The Cloud Pod | Weekly AI & Cloud News on AWS, Azure & GCP · Justin Brodley, Jonathan Baker, Ryan Lucas and Matt Kohn | Cloud Computing & AI News

319: AWS Cost MCP: Your Billing Data Now Speaks Human

·1 hr 36 min·5 clips

Matt Garman says replacing junior developers with AI is "the dumbest idea ever."

1. The Cloud Pod episode 319 centers on AWS Cost MCP and other cloud announcements from August 26th, 2025. 2. Hosts Justin, Jonathan, Ryan, and Matthew drive the discussion, and Matt Garman is the AWS CEO whose comments shape the AI labor segment. 3. The episode asks what these AWS, Google, and Amazon updates actually mean for billing, developer workflows, and database operations. 4. The first major story is xAI suing Apple and OpenAI over alleged anti-competitive chatbot distribution in the App Store and iOS integration. 5. The hosts compare that dispute to older platform fights involving Google, Microsoft, Apple, and mobile search placement. 6. The next topic is Matt Garman's warning that replacing junior developers with AI is counterproductive because it creates a future talent pipeline problem. 7. The hosts note Garman's claim that more than 80% of AWS developers use AI tools for unit tests, documentation, and code writing. 8. The conversation around AI hiring shifts into critical thinking, recursion, and why broad problem-solving matters more than syntax alone. 9. A smaller AWS feature follows: a Count Tokens API for cloud models that estimates token usage before inference calls. 10. The hosts frame that API as a guardrail for cost control, prompt engineering, and avoiding rate-limit errors. 11. AWS then introduces an open-source MCP server for billing and cost management in AWS Labs' GitHub repository. 12. The hosts say the most practical question is what "EC2 Other" means on a bill and whether MCP can explain it clearly. 13. The episode also covers RDS for DB2 read replicas, including up to three replicas per database instance and cross-region options. 14. Amazon RDS for Postgres delayed read replicas become another major topic, with a configurable lag for rollback, accidental deletion recovery, and ransomware scenarios. 15. The hosts question how often a team would notice a bad production change within the replica delay window, and they compare the feature to backups and point-in-time restore. 16. Prime Day 2025 gets a long metrics segment with 524 million commands to robots, 87,000 Inferentia and Trainium chips, 626 billion SageMaker inference requests, and 1.7 trillion Lambda invocations. 17. The tone stays informal and argumentative, with hosts interrupting each other, joking about naps, and reacting to giant AWS numbers in real time. 18. The format mixes news roundup, technical commentary, and side conversations about recursion, object-oriented programming, and security tradeoffs. 19. Listeners who follow AWS billing, AI tooling, or database reliability will get the most value. 20. Listeners looking for a tightly edited narrative or a single-topic deep dive may skip it.

As heard by us

AWS CostMTP, AI access politics, and the practical split between cloud and laptop tools.

AWS CostMTP is the anchor for a cloud-and-AI conversation about access, runtime boundaries, and the way distribution still shapes which tools people actually reach for when platforms and operating systems start bundling AI features.

Read the full review in PlayNext →

Why you'd press play

Cloud gossip with actual billing teeth.

Read the full recommendation in PlayNext →

Listen to the show on

Apple Podcasts Spotify

Play clips from this episode More from The Cloud Pod | Weekly AI & Cloud News on AWS, Azure & GCP