Shopify drops React Native and returns to native, with AI in the calculation

Shopify has announced it is dropping React Native to rewrite its core apps in native Swift and Kotlin. According to an InfoQ report, Head of Mobile Mustafa Ali re-evaluated the company's commitment to the framework after the leap in quality of AI models, and concluded that the benefit-to-cost ratio of maintaining native codebases per platform had become higher than that of using an abstraction layer.

The detail that matters for anyone building software: AI enters here not as a product feature, but as a factor that changes the architecture calculation. It's the kind of decision that used to be settled in a maintenance-cost spreadsheet, and that has now gained a new variable.

The turnaround of a company that was a symbol of React Native

The weight of the news comes from the history. In 2020, Shopify declared that React Native was the future of its mobile architecture. In 2025, it reaffirmed that commitment after successfully completing the migration of all its apps over five years. The justification at the time was threefold: not building the same feature twice, talent portability across platforms, and spending more time delivering value instead of chasing feature parity.

The numbers the company reported backed the choice: screens loading in under 500ms and sessions with more than 99.9% crash-free. Hot reloading and TypeScript were cited as productivity accelerators. Even so, Shopify never fully ruled out native: features that depend on device hardware (2D/3D scanning, running on-device models), home screen and lock screen widgets, Apple Watch apps, App Intents, Siri Shortcuts, and long background jobs always stayed on the native side. The motto was "native and React Native," not "native or React Native."

The two factors that tipped the balance

A little more than a year later, the company re-evaluated everything. According to InfoQ, two factors pushed the decision.

The first was a heavy refactor on the horizon: adopting React Native's New Architecture. That work would require revisiting native module integrations, rendering, and the boundaries between shared and platform-specific code. Since it was already going to be a major engineering effort either way, it was worth stopping to reanalyze the options.

The second was the pace of improvement in AI coding models, agents, and harnesses. A proof of concept run by a single engineer in one week went so well that six engineers rebuilt the Shop app from scratch through to production in 12 weeks. That result reinforced the choice of a greenfield rewrite over migration.

How AI was actually used

This is the point of interest for anyone working with coding agents. Shopify standardized on modern declarative frameworks, Apple's SwiftUI and Google's Jetpack Compose, precisely because the structural and mental-model symmetries between the two platforms help guide autonomous agents, while keeping each platform's native idioms.

But the smartest detail is architectural: the company isolated the factors of change (the mobile platform and the UI) from the business logic and state management, which are stable. This served to get around the slow feedback loop of mobile simulators. In practice, the AI models started triggering actions, manipulating navigation state, and checking results in milliseconds, on the desktop, without rendering UI or launching a simulator. Simulators were reserved for automated end-to-end integration passes.

It's worth noting what Shopify itself admits: despite the advance in models, the rewrite still required considerable engineering involvement in architecture, planning the agents' work, validating task completion, and fixing errors. It was not a case of "asking AI to rewrite the app."

The numbers behind the rewrite

The company reported concrete gains when comparing native to the old React Native setup:

• Cold-start latency dropping 23% on iOS and 50% on Android;
• Session stability rising from 99.5% to 99.95% (a 10x reduction in crashing sessions);
• Release build 75% faster on Android;
• Native feed scrolling sustaining 120 FPS on Android without extensive manual optimization.

The caveats the community raised

The announcement, as expected, sparked debate. The strongest criticism, according to InfoQ, is methodological: the performance and productivity results were measured against React Native's old architecture, not against the New Architecture. In other words, it remains an open question whether migrating to the new architecture would have achieved similar results without switching stacks.

The other tradeoff developers cited is the old problem of over-the-air updates. Going full native, publishing a new version now depends on the app store review process, which reduces release frequency. One developer raised an interesting cultural provocation: the perception that OTA is a decisive argument in favor of cross-platform may come from a web culture, in which changes are cheap and fast and teams feel less pressure to test everything before shipping. In the native world, since each release is harder to roll back, teams end up investing more in release tooling, thinking more carefully about changes, and testing more, which leads to larger, more spaced-out releases. His conclusion was direct: in the end, both approaches work.

What changes for those building software in Brazil

For mobile teams here, the takeaway isn't "React Native is dead." The point Shopify is signaling is that the AI variable has entered the architecture calculation, and it strikes precisely at the historical cost of native: writing and maintaining two codebases. If a coding agent can move well between SwiftUI and Jetpack Compose leveraging the symmetries between them, "don't build the same feature twice" loses part of the weight it had in 2020.

But Shopify's case comes with context that rarely replicates: a large team, a greenfield rewrite, a decision that would have run into a heavy refactor anyway, and the discipline of isolating business logic from the UI layer to give agents fast feedback. That last point is the most transferable to any project, with or without AI: separating what changes from what is stable is good architecture before it's a trick for agents.

What remains open is whether the math holds up for smaller teams, which don't have the engineering volume to sustain two native codebases even with AI helping, and which feel the pain of losing OTA much more acutely. Shopify itself is still in the middle of the transition: it plans to complete the Point of Sale and Inbox apps over the next year, and the main Shopify app (with more than 300 screens, home screen and lock screen widgets, an Apple Watch app, and Siri Shortcuts) this year. It's worth watching how these more complex cases play out, because that's where the promise of AI-driven cost reduction will really be tested.