Case study of why open-source stacks are preferable: a friend is founder of a B2B SaaS startup (it's a product for salespeople, not engineers, but SaaS nonetheless). A customer asked if they could do a bring-your-own-cloud installation on Azure. But the product depends on many AWS things such as SQS, Kinesis, Lambda, and Step Functions internally. Customer lost 😞

Wow. I didn’t know that BYOA could be a requirement of Sales teams as well.

The potential customer has a lot of sensitive data that they don't like being in other people's clouds, and this particular product requires being "trained" on some of your data; as far as I understood it, it was more of an enterprise-wide policy than sales-specific

Hey there! Curious about this -> "trained on some of your data". Does your data mean 1. train models based on the SaaS product's 'user behavior' to make the product better/grow or 2. proprietary data ( essentially owned by that customer )

2)

oh wow

It’s a SaaS AI product with LLM’s, because of course it is (:

Question not relevant to the original point - If it runs on the customer’s cloud is it still technically a SaaS? Wouldn’t it be “installed software at that point? Does “SaaS with installed on prem” have a specific name?

Look at astronomer.io— when I had a brief stint there they called it “hybrid SaaS” where the data plane ran in customer cloud and the control plane was standard SaaS in the vendor’s cloud.

I like the term “hybrid SaaS”. Having the control plane in an entirely different cloud than the data plane sounds like it creates all kinds of new “fun” problems.

It did create several problems. But I think if you had to do it, Open Cluster Management and a very good K8s Operator is the way to go. It’s on our roadmap to do that for our LittleHorse workflow engine for 2026 😉