AI is making Android phones fight for memory

AI is making Android phones fight for memory.

Here is something the AI industry does not advertise: the data centres running the models you use every day need an enormous amount of memory chips. High-bandwidth memory, the specialised kind that lets AI systems process huge amounts of data at speed, is being consumed by the hundreds of thousands of servers that power ChatGPT, Gemini, Claude, and everything else you have heard about over the last two years.

Those chips come from the same global supply chain that makes the RAM inside your phone.

And right now, the AI industry is winning that competition.

The result is a memory shortage that is working its way down from data centres to device manufacturers to the phone in your pocket. Deloitte has called it “RAMageddon.” The cost of server-grade memory roughly doubled in the first quarter of this year alone. Consumer-grade memory, the kind that goes into smartphones, is becoming more expensive and harder to source at the volumes manufacturers need.

Google just made it official. On August 26, the company published a statement to Android developers announcing new app quality requirements. The reason given was not a technical preference. It was a supply chain problem: the mobile industry is now facing “significant hardware supply constraints that are altering device memory availability,” and that is affecting the experience of real users on real devices.

This is the AI revolution, but from an angle nobody talks about. The part where your phone gets slower because a language model somewhere needed more chips.

What Google Is Actually Requiring

Starting in February 2027, apps on Google Play will need to meet new performance thresholds or face consequences. Apps that fail to comply risk reduced visibility in Google Play search and restrictions on their ability to be published in the store at all.

The requirements cover three areas.

The first is dynamic memory usage. Every app running on an Android phone takes up a share of the device’s available RAM. When too many apps grab too much, the system slows down, multitasking breaks, and phones start crashing. Google is now setting hard limits based on a device’s physical RAM. On a phone with 8 gigabytes of RAM, for example, an app can use no more than 2.25 gigabytes in the foreground, 1.5 gigabytes while running a user-facing service, and 1.5 gigabytes in the background. The limits scale with the device, but they apply everywhere.

The second area is bitmap memory, which is how apps store images. Poorly optimised image handling is one of the most common ways apps eat RAM they do not need. The new rules set specific limits here too.

The third is code optimisation. Developers will need to shrink and optimise their application code by at least 25 per cent, targeting the DEX files that represent the compiled version of an app’s code. Bloated code slows everything down, even before the app does anything. Google is requiring developers to clean this up.

To help, Google is rolling out tools that alert developers when their apps exceed the new thresholds before the enforcement deadline hits. The idea is to give the development community time to fix problems, not to surprise anyone in February with a banned app.

The Second Change: Switching Phones Should Not Mean Starting Over

Alongside the memory rules, Google also announced a second requirement with a slightly later deadline: by April 2027, apps will need to support Zero Tap Sign-In for device migration.

This one is more straightforward, and most users will appreciate it immediately.

Right now, switching to a new Android phone means reinstalling apps and then logging back into each one individually; for some apps, that is a minor annoyance. For others, especially those tied to two-factor authentication, saved preferences, or complex account states, it can take a long time and occasionally fail.

Zero Tap Sign-In changes this. When a user moves to a new Android device, supporting apps will automatically restore their credentials and account state without requiring the user to sign in again. You get your new phone, your apps install, and you are logged in and ready to go—no password hunting, no verification loops, no setting everything up from scratch.

Google is making this a requirement rather than a recommendation, which means developers who want to distribute through the Play Store will need to build support for it. For users, it means the switching experience across Android phones should become significantly smoother by the second quarter of 2027.

The Larger Point

There is something worth sitting with here.

The AI data centre boom is one of the defining infrastructure stories of the last few years. Hundreds of billions of dollars are flowing into GPU clusters and the chips that power them. The world’s biggest technology companies are racing to build capacity. This is what the headlines cover.

What the headlines cover less is the collateral effect on the consumer hardware industry. Memory chip manufacturers are shifting production priorities toward high-bandwidth memory for AI servers because margins are higher and demand is insatiable. Consumer DRAM, the ordinary RAM that goes into phones, laptops, and tablets, is getting squeezed. When supply tightens, prices rise, and manufacturers of mid-range and budget devices start making harder choices about how much memory they can afford to include.

The result is that a person buying a moderately priced Android phone today may end up with less RAM than they would have had two years ago, because the AI industry needed what the memory fabs could produce.

Google’s new requirements are a response to that reality. If phones are going to have constrained memory for the foreseeable future, apps need to get more efficient. The alternative is a gradual degradation of the Android experience for anyone not buying a flagship device.

For users with high-end phones, these changes may not be noticeable at all. But for the much larger population of people using mid-range and budget Android devices, apps that actually respect the limits of the hardware they run on are the difference between a phone that works well and one that crashes, lags, and drains the battery fighting itself.

The AI boom built the best language models in history. It is also quietly making some people’s phones worse. Google just decided that stops here.