Amazon Q Developer is dead. AWS didn’t say it that way, but the calendar did. New signups got blocked on 15 May 2026, the day after the Sydney keynote wrapped. Full end-of-support is 30 April 2027. If your team standardised on the Q Developer IDE plugin in the last 18 months, you have twelve months to migrate or stop using it.

The Sydney keynote barely mentioned Q. It talked about Kiro.

What actually happened

AWS announced the end-of-support timeline on 30 April 2026. Two weeks later, AWS Summit Singapore (6 May) and AWS Summit Sydney (13–14 May) put Kiro at the centre of every developer-track session, complete with a “Kiro Jam” coding station and a 10x credit bump for students, from 50 to 1,000 free credits. The next day, 15 May, new Q Developer Free Tier and Pro subscriptions stopped being available.

The cancellation flow is what gives the game away. According to AWS’s own migration docs, to cancel an Amazon Q Developer subscription after upgrading to Kiro you “must first enable the Amazon Q Developer console” because it is hidden by default in the new Kiro console. The product hasn’t been renamed. It has been moved into a back room with the lights off.

On 29 May, Opus 4.6 disappears from Q Developer Pro. The newest coding models, Opus 4.7 included, are Kiro-only. AWS isn’t keeping Q on life support. It is starving it.

The wrong take

The take making the rounds is that this is just a rebrand. Q became Kiro, the docs say so explicitly for the CLI, you upgrade, you move on. If you’re a VS Code shop you import your profile and you’re done.

That framing misses what AWS actually shipped. Q Developer was a chat assistant grafted into your IDE. Kiro is a “spec-driven” agentic development environment with a different mental model: you write specs, hooks fire on file save and commit, subagents run for hours. It is not the same product with a new sticker. JetBrains users have to drop the native extension and run Kiro CLI through the Agent Client Protocol. Visual Studio and Eclipse users get nothing native at all — pick a CLI workaround or switch editor.

The teams that bet on Q Developer as their enterprise AI coding standard are not migrating. They are re-evaluating, re-procuring, re-training, and re-writing the internal docs that told everyone which prompts to use.

The real pattern

This is another hyperscaler rebrand-by-deprecation. AWS positioned Q as the answer to GitHub Copilot in April 2024, sunset it 24 months later, and replaced it with a product (Kiro) renamed from earlier work. Google ran the same play: Duet AI was rebranded to Gemini Code Assist in February 2024, and the Gemini Code Assist IDE extensions are themselves deprecated as of 18 June 2026 in favour of Antigravity. Two name changes and two deprecations on the same product in roughly two years. The playbook is consistent: launch fast, brand aggressively, abandon when the architecture doesn’t survive contact with the next model generation.

The cost lands on customers. The Register’s April 2026 piece on AI vendor lock-in put real numbers on the gap between executive belief and operational reality: 90% of executives surveyed thought they could switch AI vendors in four weeks. 41% thought 2–5 business days. Of the organisations that actually tried, only 42% reported a smooth transition. The rest hit failures or significantly more effort than expected.

Switching cost is the part nobody prices in at procurement time. One consultant in the same piece put it cleanly: “switching model vendors is no longer just an API migration. It is context, workflows, and institutional memory.” None of which transfer when AWS hides the old console behind a settings toggle.

What this means for your tooling decisions

Three things to take from watching this play out.

First, separate the model from the surface. The teams least disrupted by the Q-to-Kiro shift are the ones who never let any single vendor own both the model and the IDE workflow. Claude Code, Cursor, and Copilot all run multiple frontier models behind the same surface. When a vendor changes the surface, your team’s muscle memory survives. When a vendor changes the model, your prompts survive. Couple both and a deprecation gets expensive.

Second, price the deprecation risk into the tool choice, not just the sticker price. Q Developer Pro at $19/user/month looked cheap against Cursor at $20 and Copilot Enterprise at $39. The cost was never the subscription. It was the four-week migration nobody scheduled, the internal training nobody redid, and the second-class IDE support nobody noticed until they had to ship something.

Third, assume a 24-month half-life on any hyperscaler-branded AI dev tool. The brand layer moves slower than the underlying model architecture, and AWS, Google, and Microsoft have all shown they will throw the brand away when the architecture changes. If you wouldn’t rebuild the same workflow in a year, don’t standardise on it.

Q Developer wasn’t a bad product. It got obsoleted by the speed of its own market. Kiro will probably get the same treatment in 2027 or 2028 when the next shift renders spec-driven IDEs into yesterday’s idea. Plan accordingly.

If your team standardised on Q Developer and you need to make the migration call without lighting twelve months on fire, talk to us.

Sources