From the moment Andy Jassy announced EC2 for Mac to Werner Vogels’ annual ode to builders, this year’s re:Invent was … long? So, so long. But also free! And virtual! With more than a dozen keynotes and hundreds of sessions on tap, there was something for everyone, and also too much for anyone.

Hey, virtual conferences remain a work in progress. But while the “hallway track” of this year’s event was inevitably diminished, AWS still dropped plenty of bombshell announcements and subtle hints about what the next year of cloud might bring.

We’ve covered every day of re:Invent on the ACG blog, and you can read all the analysis from our instructors and AWS guest authors in one place right here. But until next year (or at least, until the fourth week of sessions drops on Jan 12!) here are our top five trends for businesses coming out of re:Invent 2020.

Multi-cloud is inevitable, even for AWS

As IaaS solutions become more commoditized and enterprises find themselves gravitationally pulled toward multiple cloud footprints, cloud providers will have to provide higher-level services and tools to differentiate themselves and hold onto “preferred” status.

Yes, that even goes for AWS, who were coy about the use cases when they announced EKS and ECS Anywhere but eventually confirmed that yes, they’ll support Azure and GCP workloads in their new hybrid container management solution, not just on-premises VMs.

Google’s been here for awhile with Anthos, and Azure has Arc and Sentinel, so this is more a case of AWS finally showing up at the party than breaking new ground. But combined with numerous other announcements for open source and hybrid solutions, it’s clear that AWS sees the future, and it’s a lot bigger than the four walls of us-east-1.

The skills gap hits crisis mode

On the other hand .. you thought learning and managing one cloud provider at a time was tough? Try two or three. It’s why ACG’s recent State of Cloud Learning report found more than 90% of hiring managers struggling to hire qualified cloud professionals. And it’s why AWS announced they are trying to train 29 million new people on their services by the year 2025.

After all, the first thing you migrate to the cloud is the architects, security specialists, and engineers who make cloud transformation happen. They need bite-size, continuous learning that meets them where they are, then moves them forward — so they can move your business forward.

All three major cloud providers have now embraced training and skills development as a key strategic initiative. And it’s time for businesses to do so as well.

Serverless for everyone?

Serverless graduated to the big kids’ table this year, featuring prominently in Andy Jassy’s keynote and getting plenty of love throughout the event. While AWS Lambda has directionally shifted somewhat from the pure FaaS service of years past, new features like container support and per-millisecond billing make it feasible for more workloads than ever before.

That’s because serverless is ultimately not a technical spec, but an idea: own less, build more. And as AWS removes more of the reflexive objections to serverless workloads, we should expect to see more and more teams embracing the advantages of faster innovation and reduced management overhead that serverless provides.

Case in point: the strange case of AWS Proton, a new “serverless and containers deployment manager” service that is really all about enabling central cloud teams to do their best work. It may feel like Conway’s Law-as-a-Service, but Proton is the first sign that AWS is building organizational needs into their developer tools. And that’s a key step in ensuring that the serverless promise will work for everyone.

Cloud economics are undeniable

Sure, updates to EC2 and EBS aren’t the most headline-grabbing moments of re:Invent anymore, but don’t underestimate the faster, cheaper EBS volumes and Graviton-enabled EC2 instances we heard so much about this year — or the caveat-free addition of strong consistency to S3. Faster, cheaper, more reliable compute and storage is at the heart of the cloud promise. And AWS just continues to invest in their core capabilities, with pretty astonishing results.

Particularly with Graviton, AWS is taking advantage of their centralized innovation capabilities, all the way down to their custom silicon, to deliver cost and performance gains that no individual customer could achieve on their own. This is a case of the cloud getting better under the hood, at no cost to you, using the hive mind of feature requests from the collective user base to solve problems you might not even know you have yet. It’s the exact opposite of the way traditional IT works, and it’s more reason than ever to adopt cloud.

ML hogs the headlines, but responsible usage is still the challenge

Pipelines, Edge Manager, Feature Store, JumpStart, Model Monitor, and Data Wrangler aren’t third-tier Marvel heroes — they’re new ML features announced during re:Invent for just one AWS service (Sagemaker), with ML also getting sprinkled on other key services like RedShift and QuickSight.

But don’t be distracted: the most important ML-related announcement of the year is Clarify, a Sagemaker feature and open-source library that promises to help detect bias in ML systems.

Clarify deserves special mention because of the problem it seeks to address: the risk of all these ML solutions encoding and amplifying at scale the biases of the humans who train and deploy them. Clarify uses statistical analysis to help you identify imbalances in your training set before you even create a model, and can be operated as a sanity check on the trained model as well.

It’s reasonable to take this announcement with a grain of salt (who checks the bias-checker for bias?) and so AWS has also open-sourced the code for bias detection and mitigation. Conducting code review in the open air will hopefully give Clarify — and the models it tunes — increasing credibility with AWS users and customers alike.

What’s next?

We’ve got enough re:Invent talks and service launches to keep us all busy learning for months to come. As you plan your learning strategy for 2021, don’t forget to join ACG’s State of the Cloud: Preview and Predictions webinar on Jan 13. Our panel will break down re:Invent, multi-cloud, and the challenges and opportunities cloud teams need to watch out for in the coming year. See you there!