Published on: May 9, 2026
A Founder’s Confession on the Cost of “I’ll Turn It Off Later”
Most startup mistakes happen quietly. A wrong hire, a missed deadline, a feature that nobody wanted. The damage compounds over weeks and quarters, and by the time you notice, the lesson is already absorbed.
But some mistakes arrive all at once. In an email. With a number that doesn’t feel real.
For Tushar, the founder of Explaino — a Vellore Institute of Technology graduate building one of India’s fastest-growing AI video platforms — that moment came on an ordinary Tuesday morning, when he opened his AWS billing dashboard and saw a figure he had to read three times to believe.
₹50,00,000.
Fifty lakh rupees. Burned through cloud servers he had forgotten existed.
This is the story of how it happened, what he did about it, and the lesson every founder building on the cloud needs to internalize before their own Tuesday morning arrives.
The Setup
In the early days of Explaino, the team was building fast. AI voiceover models, video processing pipelines, transcription engines, smart-zoom detection — every feature required experimentation, and experimentation required GPUs.
GPUs are not cheap.
A single high-end GPU instance on AWS in the Mumbai region — the kind needed to train and test AI video models — can cost ₹250 to ₹400 per hour. Run one for a day, and you’ve spent ₹6,000 to ₹9,600. Run it for a month, and you’re at ₹1.8 lakh to ₹2.9 lakh. And that’s just one instance.
To move quickly, the Explaino team spun up multiple GPU instances across multiple AWS accounts. One for production. One for staging. One for experimentation. One for a specific feature test that a developer was working on solo.
Each account had its own purpose. Each was supposed to be temporary.
This is where the story turns.
The Mistake
The experimentation account — the one used for testing new AI models — was set up with auto-scaling enabled. The logic was simple: if a model needed more compute, AWS would automatically spin up more instances to handle the load. Convenient. Standard practice.
What wasn’t set up: budget alerts. Cost caps. A scheduled shutdown for non-business hours. A reminder to review the account weekly.
The developer who created the account moved on to other work. The instances kept running.
Auto-scaling, doing exactly what it was designed to do, kept spinning up new GPU instances as background processes consumed compute. Some of those processes had bugs that caused them to loop. Others were left running because nobody remembered to terminate them.
Days turned into weeks. Weeks turned into months.
The account, sitting in a corner of AWS that nobody was actively monitoring, was quietly burning ₹40,000 to ₹60,000 a day.
The Discovery
The realization came not from a budget alert (there wasn’t one) or a finance review (the team was small enough that cloud bills were aggregated and reviewed monthly). It came from the bank.
A standard payment to AWS had been auto-debited. The amount was significantly higher than the previous month. The bank sent a notification. Tushar opened it expecting a typo.
It wasn’t a typo.
He logged into AWS Cost Explorer and saw the breakdown. One account. One forgotten experimentation environment. Multiple GPU instances. Months of unmonitored usage.
The total damage, when traced back across the period the account had been live: ₹50 lakh.
For a bootstrapped startup, this was not a number that could be absorbed. It represented runway. Salaries. Customer acquisition. Product development. Roughly six months of operational budget — gone, not to a bad bet on a feature that didn’t work, but to a forgotten checkbox.
The Aftermath
The first reaction was panic. The second was action.
Within 24 hours, the team had:
- Terminated every running instance across every AWS account
- Audited every active service across every cloud provider they used
- Contacted AWS support to negotiate the bill (cloud providers sometimes offer partial credits for accidental over-usage by genuine customers — AWS provided a partial credit, though it didn’t come close to covering the full loss)
- Enabled budget alerts on every account at multiple thresholds
- Set up a weekly cloud cost review as a non-negotiable team ritual
- Implemented automated shutdown policies for all non-production resources outside business hours
The financial hit was real. But the operational change that came out of it — the discipline around infrastructure cost — became one of the most valuable assets the company built that year.
Today, Explaino runs leaner cloud infrastructure than most companies its size, with cost-per-video metrics that competitors would envy. Every instance is tagged. Every account has alerts. Every dollar of compute is accounted for.
That discipline didn’t come from a best-practices blog post. It came from a ₹50 lakh tuition fee.
Why This Happens to More Founders Than You’d Think
Tushar’s story is unusually transparent. Most founders who lose money this way never talk about it publicly. But cloud cost overruns are one of the most common and least-discussed startup disasters in India and globally.
The reasons are predictable:
- Engineers optimize for speed, not cost. Spinning up a powerful instance to “just test something” is faster than provisioning a smaller one and figuring out if it’ll handle the load.
- Cloud pricing is opaque. Mumbai region pricing, plus 18% GST, plus data transfer fees, plus storage, plus the currency conversion if billing is in USD — most founders can’t reverse-engineer their bill without effort.
- Auto-scaling is dangerous without guardrails. It does what you ask. If you don’t ask it to stop, it won’t.
- Budget alerts aren’t on by default. AWS, GCP, and Azure all require you to set these up manually. Most founders skip the step and forget.
- Small teams aren’t reviewing finance weekly. The bills aggregate. The shock comes once a month, sometimes once a quarter.
The mistake isn’t a sign of incompetence. It’s a sign of how easy modern cloud infrastructure makes it to spend money without realizing.
The Lesson
When asked what he’d tell other founders, Tushar’s answer is short.
“Set the alerts. Set the alerts on the day you create the account. Not later. Not when you have time. The day you create it. And review your cloud bills weekly, not monthly. By the time it shows up on a monthly review, you’ve already lost more than you should have.”
He adds one more thing.
“Talk about your mistakes. The reason this happened to me is because nobody in the founder community was talking openly about cloud costs. We were all pretending we had it figured out. We didn’t. If one founder reads this story and sets up budget alerts tonight, the ₹50 lakh wasn’t completely wasted.”
How Explaino Came Out of It
The irony of the story is not lost on the team.
Explaino, a company built to help other teams ship faster and cheaper through automation, learned its biggest lesson by failing to automate the one thing that mattered most: its own cost discipline.
Today, the platform itself reflects that lesson. Every feature Explaino ships includes built-in efficiency — AI processing happens in the most cost-optimal regions, models are quantized to reduce GPU load, and customers get usage analytics that help them understand their video infrastructure costs in detail.
The ₹50 lakh disaster, in the end, made the product better. It forced the team to build with constraints. To respect compute. To value efficiency not as an optimization but as a core value.
Closing Thoughts
The story of Explaino’s ₹50 lakh loss is, on the surface, a cautionary tale about cloud bills. But the deeper takeaway is something every founder building anything ambitious will recognize.
Speed feels free. Until it isn’t.
The features you forget. The accounts you leave open. The processes you never document. They all compound silently in the background, until one Tuesday morning you open an inbox and find out exactly what you owe.
For Tushar and the Explaino team, the lesson cost ₹50 lakh.
For every founder reading this, the lesson can cost ₹0 — if you set the alerts tonight.
