Amazon Just Approved Claude Code for 800,000 Employees — Running on Bedrock
Amazon's internal policy on AI coding tools was never subtle: use Kiro or get a waiver. For months, engineers who wanted Claude Code had to file exception requests, argue with managers, and in some cases just use it anyway through unofficial channels. Now the exception has become the rule.
Business Insider obtained an internal memo from VP Jim Haughwout confirming that Claude Code is now available to every Amazon corporate employee, effective immediately. OpenAI's Codex follows on May 12. Both tools run on Amazon Bedrock — which means all inference stays inside AWS rather than routing to third-party APIs. Kiro, Amazon's in-house tool, is no longer the only approved option, though it's still used by 83% of Amazon engineers.
The reversal did not come quietly. Around 1,500 Amazon engineers signed internal petitions pushing for formal Claude Code access, arguing the restrictions were putting Amazon at a competitive disadvantage against companies that had already embraced AI coding agents. One internal comment captured in reporting from Inc.com was blunt: "Customers will ask why they should trust or use a tool that we did not approve for internal use." On Reddit, Amazon employees claimed they'd been using Claude Code informally for a year through workarounds — which says something about how users respond when you block the tool they actually want.
The timing is notable. Amazon has invested up to $50 billion in OpenAI and up to $25 billion additional in Anthropic on top of prior pledges. The company was simultaneously one of Anthropic's largest investors and one of its loudest internal blockers. That contradiction was apparently sustainable right up until it wasn't — when enough engineers complained loudly enough, and when the competitive cost of staying on a less-preferred tool outweighed whatever internal objections had been driving the policy.
What Bedrock Changes About the Deal
Amazon running Claude Code on Bedrock rather than the public Anthropic API is the detail that matters most for anyone watching how enterprise AI infrastructure is being built. It means Amazon controls model selection, data residency, and billing — and it means Anthropic's revenue from Amazon's internal usage flows through AWS, not as direct API revenue. That's a tighter integration than a typical API customer relationship. It's also a model for how cloud providers can offer AI coding tools without ceding control to third-party inference providers.
For practitioners, the architecture question is worth sitting with. When your employer runs a coding agent on your own cloud infrastructure, the data residency and latency advantages are real. When that same employer also owns the model provider through investment, the vendor relationship gets complicated in ways that aren't obvious from the outside. Amazon's engineers now have production access to Claude Code on infrastructure Amazon controls, funded by a bet Amazon made on both sides of the competition.
The Kiro Problem Nobody Is Talking About
Amazon built Kiro for internal use and spent years requiring engineers to use it. The fact that 83% of engineers still use it tells you something: the tool has inertia, switching costs, and integration depth that a new Claude Code deployment won't immediately overcome. The more interesting question is whether that 83% drops meaningfully over the next six months, and what Amazon's internal data shows about relative performance.
If you're at a company that's been cautious about AI coding agents, the Amazon rollout is a natural experiment you should be watching. Amazon cited security and performance concerns as reasons to prefer Kiro for months. Now those concerns are apparently resolved — or were always manageable, or were outweighed by the cost of losing engineers to better tooling. If Amazon publishes any internal data comparing Kiro versus Claude Code outcomes, that's the strongest proof (or counterevidence) the industry will get that these tools are ready for production.
The deeper signal is about adoption velocity. Amazon's internal AI tracking showed the company monitoring adoption rates closely — a level of instrumentation that suggests leadership was watching this metric like a business KPI, not just a technology preference. The rollout to all employees suggests AI coding tool adoption is no longer optional at one of the world's largest engineering organizations. By extension, developers who are still resisting these tools may find themselves increasingly outside the mainstream of how software is actually being built.
The Bottom Line
Amazon just validated Claude Code as production-ready by deploying it to 800,000+ corporate employees. When your biggest investor — with every strategic incentive to protect its own competing tool — decides the cost of blocking Claude Code is too high, that's a market signal worth paying attention to. This isn't a startup experimenting. It's Amazon concluding that Claude Code is the tool that wins, and that the right move is to get out of the way.
For the rest of the industry, the question is what "Amazon uses this" means for procurement conversations at companies that take their cues from Amazon's technical choices. The answer, probably, is: quite a lot.
Sources: Business Insider, Inc.com, Times of India