From the perspective of someone not working in a major US-based company, the costs for some of the tools in the ML/Data Science/Data Engineering space are kinda bonkers tbh.

Curious what your perspective is on cost/difficulty of self-hosting & managing OSS tools vs paying for commercial versions.

If you’d like I’m happy to provide more details but it’s much harder to motivate the ROI from better tooling vs the (more visceral) added budgetary burden. In my experience, it’s much easier to motivate costs if a) it’s clear how it will lead to increased revenue (CRM tooling for example), b) the project has a high-level sponsor that is technologically motivated, or c) there’s a major incident that gets the attention of upper management/execs that can motivate the expense. With that said we’ve had quoted costs from companies that have been insanely absurd in some cases (“ML monitoring software for the cost of 2 full time staff engineers year for up to 2 ML models”).

A former colleague claimed that the average discount Oracle gave to Norwegian customers for Enterprise-type was somewhere around 70%. I know other software companies that struggle to get into the Norwegian market because their pricing is based on what American enterprise customers are willing to pay. Although average pay in Norway is high, the wage scale is more compressed which means the math is different on build-vs-buy decisions. With FAANG-level pay, software looks really cheap in comparison. But the US is big, I imagine there are plenty of states, plenty of companies, with lower pay and different math. I used to work in government, which was practically allergic to buying software - managers seemed incapable of comprehending that pay is an expense too. I’m in the private sector now where they actually try to do the math.