Google is undertaking a mass migration to Arm - find out the secrets behind what it takes for the world's biggest companies to port their internal workloads to new hardware

• Google is moving thousands of internal workloads from x86 to Arm CPUs
• The company built an AI tool called CogniPort to automate migration fixes
• Google engineers spent months fixing test failures linked to x86 infrastructure

Google has embarked on a hugely ambitious project to migrate all its internal workloads from x86 to Arm-based CPUs, a process that involves one of the largest hardware transitions ever attempted by a global tech company.

The effort aims to allow its systems to run efficiently on both x86 processors and its custom-built Axion silicon.

With roughly 30,000 applications already converted, Google continues to rely heavily on automation to handle the massive codebase involved in the process.

Porting workloads at warehouse scale

In a blog post outlining the project, Google’s engineering fellow Parthasarathy Ranganathan and developer relations engineer Wolff Dobson noted the migration began with some of the company’s most critical systems, including F1, Spanner, and Bigtable.

Initially, teams relied on conventional software development practices with dedicated engineers and weekly coordination meetings.

Although they expected major architectural hurdles, modern compilers and debugging tools helped reduce many of the anticipated issues.

However, a large amount of time was still devoted to adjusting thousands of tests that were closely tied to Google’s existing x86-based infrastructure.

Engineers also faced challenges updating legacy build and release systems, managing production rollouts, and ensuring stability across mission-critical environments.

To accelerate the transition, Google developed a new AI tool known as “CogniPort.”

The system works by analyzing build and test errors and then attempting to automatically fix them, particularly in cases where an Arm-specific library or binary fails to compile.

CogniPort has shown a success rate of around 30%, performing best when handling test corrections, data handling inconsistencies, and conditional platform code.

While not flawless, the tool represents a key step in enabling automation at warehouse scale and reducing the human workload required for such conversions.

The long-term motivation behind Google’s move lies in performance and efficiency - its Axion-powered Arm servers reportedly deliver up to 65% better price-performance and can be as much as 60% more energy-efficient than comparable x86 instances.

This shift could result in fewer x86 processors across Google’s vast data infrastructure, potentially transforming the makeup of its internal compute clusters.

For now, major applications such as YouTube, Gmail, and BigQuery already operate on both x86 and Arm-based systems.

As Google migrates the remaining 70,000 packages, doubts persist about whether AI tools can handle such scale without adding new maintenance challenges across its systems.

Via The Register

Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds. Make sure to click the Follow button!

And of course you can also follow TechRadar on TikTok for news, reviews, unboxings in video form, and get regular updates from us on WhatsApp too.