I don't remember the numbers (90% is probably a bit exaggerated) but our savings of going from Atlas to MongoDB Community on EC2 several years ago were big.

In addition to direct costs, Atlas had also expensive limitations. For example we often spin up clone databases from a snapshot which have lower performance and no durability requirements, so a smaller non-replicated server suffices, but Atlas required those to be sized like the replicated high performance production cluster.

Was it? Assuming an M40 cluster consists of 3 m6g.xlarge machines, that's $0.46/hr on-demand compared to Atlas's $1.04/hr for the compute. Savings plans or reserved instances reduce that cost further.

There's definitely MongoDB markup, but a full 33% of their bill was AWS networking costs that have nothing to do with Atlas.

Highly doubt that. MongoDB has 5000 well paid employees and is not a big loss making enterprise. If most of the cost was pass through to AWS, they’d not be able to do that. Their quarterly revenue is $500M+ but also spend $200M in sales and marketing and $180M in R&D. (All based on their filings)

You can look at this particular bill and observe that more than 50% of the cost was going to AWS.

If they’re a reseller of AWS, which they will be, they decide the rates that get charged.

Yes, and my point is that this customer switching to running their own MongoDB instances on EC2 like Atlas does would reduce the bill by less than 50% because the rates that they are charging mean that their cut is less than what AWS is getting from this customer.