Good Morning!
It’s over. re:Invent is finally over. If you need me, I’m going to be sleeping for a week. If you need a trip report / additional announcement snark coverage, be sure to visit shitposting.ai which does exactly what you think it will do.
Podcasts
Last Week In AWS: Welcome to re:Invent, Where the Roadmap Is Made Up and the Quotas Don’t Matter
Screaming in the Cloud: AI Agents, Enterprise Risk, and the Future of Recovery: Rubrik’s Vision with Dev Rishi
Choice Cuts
Introducing Amazon Route 53 Global Resolver for secure anycast DNS resolution (preview) – They’re renaming Route 53 Resolver to "VPC Resolver" while launching Global Resolver, because nothing says customer clarity like forcing everyone to relearn your product taxonomy, like they just did with Security Hub –> Security Hub CSPM –> a new thing called Security Hub. "Y’know our trash naming? How can we make that somehow worse?"
Introducing AWS Lambda Managed Instances: Serverless simplicity with EC2 flexibility – So now you have serverless functions running atop servers plus a management fee of 15% and additional complexity? Aren’t there a bunch of open source projects that purported to do this that nobody uses because, honestly, why would you? I’m gonna have to sit with this one for a while.
AWS announces preview of AWS Interconnect – multicloud – This is either transformative or a waste of everyone’s time, and it’s impossible to tell which because the one thing that matters most to settling that question is "what’s the price." They aren’t disclosing it yet, so at the moment it occupies a superposition of "excellent/crap." Please collapse the waveform so we know which one it is.
Introducing AWS Transform custom: Crush tech debt with AI-powered code modernization – The sign at re:Invent said "eliminate tech debt forever with AWS Transform." I will bet you a freaking house that it won’t. Any takers? It’s a good feature but the marketing is so far over its skis that it’s already back in the lodge sitting in the hot tub by now.
Amazon CloudWatch introduces unified data management and analytics for operations, security, and compliance – I think this is priced reasonably, and I think I’ve activated it properly in my test account, but honestly it was so confusing to enable that I’m not entirely sure about either of those. Yes, the problem may well be that I’m an idiot—but it’s not likely that I’m the stupidest customer that AWS is going to encounter with this feature, so I’m a useful test case.
Amazon EC2 P6e-GB300 UltraServers accelerated by NVIDIA GB300 NVL72 are now generally available – AWS just dropped hardware so expensive you need to "contact your sales representative" to even ask the price, but it doesn’t matter because it’s not available to most customers who don’t rhyme with "Shmanthropic."
Introducing AWS AI Factories – AWS will now rack servers in your data center, charge you cloud prices for on-premises hardware, and call it innovation. Nothing screams "digital sovereignty" quite like letting AWS control the infrastructure while you pay the power bill and real estate costs. Oh, and if its upstream connection to a region goes down for too long it basically becomes a space heater.
Introducing AWS DevOps Agent (preview), frontier agent for operational excellence – An AI agent that investigates incidents at 2 AM sounds great until you realize it’s learning from your messy runbooks and spaghetti code, so it’s going to be just like your worst engineer if you aren’t careful with it. I want to like this, but I’ve seen so many abortive attempts at building such things over the last fifteen years that my natural kneejerk response is "there’s no way this can work." Prove me wrong, please?
Amazon S3 Storage Lens adds performance metrics, support for billions of prefixes, and export to S3 Tables – I’m really happy about how much of my own crapass tooling S3 Storage Lens has managed to deprecate. This is a great release, and I’m excited to suddenly present to my colleagues as a Dark S3 Usage Wizard.
Build multi-step applications and AI workflows with AWS Lambda durable functions – "We have Step Functions at home." This is going to take me kicking the tires on before I have an educated opinion, because last week was just a smidgen busy.
Amazon S3 increases the maximum object size to 50 TB – Storing a 50TB file in S3 means that this unlocks ~4 excellent use cases, and thousands of absolutely monstrous ones like "the data warehouse is now a SQLite file." Guess which one I’m more excited about?
Amazon S3 Tables now offer the Intelligent-Tiering storage class – S3 Tables gets Intelligent-Tiering, which is genuinely useful for analytics workloads with unpredictable access patterns. Be aware though, you’ve gotta not reset the timer every time your workload runs, so you’ve gotta be smart about how you’re querying this database.
China-nexus cyber threat groups rapidly exploit React2Shell vulnerability (CVE-2025-55182) – A huge discount on your AWS bill once you leverage this to run in others’ accounts. Previously that kind of bill pro tip was only available to Azure customers.
Introducing Database Savings Plans for AWS Databases – AWS Database Savings Plans are here. RDS, Aurora, ElastiCache, Neptune, DocumentDB, DynamoDB- the whole family. I’d like to thank the Academy, my parents, and the heat death of the universe which we almost reached first.
I’ve been asking for this since 2019. Twenty. Nineteen. My hair was a different color. I had different enemies. The world was younger. Some of you in this room weren’t even in cloud yet. You were innocent. Unscarred by the Reserved Instance management experience.
For six years – SIX YEARS – AWS said "just use Reserved Instances for each service separately." That’s like telling someone to use different currencies for each type of purchase. "Oh you want to buy groceries AND gas? Better maintain separate wallets with separate exchange rates and separate commitment terms." Totally reasonable. Very normal.
Compute got Savings Plans in 2019. The SAME YEAR I started asking for database coverage. But databases? Databases had to wait until 2025. That’s six years. And look – I’m not saying AWS moves slowly. I would never say that. But I will say that geological epochs have been sliding into my DMs asking if AWS needs motivational coaching.
You folks have heard me complain about this so many times you have PTSD. Every time a database service launched, every time Savings Plans came up, every time someone asked about cost optimization – there I was, like Statler and Waldorf from the Muppets, heckling from the balcony. "Where are the database Savings Plans?" I’d ask. "Wouldn’t it be nice if databases worked like compute?" I’d muse. "Is this the year?" I’d wonder, hope dying a little more each time.
Some of you don’t remember the before times. Those were dark days. Days of managing seven different Reserved Instance portfolios like some kind of medieval warlord, each with their own commitment terms, their own utilization tracking, their own special blend of vendor lock-in. It was like having seven different gym memberships that you couldn’t transfer between locations.
For context – this is the pricing commitment flexibility we’ve had for EC2 since 2019, but for the databases that power literally everything you actually do. Because let’s be honest: your compute is replaceable, but your database is your database. That’s where your actual business lives.
Reserved Instances were commitment without flexibility. A prison of your own cost-optimized making. Now we have commitment WITH flexibility across services. You can move between Aurora and RDS without sacrificing your savings commitments. You can shift from ElastiCache to DynamoDB without starting over. The year is 2025 going on 2026. We can prompt AI to write code. We can finally manage database commitments like rational human beings. What a time to be alive.
I started asking for this as a young man with dreams. I’m now old and bitter. But I have Database Savings Plans. So I guess everything worked out.
Anyway – Database Savings Plans: they’re here, they’re great, they only took six years and the collective screaming of the entire FinOps community. Thank you AWS. No really, this is huge, this is meaningful, this will save companies real money and real operational headaches.
But also: what took so long, you absolute maniacs?
… and that’s what happened Last Week in AWS.
