Invoices in the thousands of dollars for minimal mistakes, unexpected attacks, misunderstandings of fees, and limits that don’t work: this is the dark side of serverless, compiled on a chilling website.
With a simple yet straightforward design, the website ServerlessHorrors is gaining notoriety in the tech community by becoming a sort of “haunted house” of the cloud world. Its proposal is clear: to gather real stories from developers and teams who have suffered exorbitant bills, billing errors, misunderstood configurations, and abuses of the serverless model from providers like Vercel, Google Cloud, AWS, Cloudflare, Netlify, or PostHog.
Created by András, a developer behind the project Coolify, an open-source, self-hostable alternative to platforms like Heroku, Vercel, or Netlify, this website serves as both a warning and a channel for catharsis for those who have had to learn the hard way.
From $50 a month… to $70,000 in one go
One of the most extreme cases featured on the site is that of a Google Cloud user who, after paying only $50 a month, woke up one day to a bill of $70,000. Another story recounts how a simple query in BigQuery, about a public dataset, generated a bill of $22,639.69, with the user not expecting such consumption.
The common denominator: a lack of real control over spending and the collateral effects of a serverless infrastructure that promises simplicity but can become unpredictable and expensive if its operation is not understood.
When the “free tier” isn’t so free
Another testimonial illustrates how a free account on AWS led to a bill of over $100 due to a series of misunderstood configurations. “Why is $103 a horror story? Because it was a free plan,” the author quips. There are even documented cases where publishing a simple sitemap.txt led to the use of hundreds of GB per hour, resulting in unexpected charges.
There are also examples where DDoS attacks caused abnormal traffic that was billed without mercy, as in the cases of Netlify or Cloudflare, where platforms demanded payments of up to $120,000 in less than 24 hours or directly deactivated the sites.
A reminder of the risks of poorly managed serverless
The serverless philosophy has gained popularity in recent years for its promise of automatic scalability, simplicity in management, and pay-for-use. However, ServerlessHorrors shows the dark side of this model when effective limits are not implemented, use is not monitored, or when providers’ billing policies are opaque.
Among the recurring lessons from the published stories are:
- The risk of traffic attacks on pay-per-use services (DDoS, scraping, form spam…).
- The absence of real-time alerts for abnormal usage.
- The failure of configurable spending limits, as seen in Vercel, where even with a cap of $120, some users received bills over $700.
- The confusion between storage, transfer, and computational usage pricing, especially on platforms like Firebase or BigQuery.
- The impact of malfunctioning AI, such as the recent case where a suggested change by Devin (Cognition Labs’ AI) caused an error that cost $1,273.69.
A call to the community: “Share your story or open a PR”
ServerlessHorrors is not just a website for reading. It is also participatory. András invites other developers to submit their stories or open a pull request on GitHub, turning the platform into a community project that raises an essential debate: the need for transparency, real limits, and greater technical education regarding the use of cloud services.
In times where “deploy in minutes” is promoted, this website reminds us that ease can come at a high cost if the underlying model is not understood. A must-read for system administrators, developers, startups, and infrastructure managers.
