Cloud costs are in a bubble

Edit: An interesting discussion developed at HN

Cloud costs are out of control. Many investors extrapolate cloud profits into the future but fail to consider the possible margins shrinking as prices will inevitably go down. In this post, I'll lay out why I think margins will shrink.

But first - who am I to talk about it? Even though the name of this newsletter is "The Investing Desk", I'm only a part-time investor and most-time a software engineer. I've been CTO of 2 startups, and I've been a tech lead on bigger companies (such as Monday.com).

What are cloud and managed services

Cloud, in the basic sense, is renting computing from other providers, but today it also refers to managed services.

As one can go deeper into how managed his services are, we can refer to it as being more "cloudy".

Renting simple VMs from AWS is not like paying for a managed SQL service. And serverless services are not like full PaaS services like firebase. The deeper you go on the managed service scheme, the more you'll pay, and the more you'll be locked into the platform.

The TCO collapse

The leading KPI for decision-making in choosing a cloud/managed service is the Total Cost of Ownership calculation. The manager tries to incorporate all the costs coming with self-hosting the service vs. paying for a managed one.

The thing is, as a CTO myself and as someone who pitched solutions to other executives - you inevitably integrate vague measures into the calculation.

You start with the constants: how much the service costs vs. its server costs self-hosted. Then you add less accurate measures: "It will take 2 engineers full time to maintain, which costs ~$300,000 a year". Then, it becomes weird: "Actually, no one will be full-time, but 5 engineers will spend 20% time a year on emergencies and then another 10% time to be familiar with the system, so that amounts to ~$250,000 a year."

And before you know it, you have an excel diagram with a field called "attention span of Ops team - 10%, team of 14, that's ~$400,000 a year". I've been there, I've seen that - it's bullshit.

These TCO calculations may seem scientific, but more often than not, they are tools to excuse a decision the management has already made.

And in times of stress - like we may have go into right now - the bullshit floats, and better decisions are going to be made.

MongoDB Atlas example

A few months ago, I decided to try MongoDB Atlas managed services for a client I work for.

I was angry. But I now came to terms with it - that's a bubble. Nothing personal.

My $800 Azure VMs deployment was of a highly available replica set with a delayed replica for a disaster case and automatic backups set up.

I can't ever excuse an $800 VMs cost per month (and let's say +$50 for Azure space for backups) against $75,000 per month for an equivalent service. This is extreme.

I know, usually talking to sales will give me a 50% discount. But that's still $37,500 a month. What the hell.

Additional data following HN comment: I used the "serverless" product, not the "dedicated" product. The "dedicated" require much more management than the "serverless" one, but it is cheaper. So the pricing gap is much lower for the "dedicated" product. FYI.

There will be a reckoning

My theory is that while money was cheap and the markets were hot, startups with millions in funding and no pressure to show profitability just didn't care.

That led to cloud providers across the board becoming ever more greedy. Now that the money and market pressures are changing, so does the business approach.

Cloud services are useful and valuable. The question is - to what degree and at what price point?

Some businesses will go down a tier in the "cloudiness" chart. From a managed DB to self-hosted in VMs. From managed queues to self-hosted Kafka or RabbitMQ.

Other businesses will re-negotiate terms and lower the price points to be justified again.

We may see the common gross-margins figure of ~70% for cloud providers, which is absolutely insane, go down to more sustainable levels. Or - we may see revenue growth slow down significantly in an attempt to keep the margins high. Whatever rock each cloud provider executives will choose to play.