Google is rolling out a significant behind-the-scenes upgrade to Android that could quietly improve how your smartphone performs every day. The company is introducing a new optimisation method called Automatic Feedback-Directed Optimisation (AutoFDO), aimed at making Android devices faster, smoother and more power-efficient.
Unlike flashy feature updates, this improvement works deep within the Android system — at the kernel level. The Android kernel functions as the core bridge between a phone’s hardware; processor and the apps users rely on daily. It plays a crucial role in managing system resources and performance. According to Google, the kernel accounts for roughly 40 per cent of a smartphone’s CPU usage, meaning even minor efficiency gains can translate into noticeable real-world improvements.
Google’s Android LLVM toolchain team recently shared details about how AutoFDO works and why it matters. Instead of relying solely on traditional assumptions about how software is typically used, the new approach uses real-world behavioural data to guide optimisation decisions.
To achieve this, Google conducts-controlled lab simulations that mimic everyday smartphone activity using popular Android apps. During these tests, profiling tools monitor which parts of the kernel are accessed most frequently. These heavily used sections — often referred to as “hot” code paths — are then prioritised during the compilation process.
AutoFDO leverages this usage data to guide the compiler when rebuilding the kernel. By reorganising and optimising the most frequently executed code first, the system can operate more efficiently under typical usage conditions. This refined process allows the kernel to respond faster to routine tasks such as opening apps, switching between functions, or booting up the device.
Traditional kernel compilation methods generally depend on broad assumptions about usage patterns. AutoFDO changes that model by using actual performance data to optimise the system based on how devices are realistically used. The result is a smarter, more tailored optimisation process designed to improve day-to-day performance.
In early internal testing, Google reported measurable improvements. Devices demonstrated a 2.1 per cent faster boot time and a 4.3 per cent improvement in cold app launch speeds. Beyond these metrics, the company observed performance gains across several additional benchmarks, contributing to an overall smoother user experience.
Importantly, Google has designed AutoFDO to operate in what it describes as a “conservative by default” mode. If the system encounters scenarios outside the tested usage patterns, it can revert to traditional compilation methods. This ensures stability and reliability while still delivering performance benefits where possible.
The rollout has already begun for the latest Android kernel branches, including android16-6.12 and android15-6.6. Future versions, such as android17-6.18, are also expected to support the optimisation framework.
While users may not immediately notice a dramatic change after an update, the cumulative effect of faster app launches, smoother multitasking and improved efficiency could enhance everyday smartphone experiences. By refining performance at the kernel level, Google is focusing on long-term improvements that help Android devices feel responsive and reliable over time.
