Sustainable SaaS: cut your cloud cost and carbon

Your cloud bill went up 30% last year and nobody in the company can really explain why. Your customer count did not grow 30%. The team did not double. But the bill went up.

Meanwhile, the UN published a report saying the world will blow past the 1.5 degree warming limit. It sounds like one thing has nothing to do with the other. It does. Both bills come from the same place: energy being spent with nobody watching. And that is exactly where the idea of a sustainable SaaS stops being a nice speech and becomes one less line in your fixed costs.

I am not going to ask you to save the planet. I am going to show you that the part of your system that pollutes the most is, almost always, the part you pay for and never use.

What a climate report has to do with your bill

Cloud is a very good marketing word. It makes your data sound like it floats in the air.

It does not float. It lives in warehouses full of machines running 24 hours a day, with industrial air conditioning working nonstop so the heat does not melt everything. The International Energy Agency estimated that data centers consumed around 415 TWh in 2024, close to 1.5% of all the electricity on the planet. The projection is that this number almost doubles by 2030, pushed mostly by AI.

The cloud is not a cloud. It is someone else's computer, plugged into a wall.

Once you get that, one thing becomes obvious: every dollar on your cloud bill matches real energy burned. If the bill has fat, the fat is physical too. You are paying to heat up a warehouse in Virginia.

And here is the good part for anyone running a business: you do not have to choose between saving money and reducing impact. In this specific case, it is the same button.

Your system burns energy even when nobody is using it

This is the point almost every business owner finds out too late.

Software does not work like the lights in your house. It is not "use it, pay for it". A large share of a SaaS infrastructure sits there running, waiting for something to happen. And waiting costs money.

Cases I have found at clients, all real, all boring:

• A staging environment created for a test in 2023, never turned off. It cost R$ 900 a month to serve exactly zero people.
• An automated report running every hour, querying the entire database. The recipient had left the company eight months earlier. The email landed in a dead address.
• Daily backups of a 40 GB database, kept forever, with no expiration rule. Three years of copies nobody would ever restore.
• An integration trying to sync with a system that had been shut down. It failed, retried, failed, retried. A hundred and forty thousand requests a month just to throw an error.

None of these is a bug. The system was working perfectly. It was doing brilliantly something that did not need to be done at all.

It is the air conditioning running with the window open, digital edition. And the window has been open for two years.

"But my company is small, this changes nothing"

Fair. If your bill is R$ 2,000 a month, cutting 30% gives you back R$ 600. It will not change your quarter.

But the waste argument is not only about the amount. It is about what the waste tells you.

A system with cron jobs running for nobody, a broken integration stuck in a loop and three forgotten environments is a system nobody is watching. And a system nobody watches is where the next serious problem lives: the bill that triples on its own during a peak month, the database that fills up, the integration that stayed broken for six weeks without anyone noticing, the customer data still sitting there when it should not be.

Finding waste is a side effect of having visibility. Visibility is what you actually want.

And there is a part already knocking on the door. Large companies have started asking suppliers for emissions data. If you sell software to a manufacturer or a multinational, the odds of an ESG questionnaire showing up in your sales process over the next few years are high. Walking into that conversation with a number in hand is very different from walking in with "I think we do not spend much".

How to look at this without turning it into a six month project

You do not need a consultancy, a new dashboard or a committee. You can do it in one afternoon. This is my order:

1. Open your detailed cloud bill. AWS, Google Cloud, Azure, Vercel, they all have one. Sort by cost, highest to lowest. Look at the top five items.

2. Ask about each one: who uses this and when? If nobody can answer in two minutes, you found a candidate. Mark it, do not delete it yet.

3. List everything that runs on its own. Scheduled jobs, automated reports, syncs, recurring emails. For each one: who reads the result? If the name that comes up belongs to someone who left, shut it down.

4. Check the error logs for the last seven days. A repeated error in high volume is almost always a process trying to talk to something that no longer exists. Every attempt is a request you pay for.

5. Set a lifespan for cold data. Backups, logs, temporary files, old attachments. Almost every provider has automatic rules for expiration or for moving data to cheap storage. Setting it up takes ten minutes and the effect is permanent.

6. Turn it off before you delete it. Shut the suspicious resource down and wait two weeks. If nobody complains, delete it. If someone complains, you found out who was using it. You win either way.

This routine usually cuts 20% to 40% off the bill in systems that have never been reviewed. The first time is always the most generous, because there are years of accumulated stuff in there.

Where AI changes the math

If you are putting AI into your product, pay attention to this part.

A model call is expensive and heavy, in money and in energy. Plenty of people ship an AI feature and never look at it again. So it sends the entire document when a snippet would do. It runs the big model where the small one would solve it. It reprocesses the same question fifty times a day without saving the answer.

Caching solves most of that. If ten customers ask the same question, answers two through ten can come from storage. It is instant, it is free and it burns nothing.

Here is my strong opinion: most of the AI features I see in products today could spend a fifth of what they spend and deliver exactly the same result to the user. The difference is not technical, it is attention. Nobody went back to look after it worked.

The right way to think about this

Do not treat it as a sustainability campaign. Campaigns end.

Treat it as maintenance. Once a quarter, someone opens the bill, checks what runs on its own, checks the errors and shuts down what is no longer useful. One hour of work, four times a year.

The UN report says we will pass 1.5 degrees and that the way back depends on cutting waste everywhere, not just at the smokestacks. Your share is small. But it is the only one entirely under your control, and it gives you money back while you do it.

If you do not know where to start, or you suspect your operation has hidden fat and nobody has time to hunt it down, this kind of cleanup is what I do. I look at the infrastructure, show you what is burning for nothing and leave the shutdown configured so it does not come back. See how I work.

Your cloud bill went up 30% last year and nobody can explain why.

Your customer count did not grow 30%. The team did not double. But the bill went up.

I have found a staging environment forgotten since 2023 costing R$ 900 a month to serve exactly zero people. An automated report running every hour into the inbox of someone who left the company. An integration failing 140 thousand times a month against a system that was shut down.

None of that is a bug. The system was doing brilliantly something that did not need to be done at all.

And since the cloud is just someone else's computer plugged into a wall, that fat is not only financial. It is real energy burned.

One afternoon of review usually cuts 20% to 40% off the bill the first time around.

If you suspect there is hidden fat in there and nobody has time to hunt it down, reach out.

#SaaS #CloudCost #FinOps #Sustainability #Technology