Self-hosted pricing details
I was looking at the pricing page but I'm a little confused about the info there. For example, I have self-hosted selected, yet both Pro and Enterprise has vCPU and seats sliders and says "~52M executions of 100ms per month". Does this mean that you are charging based on the server specs we are self-hosting Windmill on?
13 Replies
Correct, we are charging based on usage and use size of the infra allocated to workers on prod as the relevant proxy for it
So if we decide to get a bigger server to handle multiple services in one server, including windmill, we end up paying more for nothing and prices will keep going up regardless what we actually do with windmill. Alright, seems like a completely no go pricing model for us and we will stick with the open-source, thanks for the answer.
That's not exactly how it works, you can define limits of resources for the workers in docker compose and helm and this is actually what we count rather than the servers it is hosted on directly.
Yes, of course. The issue is that your pricing model is basically taxing our investments in our infra to be able to run software faster. Seat-based pricing works because that is directly tied to human usage but, in our case, having to pay extra for wanting our jobs to run faster is something we are not looking for at the moment.
Sure, in this case it's not so much necessarily about running faster (unless you are very much compute bounds for some jobs) but running more jobs, having more throughput and for most of our customers, including the ones that use very large clusters, it seemed fair that when they used windmill at larger scale they would pay more as they derive more value out of it and it's exponentially harder to make software that run at that scale reliably. But happy to see the open-source is sufficient for your needs, it's the goal that users have the choice and only need to pay for things they get a clear return on investments on.
I have to say, I had a similar question. I'm actually impressed with your pricing model - it's so good, in fact, that I'm adjusting my thinking instead of moving on. Windmill is an extremely high value product and is unbelievably well positioned to take advantage of the generative AI adoption needs of all businesses over the next few years. Your pricing model is appropriate.
@rubenf regarding adding the core package to enterprise cloud, when you say without restrictions, do you mean user, cpu restrictions? Does it mean for $14.4k/year we get unlimited users, etc.?
Or is $14.4k for 1 seat?
it's neither, the 6k is the fixed price that offset the price we pay to spawn a dedicated cluster for you which is indepdent of the number of workers you have on it (because some parts have fixed cost, database, EKS, and some manual labor)
The rest is per-seat and per-vCPU
you can see the price per-seat and per-vCPU on that screenshot, +$400/yr/seat, +1k/y/vCPU
Ok, that makes sense. So the 6k is a part of the 8400
It's a bit confusing the way it's presented
perhaps a FAQ on the pricing page?
How would you have preferred it to be presented ?
I think if the core package was presented above the sliders with the price above the vCPUs ($6000/yr)
that way the price is clearly composed of three elements
Price $8400/yr
1 Core package /t $6000/yr
2 vCPUs /t $1000/yr/cVPU
1 seats /t $400/yr/seat
perhaps it makes sense to most people and I'm an outlier
I think if we add "+" next to $6000 it might go a long way