Most new products launch fastest on Vercel for the app and Supabase for the database, for about US$45 a month once both are on paid plans. AWS earns its place when a product has heavy, steady workloads, strict compliance requirements or engineers to run it. The deciding questions are who will look after the system and how your bill grows as customers arrive.
Vercel and Supabase do different jobs
Every product needs somewhere to run its app and somewhere to keep its data. Vercel runs the app: it builds your code every time it changes and serves it from a global content delivery network, with a Sydney region for serverless functions. Supabase keeps the data in a Postgres database and adds user logins, file storage and live updates on top, which makes it a backend as a service. AWS covers both layers with hundreds of separate services, and your team assembles and configures each one.
Most teams pair Vercel with Supabase. The real decision is whether that pairing or building directly on AWS fits your product better, with Firebase and Cloudflare as the main alternatives. Hosting is one layer of your tech stack, and it's the layer that's hardest to change once customers depend on it.
| Provider | What it runs | How it charges | Best for |
|---|---|---|---|
| Vercel | The app and its server code, with first-class support for Next.js | Pro is US$20 a month per developer and includes US$20 of usage | Web apps and sites where releasing changes quickly matters most |
| Supabase | Postgres database, user logins, file storage and live updates | Pro starts at US$25 a month and includes US$10 of database compute | Products built on structured records such as customers, bookings and orders |
| AWS | Every layer, from servers and databases to queues and AI models | Pay per use for each service, with up to US$200 of credit for new accounts | Heavy steady workloads, compliance requirements and teams with engineers |
| Firebase | Google's document database, logins, hosting and functions | Free Spark plan, then pay per read, write and gigabyte on Blaze | Mobile apps that need offline sync and live updates |
| Cloudflare | Code on Cloudflare's network, plus R2 file storage and the D1 database | Workers Paid starts at US$5 a month, and data transfer is free | High-traffic sites, APIs and media where bandwidth would dominate the bill |
| Railway and Render | Platform as a service hosting for servers that run all day, background workers and databases | Railway bills per second of CPU and memory; Render instances start at US$7 a month | Long-running jobs, queues and apps that keep a connection open |
What it costs at each stage
The Vercel and Supabase pairing has the most predictable bill for a small team, because both charge a flat plan fee and include a usage allowance. Here's how it typically grows.
| Stage | Vercel and Supabase each month | What drives the bill |
|---|---|---|
| Prototype | US$0 on free plans | Vercel's Hobby plan is for personal, non-commercial use, and free Supabase projects pause after a week without activity, so both move to paid plans at launch. Our guide to no-code vs custom code covers when a prototype is ready to become a product |
| Launch, one developer | About US$45 | Vercel Pro (US$20) and Supabase Pro (US$25), with the smallest database covered by Supabase's compute credit |
| Growing, three developers | About US$90 | Three Vercel seats (US$60), Supabase Pro (US$25) and a step up in database size (about US$5 after the credit) |
| Scaling | Usage-based | Function CPU time on Vercel, plus database size, compute and data transfer beyond Supabase Pro's 100,000 monthly active users and 250 GB of egress |
Vercel's Pro plan now bundles website traffic into a flat-rate CDN, so a busy month of visitors moves the bill far less than it used to. Function CPU time is the line worth watching, at about US$0.13 per active CPU hour beyond the included credit. On Supabase, each extra project adds its own compute cost, which matters if your multi-tenancy design gives each client a separate database.
When AWS is the better choice
AWS pays off once a product runs heavy compute around the clock, such as video processing, large data pipelines or AI workloads. At that volume, paying per unit directly to AWS beats the margin a platform adds on top. It's also the usual pick when a government or enterprise customer requires specific certifications, or when a business already has AWS credits and engineers who know it well. Comparing build vs buy at this point helps decide which workloads are worth owning outright.
The trade is ownership. Someone has to set up networking, permissions, backups, monitoring and deployments, and keep them current. Teams that move to AWS successfully usually move one workload at a time, such as a background job or a data pipeline, and keep the app on the platform that already works. Our platform migration checklist covers how to make that switch without downtime.
Where Firebase and Cloudflare fit
Firebase is strongest on mobile apps, where its offline sync and live updates save a lot of build time. Its free quotas are generous, with 50,000 document reads and 20,000 writes a day and 50,000 monthly active users for logins. Firestore stores data as documents, which suits chat and activity feeds well and makes reporting across customers and orders harder. Moving off Firestore later means reshaping the data, whereas a Supabase database is standard Postgres that moves to any Postgres host.
Cloudflare charges nothing for data transfer, which changes the maths for media-heavy products and busy public APIs. Workers Paid starts at US$5 a month with 10 million requests included, and R2 stores files with free egress. We put Cloudflare in front of media-heavy commerce builds for exactly this reason.
Keeping your exit open
Every provider makes some features easy to adopt and hard to leave, which is how vendor lock-in builds up. The way to stay flexible is to keep your business rules in your own code and your data in a format other hosts can read.
- Postgres is the most portable choice of database. Supabase is open source and supports self-hosting, and its data exports to any Postgres provider.
- Next.js runs on other hosts, though Vercel features such as image optimisation and incremental page caching need replacements elsewhere.
- Firestore and DynamoDB use their own data models, so leaving either one is a rebuild of the data layer.
- An API-first design lets you swap the front end or the backend separately as the product grows.
Our founder's notes on platform lock-in risk go deeper on when owning more of the stack is worth it, and a fractional CTO can review those choices before the build starts.
Keeping Australian data in Australia
Vercel, Supabase, AWS and Google Cloud (which runs Firebase) all offer Sydney regions, and AWS and Google Cloud add Melbourne, which covers most data residency requirements. Choose the region when you create the project, because moving a Supabase database to another region later means migrating it to a new project.
Under the Privacy Act, a business that sends personal information overseas generally remains accountable for how it's handled there. Health, NDIS and government clients often write Australian hosting into their contracts, so settle the region before the first customer record goes in.
How we choose for client builds
Our default for a new product is Next.js on Vercel with Supabase behind it, because a small team can ship to it daily and the monthly bill stays predictable. This site and SportsBlock both run on Vercel, and the internal CRM and support product we run for Marshall Tech clients is built on Supabase. Supabase's growth reflects how common this pattern has become: its CEO told TechCrunch in June 2026 that more than 60% of new databases on the platform are launched by AI tools.
We move individual workloads to Cloudflare or AWS when the numbers or a customer's requirements call for it, and we price that decision before anything is built. A custom platform MVP starts from $15,000 AUD + GST and ships in four to eight weeks, with the stack and running costs agreed up front. For a business weighing up a product or an existing build, a fractional CTO makes these calls alongside your team, and our pricing page lists every starting price.
If the bill on an existing product has crept up, a tech stack audit usually finds the services you can consolidate or downsize, and our tech stack optimisation work carries those changes through. When you're ready to talk through your own product, get in touch.