That article is garbage, basically boils down to “did you know that this thing that’s not suitable at all for a model where you pay by milliseconds doesn’t work well on serverless?”

Did that person even read what Lambda etc. are useful for? Like, what’s their advantage? Hint: it’s not 15+ minute execution nor is it having a shared state with previous instances.

You know what works?

A container.

(…)

Runs anywhere

In theory: yes. In practice: not always. Last week I’ve started debugging an image which in fact does not run everywhere. I’ve deployed it on 3 different systems, same Docker command, same Docker version, 3 different outcomes.

«Bandwidth is especially overpriced. $0.55/GB egress in 2025? Why?»

Because you get so much else for free. They have to charge for their services somewhere.

You end up paying for everything, but the costs are artificially shuffled around.

Containers aren’t an alternative to serverless computing. That’s like saying “roads are a scam, I’ll just drive my car instead.” The only serverless compute service they mention, Lambda, literally allows you to deploy custom containers.