Most of the time, putting a price on a product discourages use of the product.

That’s one reason governments tax things like gasoline, cigarettes, and alcohol, because making it cost money discourages consumption of the thing.

The same principle is true in software, where if you want explosive use of your product, what you ought to do is make it free.

Since we recently brought back the free plan, we thought we’d share, well, the bigger business plan behind building a free plan for computing.

It’s not something we’ve been able to read about much on the internet, so we thought you’d find it interesting.

Let’s get into it.

The conventional high-growth startup calculus is that if you can earn an exponential free user growth curve, you should be able to scaffold viable unit economics underneath it before you run out of money.

But what if you’re selling something with fixed costs, like compute?

These days, some really well-funded companies are taking the margin-negative computing thing to its logical extreme. The “main event” right now is a handful of gigafunded startups racing to secure high-scarcity commodity computing equipment.

That may work for training and inference workload capacity which is currently in short supply but our market (general computing infrastructure) is a bit different.

General computing is a low-scarcity commodity market, which is to say a highly available and undifferentiated market. It’s dominated by the biggest companies in the history of the world — Amazon, Microsoft, Google, etc — which are capable of pushing $100K free credit deals to users.

So on one side you have the “new clouds” which are some of the fastest growing companies in the history of the world because they are pushing margin-negative compute, and on the other side you have the “hyperscale clouds” which are some of the biggest companies in the world which are front-loading enormous free credit deals to lock-in new customers.

We’re challenging both.

To make this business work, you have two options: you can wrap one of the big clouds, or you can build your own data centers.

There was a time we were spending $16 for every $1 of topline revenue. Today, without raising prices whatsoever, we’re running aggressively gross margin positive and that number is climbing rapidly.

While at first we chose to build on top of GCP, we soon learned that you can’t build a cloud on another cloud unless you want to charge enormous premiums on commodity compute to make your business make sense.

In addition to good margins, we also needed beat big cloud pricing because businesses don’t like to use tools that mark commodity prices up 30%, let alone 2-5x.

Apart from that, building on another cloud sticks you with risky upstream dependency chains and limits the low-level control you have when it comes to networking, storage, and more.

We had to exit the cloud to build the cloud we ourselves wanted, which is a cloud that is cheap like the hyperscalers, and easy to use like a PaaS.

What should you do if your competitors are offering negative margin (and thus heavily subsidized) on one side and six figures in free credits on the other?

Our answer to this conundrum is to build perfect incentive alignment with our customers.

If we did margin-negative computing, we’d eventually have to raise prices, which would screw our customers.

If we did huge free credit deals, we would lock our customers into our ecosystem, which would screw our customers.

Instead, we think that building a durable business, with predictable margins, where we help customers to the lowest bill possible for compute through a “pay only for what you use” pricing model is the way to go.

Typically the domain of serverless compute providers, we offer long-running servers by the minute, that scale up and down with your workload.

This is possible because 1) We’ve written our own orchestrator, and 2) We have pricing power because we’ve racked and stacked our own hardware.

(If you’d like to read more about the Railway orchestrator, check out Pierre’s blogpost So You Think You Can Scale?)

In addition to squeezing our competitors, pricing power lets us do lots of cool things, in addition to offering a free plan. It also lets us offer revenue sharing to partners who build on Railway as well as $1M for open source template creators.

The premise here is the same as before, we’re investing money to encourage growth. We’re convinced that templates can reach $1B GMV over time and we want the community to reap most of those rewards.

(If you’d like to read more about Railway’s template marketplace, check out Sarah’s post Deploy Together, Earn Together.)

So that’s our philosophy, to align on incentives with our customers by pricing to encourage growth. All while making sure we build a sustainable business to serve you long term.

If you’re interested in building the most developer friendly cloud with us, check out our hiring page.