core cluster, is projected to provide a marked improvement in multi-core performance and daily efficiency, which is a priority for Google.  

Recently, a mysterious device, believed to incorporate the Tensor G5 and codenamed ‘Google Frankel,’ appeared on Geekbench (via GSMArena) achieving 1323 points in single-core and 4004 points in multi-core assessments.  
As a point of reference, the Pixel 9 secured 4185 points in our Geekbench 6 multi-core evaluation.  

The leaked specifications closely align with the expected configurations of the Tensor G5, boasting a CPU layout comprising one prime core (up to 3.40 GHz), five performance cores (up to 2.86 GHz), and two efficiency cores (up to 2.XX GHz).  

However, the listing indicates a Power VR graphics unit rather than the anticipated Imagination Technologies DXT-48-1536 GPU.  
Since Power VR is also a product of Imagination Technologies, this inconsistency may be attributable to early testing stages, and the Geekbench scores should be interpreted cautiously, as they may not accurately depict the chip’s final performance.  

GPU and gaming  
For the first time within the Tensor lineup, Google may be transitioning to an Imagination Technologies GPU, specifically the dual-core IMG DXT-48-1536 operating at 1.1 GHz.  
This GPU enhancement could facilitate ray tracing capabilities, a first for Google’s Tensor chipsets, thereby enabling more lifelike graphics and enhanced gameplay experiences in mobile gaming.  

Moreover, the GPU may incorporate virtualization support, enabling it to manage accelerated graphics within virtual machines.  

This feature aligns with Google’s broader emphasis on virtualized applications and productivity tools, where elevated performance graphics can improve functionality.  

Enhanced AI  
The Tensor G5 is anticipated to exhibit moderate advancements in AI processing, featuring an upgraded TPU (Tensor Processing Unit) that boosts machine learning capabilities by approximately 14% compared to its predecessor.  

This TPU may also incorporate embedded RISC-V cores that would facilitate custom operations, rendering the TPU more adaptable and proficient at managing various on-device machine learning tasks.  

These enhancements, in conjunction with the TPU’s support for on-device training, could broaden the methods by which the Pixel 10 series learns and adapts to user preferences over time, further enabling real-time personalization within the Pixel line.  

Integrated Fan-Out Package on Package (InFO_PoP)  
The Tensor G5 is reported to utilize TSMC’s InFO_PoP technology, integrating RAM directly atop the chipset, which improves thermal management and electrical efficiency.  

This arrangement could facilitate thinner 1.16mm packaging and assist in mitigating overall power consumption while maintaining performance.  
Updates to this article will be provided as new information becomes available.  

In the interim, you may explore the current flagship Pixel 9 series offerings, as well as the historical development of Pixel phones, wherein we document the evolution of Google’s mobile devices throughout the years.  

Tensor G5 vs Tensor G4