The cheapest way to host a SaaS app or video generation app.Real numbers, not estimates.
If you have ever searched for the cheapest cloud to host your side project or your video generation app, you have probably found blog posts full of vague percentages and no actual dollar amounts. This post shows the real monthly bill for two common app shapes: a SaaS with 5,000 users and a video generation app running on GPU compute.
Published April 26, 2026
SaaS side project: what you actually pay each month
A typical SaaS side project needs four things: a server to run the app, a database, authentication, and storage. Most developers price just the hosting and forget the rest. The real bill includes all four.
Here is the monthly cost for a Next.js SaaS with a Postgres database, authentication, and 5,000 monthly active users on 10 GB of storage. Numbers are from the Varity cost calculator, based on published pricing as of April 2026.
| Service | AWS | Vercel | Varity |
|---|---|---|---|
| Hosting | $250/mo | $100/mo | $25/mo |
| Database | $750/mo | $400/mo | $100/mo |
| Authentication | $250/mo | $150/mo | $0 (included) |
| Total | $1,250/mo | $650/mo | $125/mo |
At 5,000 users, Varity costs $125 per month. AWS costs $1,250. Vercel costs $650. The gap is not mainly from compute pricing: it is from authentication. AWS charges per user per month for Cognito. Varity includes authentication at no extra cost.
As your app grows to 100,000 monthly active users (50 GB storage), the gap widens further:
| Service | AWS | Vercel | Varity |
|---|---|---|---|
| Hosting | $5,001/mo | $2,002/mo | $501/mo |
| Database | $15,000/mo | $8,000/mo | $2,000/mo |
| Authentication | $5,000/mo | $3,000/mo | $0 (included) |
| Total | $25,001/mo | $13,002/mo | $2,501/mo |
Cost estimates based on published pricing as of April 2026. Actual costs vary by usage, region, and plan selection.
Video generation app: the two-part cost you need to know
A video generation app has two separate cost drivers that most hosting calculators do not show together: the web layer and the GPU compute layer.
Web layer (API, queue, database, auth)
This is the part that handles user requests, stores job state, and returns results. It is a standard web app. Using the same cost model above at 1,000-5,000 users, you are looking at roughly $30-125 per month on Varity versus $200-650 on Vercel or $300-1,250 on AWS.
GPU compute (inference)
This is the part that actually generates the video. On AWS, a GPU instance (G4dn.xlarge, one T4) runs about $0.53 per hour and you pay for it whether it is generating video or idle. For 1,000 videos per month at 30 seconds of compute each, you need about 8 GPU hours, which is $4-5 on AWS in raw GPU cost. But since AWS instances run continuously, a team doing burst workloads often pays for 24 hours even when they use 8.
Varity bills GPU compute per second of active use, not per hour reserved. For the same 1,000 videos at 30 seconds each, you pay for those 8 GPU hours and nothing more. For bursty video generation workloads, that is typically 60-80% cheaper than running a reserved AWS GPU instance.
The total monthly cost for a video generation app with 1,000 jobs per month on Varity is roughly $35-50 all in. The same stack on AWS is typically $150-300 per month, depending on how efficiently you size and scale the GPU instance.
Why Varity costs 60-80% less than AWS
The savings come from three things working together:
- +Authentication is included. AWS Cognito and Vercel's auth integrations charge per monthly active user. At 5,000 users, that is $250-300 per month just for login. Varity includes authentication in the platform fee with no per-user charge.
- +Usage-based pricing, not reserved capacity. AWS and Vercel both have pricing models that reward you for committing to capacity in advance. Most side projects and early-stage apps do not have predictable traffic, so you over-provision. Varity charges by actual usage with no minimum commitment.
- +One bill, not four. When you use AWS, you pay separate invoices for EC2 (compute), RDS (database), Cognito (auth), and S3 (storage). Each has its own pricing curve and minimum. Varity bundles these into a single usage-based bill, which removes the additive cost of minimum charges on services you barely use.
How to deploy on Varity
Two commands. Your app is live in under 60 seconds.
In your project directory:
$ npx varity@latest init $ varity deploy Detecting framework: Next.js Configuring database...done Configuring auth...done Deploying... Live at https://varity.app/your-app/
Already on Vercel? The migration path is one command and takes about 10 minutes:
$ npx varity@latest migrate --from vercel $ varity deploy
If you use an AI IDE like Claude Code or Cursor, you can install the Varity MCP server and deploy without leaving your editor. Type "deploy this with Varity" and it runs both commands for you.
Which platform should you use?
Here is the honest summary:
- +Use Varity if you want the cheapest full-stack option and you do not want to configure anything.
- +Use Vercel if your app is mostly static or server-rendered and you do not need a database or auth included.
- +Use AWS if you have a dedicated infrastructure team and need fine-grained control over every service.
- +Use Varity for GPU compute if your video or AI app runs burst workloads rather than continuous inference.
Frequently asked questions
What is the cheapest place to host a SaaS side project?
Varity is the cheapest full-stack option for SaaS side projects. At 5,000 monthly active users with a database and auth, Varity costs around $125 per month compared to $650 on Vercel and $1,250 on AWS. Varity includes authentication and database at no extra charge, which is where most of the savings come from.
What is the cheapest way to host a video generation app?
A video generation app needs two things: a web API layer (for handling requests, queueing jobs, and serving results) and GPU compute (for running inference). Varity covers both in a single deployment. The web layer costs 60-80% less than AWS, and GPU compute is billed by the second rather than by the hour, so you only pay when a video is actually being generated.
Is Varity cheaper than Vercel for a full-stack app?
Yes. For a full-stack app with a database and auth, Varity is consistently cheaper than Vercel. The main reason: Varity includes database and authentication in the platform cost. On Vercel you pay for those separately, and the combined bill adds up quickly once you have real users.
Does Varity support GPU compute for video and AI apps?
Yes. Varity supports GPU compute for inference workloads including video generation, image generation, and LLM serving. GPU is billed per second of active compute, not per hour reserved. For workloads that generate video on demand (rather than continuously), this per-second billing cuts costs significantly compared to always-on GPU instances on AWS.
How do I deploy a video generation app on Varity?
Run 'npx varity@latest init' in your project directory and then 'varity deploy'. Varity detects that your app requires GPU compute and configures it automatically. You do not need to select an instance type, configure networking, or set up a separate GPU service.
See your app's cost on Varity
Deploy for free during beta. No credit card required until your app has real traffic.