Google’s Gemini Go Brings AI to Android’s 2GB Budget Phones

Google rolled out Gemini Go on June 4, 2026, bringing a version of its AI assistant to Android Go devices with as little as 2GB of RAM. Apple Intelligence requires 8GB to function. Google’s own flagship Gemini Intelligence tier needs 12GB. For hundreds of millions of users on sub-$100 phones in India, Nigeria, Indonesia, and Latin America, the new assistant arrives on hardware that costs a fraction of what those platforms demand, in markets that most AI coverage has barely acknowledged.

The rollout also ends Google Assistant Go, the cut-down voice assistant that had served this hardware tier since Android Go launched in 2017. What replaces it is something considerably more capable, and its arrival across more than 180 countries signals where Google has decided the actual scale of the AI market lives.

The End of Assistant Go

Gemini Go replaces Google Assistant Go on any Android Go device running at least 2GB of RAM, the certified minimum for new hardware since Android 13 (Go edition) arrived in 2022. The transition had been signaled for months. Google announced in March 2025 that Gemini would replace Assistant across Android by year-end. It didn’t happen on schedule: Anish Kattukaran, chief product officer at Google Home and Nest, wrote in December that the company was “adjusting our previously announced timeline to make sure we deliver a seamless transition,” pushing the full migration into 2026. The Android Go tier is now the first complete segment where the old assistant is gone and has nothing running alongside it.

Finding Gemini Go on an eligible device requires no separate download and no additional install. The experience runs through the existing Google Search app, consistent with the Go platform’s resource-first design philosophy. The official Android (Go edition) developer documentation spells out the core principle: every app and service on Go devices is built to minimize memory, data, and storage demands. Gemini Go follows it. A long press of the Home button or Power key on a supported device launches the assistant.

Google is pushing the update gradually, so compatible devices won’t all receive it simultaneously. Updating the Google app through the Play Store is the most reliable way to check eligibility. Gmail Go and Maps Go already exist as lightweight Go-tier counterparts to their full versions. Gemini Go fits that naming pattern but not the logic: those apps run alongside full alternatives a user can reach through a browser. Gemini Go replaces the only assistant this class of hardware has ever run, with no parallel version to fall back to.

The Scale of the Android Go World

Where Budget Phones Dominate

Android Go launched as a category in December 2017, built around a specific observation: entry-level phones already accounted for more than half of all global device shipments at the time. That figure came from Google’s own tracking. The category has grown since. Android 15 Go Edition, released in March 2025, supports devices across more than 180 countries, and where those devices concentrate explains why the Gemini Go launch carries weight that its hardware specs alone don’t convey.

According to Android market share data compiled by Command Linux, Android commands overwhelming smartphone adoption across the world’s largest developing regions. The brands moving sub-$100 volume in those markets, Tecno, itel, and Transsion, are the ones Samsung and Apple don’t directly compete against at this price point. Android Go is the operating system those devices run.

• 95.21% Android smartphone adoption in India
• 85.15% Android market share across Africa
• 90.9% Android penetration across Latin America
• 16,000+ device models certified for Android 15 Go Edition

The Feature Gap Those Users Lived With

In each of those regions, Android Go devices frequently serve as first smartphones, not upgrades from older premium hardware. Google’s developer documentation for the platform describes the target as users “transitioning from feature phones.” Those users, until June 4, had Google Assistant Go: capable of calls, alarms, and basic queries, but unable to hold a contextual conversation, reference an uploaded file, or parse a request combining location, availability, and amenity requirements. Gemini Go handles the second category of query; the old assistant never did.

In markets where Android’s share runs above 85%, Gemini Go is now the most capable AI assistant available to the majority of smartphone users, because Apple Intelligence and Samsung Galaxy AI don’t reach the hardware they’re on. The vacancy at the 2GB tier wasn’t an oversight; it was the space Google has moved to occupy.

What Gemini Go Does on 2GB Hardware

Google’s own launch description is measured: “a streamlined version of Gemini designed to help you stay connected and get things done, even on devices with lower storage.” The feature set goes further than that framing implies.

• Voice calls and text messages on demand
• Drive-time estimates and multi-part local queries (Google’s documented example: “Help me find a ramen restaurant open for lunch on Tuesday with an EV charger nearby”)
• Alarm setting and calendar event creation
• Music playback by mood or activity (“Play quiet acoustic songs for a dinner party”)
• Document and photo uploads to add context to conversations

The file upload capability arrived roughly two years ahead of what most industry observers had projected for this hardware tier. Sending a document or photo and asking a contextual question requires the model to hold image or text alongside conversation history, a meaningful computational demand on a 2GB device. The mood-based music queries are a smaller technical ask, but both represent a genuine step beyond literal command execution and closer to the conversational AI that Pixel owners have had since 2023.

One detail Google hasn’t provided: what the overlay looks like. 9to5Google noted on June 3 that no interface screenshots accompanied the announcement. What is confirmed is that Gemini Go lives inside the Google Search app rather than a standalone Gemini application, keeping memory overhead compatible with the platform’s limits.

The RAM Gap Nobody Else Is Crossing

Every major AI platform has set a hardware floor. The spread between them is stark.

For Apple, AI has no budget tier. Every device running Apple Intelligence sits at 8GB or above, corresponding to hardware priced at $999 and higher. Apple has tied its AI feature set to premium devices since launch, treating it as a differentiator rather than platform infrastructure. Samsung’s Galaxy AI begins with the S24 series, leaving its own Galaxy A-series budget lineup excluded. Neither company has built a reduced-capability AI tier for entry-level hardware, and neither has signaled plans to.

Google is the outlier. Android powers an estimated 3.9 billion active devices globally, with the sub-$100 segment representing a footprint Apple and Samsung’s AI bets don’t touch. Running Gemini Go at 2GB, six times below Apple’s minimum and one-sixth of what its own Gemini Intelligence tier requires, is Google treating its AI assistant as platform infrastructure rather than a premium selling point.

Google’s Bet on the Next Billion Users

What Gemini Go Can’t Do

Gemini Go is a replacement for Google Assistant Go, not a compressed Gemini Ultra. Advanced reasoning tasks, Deep Research, Gemini Advanced capabilities, image and video generation, and Pixel-specific system integrations are absent. The experience is built for communication, scheduling, local information, and basic file context: the tasks that matter most on a primary device with limited storage and often variable connectivity.

Google hasn’t published a side-by-side breakdown of what Gemini Go includes versus what it excludes. The “Go” branding does that work implicitly: it tells the user this is a tiered product calibrated to the hardware, without specifying which capabilities didn’t make the cut. Google chose not to call this simply Gemini on Android Go devices, a naming decision that maintains a visible hierarchy between its premium and budget product lines. The technical constraints driving the tier are real; the decision to mark the tier with a suffix is a choice, and it signals the hierarchy is meant to last.

The Unified Platform Logic

Platform unification is the broader frame. Google’s 2026 migration is moving every Android surface onto some version of Gemini: Wear OS, Google TV, Android Auto, smart speakers, and now the entry tier. The company’s I/O 2026 keynote roadmap for Gemini across Android 17 and agentic features mapped a full stack that Gemini Go now fills from the bottom. There is no longer a gap in Google’s Gemini coverage by device tier.

The commercial logic behind this isn’t about competing with ChatGPT for power users. Every Android user, whether on a Pixel 10 or a Tecno Pop budget phone in Lagos, is a potential search query source for Google’s advertising business. Keeping users inside a Gemini interface keeps them inside Google’s infrastructure for those queries, rather than routing them to a third-party assistant that sends searches elsewhere. Mobile market data from Digital Applied’s 2026 mobile OS report puts the average Android user’s annual app spend at $13, against $63 for the average iPhone owner. Google doesn’t rely on these users buying apps. It relies on them doing searches.

In Lagos, Dhaka, or Bogotá, an AI habit formed on a 2GB phone is a Google search habit, and the June 4 rollout establishes that habit before a competing assistant can.