I opened up the Perks page after graduating from On Deck and was flabbergasted to see that we had gotten FREE money. Well, maybe not free money but basically free. It was $10K worth of AWS credits to build our startup. We're now 6 months in and honestly, redeeming them was a mistake.
AWS perks page on On DeckThe infrastructure for engyne.ai is not that complicated: there's frontend & backend hosting with Vercel (Next.js), a DB and an S3 for images.
Earlier in the year, we were paying $20/month total - Vercel, S3 free tier and $20/month Digital Ocean Postgres.
We honestly just should have stayed there but we pigged out.
Where we spent our AWS credits
On Day 1, I provisioned a db.m5.large, 20GiB, multi-AZ, all the bells & whistles Postgres instance and migrated all of our data to it. I even got a second instance for staging (cuz why not right?). The cost for all of this is ~200 USD/month. Here's the kicker: we only use 2% of this storage currently.
AWS Cloudwatch dashboard showing 18GiB free storageThat's why using the credits was a mistake. We have 10 customers. The infrastructure usage is barely anything. Our bottleneck is not on the infrastructure level but on the customer acquisition - How do we get the next 100 customers. This is the classic technical founder trap.
As technical founders, we think that launching a startup means having a robust scaling policy in place on Day 1. That's just not true. You'll go through a hundred different iterations of your idea, design, sales offer, landing page copy before you get to any sort of Product Market fit where people are using the platform so much that it hits any sort of limits.
Don't think about scaling your infrastructure before you have a business that is scaling.
Since the credits expire after 12 months, we have a massive $200USD/month bill waiting for us if we don't off-board soon.
We are migrating off of AWS soon
Digital Ocean's $15/month Postgres plan is more than enough storage for us at the moment. I would use pg_dump to export all the data from the AWS Postgres to the Digital Ocean one. They also now have the pg_vector extension support so we can save our vector embeddings. Hopefully, this goes smoothly.
We should have waited at least a year before redeeming the credits to really take advantage of the amazing services that AWS provides. Since the credits expire after a year, we will end up with about 80% of the credits unused.
If I was to start again, I would continue using Digital Ocean and Vercel until regular customer usage hits ~80% of the capacity. Only then I would redeem the credits and move the infrastructure over on a weekend. This way you know you are in a good position to actually utilize the free gift and that you have enough revenue to justify continuing to stay on even after the credits expire.