Where to host your African startup: VPS vs serverless vs PaaS

You've built something people want. Now it needs to live somewhere that won't collapse when your first 10,000 users arrive-and won't drain your runway while you're waiting for them. For African founders, the hosting decision is rarely straightforward. Your options span continents, your network is unpredictable, and the wrong choice can mean your app feels sluggish in Lagos but blazes in London.

This article cuts through the noise. We'll compare VPS, serverless, and Platform-as-a-Service (PaaS) on the factors that actually matter to you: cost in naira, latency for your users, operational overhead, and whether you'll wake up at 3 a.m. fighting a database fire. By the end, you'll know which model fits your stage, your team, and your users' reality.

The three hosting models explained

What each one actually does

Let's start with definitions that matter:

VPS (Virtual Private Server): You rent a slice of a physical server. You get root access, you manage everything-the OS, the runtime, the database, security patches, backups. Providers like DigitalOcean, Linode, and AWS EC2 offer this. You pay per month for a fixed amount of compute.

Serverless: You write functions, upload them, and the provider runs them on demand. You don't manage servers, scaling, or infrastructure. You pay per execution-per 100 milliseconds of CPU time, per API call, per database operation. AWS Lambda, Google Cloud Functions, and Vercel are examples.

PaaS (Platform-as-a-Service): A middle ground. You push your code (usually via git), the platform builds and deploys it, handles scaling, and you manage your application layer. Vercel, Heroku, Render, and Railway sit here. You pay per app or per resource consumed, but you're not managing servers or OS-level patches.

Each has a different cost curve, operational burden, and latency profile. None is universally "best."

Cost comparison: what you'll actually pay

Let's be concrete. Here's what a typical small-to-medium startup spends monthly on each model:

Hosting model	Entry price (USD/month)	Typical cost at 10k MAU	Typical cost at 100k MAU	What's included
VPS (DigitalOcean)	$4-6	$12-24	$24-60	Fixed compute, you manage everything
Serverless (AWS Lambda + RDS)	$0-5	$20-80	$100-500+	Pay per invocation; database costs separate
PaaS (Vercel/Railway)	$0-20	$20-100	$100-300	Includes scaling, some free tier

But this table hides the real story. Let's dig in.

VPS costs are predictable but inflexible. A $12/month DigitalOcean droplet gives you 1GB RAM, 1 vCPU, and 25GB SSD. That's enough for a Django app and a small Postgres database if you're lean. But if you're running out of memory, you upgrade to the $24 tier-you're now paying double whether you need it or not. In naira terms (at ~₦1,500/$1), that's ₦18,000 to ₦36,000 monthly. Over a year, that's ₦216,000 to ₦432,000. Most Nigerian startups we've worked with at LaunchPad keep a VPS for side projects precisely because the cost is boring and predictable.

Serverless can surprise you. AWS Lambda's free tier is generous: 1 million requests per month, 400,000 GB-seconds of compute. But the moment you exceed it, costs climb fast. A spike in traffic-say, a viral tweet or a Product Hunt launch-can cost you ₦50,000 in a single day if you're not careful. We've seen founders shut down Lambda functions mid-launch because the CloudWatch bill was spiraling. However, if your traffic is genuinely bursty (seasonal, event-driven), serverless can be cheaper than paying for idle VPS capacity.

PaaS sits in the middle. Vercel's free tier runs one project with generous limits. Their Pro plan is $20/month. Railway starts at $5/month and scales with usage. The advantage: you're paying for what you use, but there are guardrails. Railway won't bill you ₦500,000 because of a runaway loop-there are spending caps.

For a Nigerian startup bootstrapping, the cost hierarchy is usually: VPS (cheapest at small scale) < PaaS (predictable, fewer surprises) < Serverless (unpredictable, but can be cheapest at very high or very low volume).

Latency and performance for African users

This is where hosting choice becomes non-negotiable. If your app takes 5 seconds to load in Kano, users won't wait.

VPS latency depends entirely on where the server lives. If you host on DigitalOcean's Frankfurt data centre, a request from Lagos travels ~9,000 km and takes ~150-200 ms one-way. Add your app's processing time, database queries, and you're at 500-1000 ms total. Noticeable, but acceptable. If you use DigitalOcean's New York region, you're looking at 250+ ms one-way-now your app feels sluggish.

The solution: host in Africa. DigitalOcean doesn't have a data centre in Nigeria yet. But AWS has regions in Cape Town (Africa South) and Lagos (Africa Central, launched 2024). Hosting on Africa Central cuts latency to ~20-30 ms for Lagos users. In our experience at LaunchPad, founders who switched from US-based VPS to AWS Africa Central saw immediate user complaints disappear.

Cost trade-off: AWS Africa regions are pricier. A t3.micro instance in Africa Central costs roughly 30% more than in us-east-1. But if 80% of your users are in Nigeria, the latency win is worth it.

Serverless scales latency differently. AWS Lambda functions deployed in the Africa Central region have the same latency advantage. But cold starts are real. The first invocation of a Lambda function can take 1-3 seconds (for Node.js; Python is slower). Subsequent invocations are fast. For APIs that are called frequently, this is fine. For sporadic batch jobs, cold starts don't matter. For a user-facing endpoint that's idle for 15 minutes, then gets called? The user waits.

Vercel, Netlify, and other edge-first platforms use CDNs. Your static assets and API responses are cached at edge locations close to users. Vercel has edge locations in Africa, but your origin server might still be in the US. This is fine for static sites and JAMstack apps but adds latency for dynamic, personalized content.

PaaS providers like Railway and Render let you choose your region. Railway has a Singapore region; Render has Oregon. Not ideal for African latency. However, both support custom domains and can integrate with CDNs like Cloudflare, which has presence in Nigeria. The latency is better than VPS in distant regions but worse than a native African VPS or serverless function.

For low-bandwidth scenarios-common across much of Africa-see our guide on designing for low-bandwidth. Hosting choice alone won't save you if your app is bloated.

Operational burden: who wakes up at 3 a.m.

This is the hidden cost most founders underestimate.

VPS requires you to be a sysadmin. You manage OS updates, security patches, database backups, monitoring, and scaling. A security vulnerability in your OS? You patch it. Your disk fills up? You debug it. Your database gets slow? You optimize queries or add indexes. For a solo founder or a team without a DevOps engineer, this is a drag. In our experience, founders spend 5-10 hours per month on VPS maintenance-time they could spend building.

However, VPS teaches you systems thinking. You understand what's happening under the hood. This is valuable.

Serverless abstracts away infrastructure but adds complexity elsewhere. You don't manage servers, but you manage function concurrency, timeouts, cold starts, and vendor lock-in. AWS Lambda has a 15-minute timeout; if your task takes longer, it fails. You need to architect around this. You also need to understand cost drivers: every database query, every API call, every MB of data transferred costs money. A N+1 query bug becomes a ₦10,000 bill. The operational burden shifts from "keep the server alive" to "keep costs predictable."

PaaS is the least operationally demanding. You push code, the platform deploys it, and scaling happens automatically. You still monitor logs and errors, but the platform handles OS updates, database failover, and scaling decisions. For a founder who wants to focus on product, PaaS is a win. The trade-off: you're locked into the platform's ecosystem. Want to run a custom binary? Not easily. Want to use a specific database version? Limited options.

Here's a decision matrix:

• Choose VPS if: You have a DevOps engineer, you need full control, or you're running something the other platforms don't support well (e.g., a Rust service, a custom daemon).
• Choose Serverless if: Your traffic is bursty, you're cost-optimizing, and you're comfortable with vendor lock-in and cold-start latency.
• Choose PaaS if: You want to ship fast, you're not optimizing for extreme scale, and you prefer predictable costs over maximum control.

Tech stack considerations for your choice

Your framework and database choices constrain your hosting options. Consider the alignment we outline in tech stack choices for Nigerian startups in 2026.

Node.js, Python, and Go work on all three models. No constraints.

Stateful applications (WebSockets, long-lived connections) are awkward on serverless. Lambda functions are designed for request-response. If you're building a real-time chat app, a VPS or a PaaS with persistent connections is better. Vercel's serverless functions support WebSockets, but it's not the natural fit.

Databases are a major constraint. On a VPS, you can run Postgres, MySQL, MongoDB, Redis-anything. On serverless, you usually connect to a managed database (AWS RDS, Neon, Supabase). PaaS platforms often bundle a database or let you connect to external ones.

For African startups choosing between managed Postgres options, read Supabase vs Neon vs self-hosted Postgres for African startups. The choice affects your hosting tier.

Scheduled jobs and background tasks are simpler on a VPS (cron jobs). Serverless requires AWS EventBridge or similar. PaaS platforms vary-Vercel doesn't handle scheduled jobs natively; Railway does.

Regional considerations specific to Africa

Africa's infrastructure landscape is improving but fragmented.

Nigeria: AWS Africa Central (Lagos) is now live. This is a game-changer for Nigerian startups. Latency to Lagos is <30 ms. Pricing is higher than US regions but justified for user-facing apps. DigitalOcean still has no Lagos presence, but you can use their Frankfurt region as a fallback. Vercel's edge network includes Nigeria, making it a solid choice for frontends.

Kenya: AWS Africa South (Cape Town) is ~1,500 km away, so latency is ~60-80 ms. Not ideal. M-Pesa and Safaricom-adjacent startups often host on AWS us-east-1 or use Vercel for frontends. For backend APIs, the latency penalty is real.

South Africa: Cape Town has AWS Africa South. Latency is <20 ms. Costs are lower than Lagos. Many SA startups host here.

Pan-African: If your users span the continent, no single region is optimal. Consider a multi-region setup: primary in Lagos or Cape Town, secondary in Frankfurt or Singapore. This adds complexity and cost. For early-stage, pick the region with the densest user base.

Connectivity: Africa's internet is improving but remains variable. ISPs sometimes throttle or route traffic inefficiently. Hosting in-region helps but doesn't fully solve this. A CDN like Cloudflare (which has a presence in Nigeria) can help cache static assets and reduce round trips.

Practical recommendations by stage

Pre-launch (0-100 users): PaaS free tier (Vercel, Railway, Render) or a $5-10 VPS. You're not worried about scale; you're validating. Vercel is ideal for frontends; Railway for full-stack apps.

Early traction (100-10k users): Stick with PaaS if costs are acceptable. If you're hitting ₦20,000+ monthly, consider a VPS in Africa Central (AWS). The latency improvement will be noticeable. If you have a DevOps person, a VPS is a good investment.

Growth (10k-100k users): Evaluate based on your traffic pattern. Bursty traffic? Serverless or PaaS with auto-scaling. Steady traffic? A VPS or PaaS. Start thinking about multi-region if your users span Africa. At this stage, latency is a feature, not a luxury.

Scale (100k+ users): You likely need custom infrastructure. A Kubernetes cluster, managed databases, a CDN, and multi-region setup. Most PaaS platforms don't scale cost-effectively beyond this. You're moving to VPS or managed Kubernetes (AWS EKS, Google GKE). Alternatively, if you're serverless-native, you're optimizing cold starts and database connections aggressively.

Making the decision: a checklist

Before you commit to a hosting model, answer these:

1. Where are 80% of your users? (Determines region; drives latency requirements.)
2. What's your expected traffic pattern? (Steady = VPS or PaaS; bursty = serverless or PaaS.)
3. Do you have a DevOps person or the time to learn? (Yes = VPS; no = PaaS.)
4. What's your runway in months? (Less than 6 = minimize fixed costs; more than 12 = optimize for latency and reliability.)
5. Are you stateful? (WebSockets, long connections = VPS or PaaS; stateless = any.)
6. Do you have database needs that constrain you? (Check Supabase vs Neon vs self-hosted Postgres for African startups.)

FAQ

Q: Should I host in Africa or the US? A: If 70%+ of your users are in Africa, host in Africa (AWS Lagos or Cape Town). Latency matters for user experience. If your users are global, host in the US and use a CDN. The trade-off is cost vs. latency.

Q: Is Vercel good for African startups? A: Vercel is excellent for frontends and JAMstack apps. It has edge locations in Africa, so latency is low. For backend APIs or stateful services, it's less ideal. Many African founders use Vercel for the frontend and a separate VPS or PaaS for the backend.

Q: Can I switch hosting later without rewriting my app? A: Mostly yes. If you avoid vendor-specific features (e.g., AWS Lambda-specific APIs), you can migrate. Expect some downtime and effort. Serverless apps are harder to migrate than PaaS or VPS apps because they're often tightly coupled to the provider's ecosystem.

Q: What about costs in naira? Will currency fluctuations kill me? A: Yes. If you're paying in USD and earning in naira, a weak naira makes hosting more expensive. Hedge by raising in USD or using stablecoins (USDC via Paystack or Flutterwave). Some founders lock in annual plans to reduce this risk.

Q: How do I know if my app is too slow? A: Use tools like GTmetrix, WebPageTest, or your analytics. If your Time to First Byte (TTFB) is >1 second or your full page load is >3 seconds, latency is a problem. Test from Nigeria, Kenya, and other key markets. If latency is the issue, moving to an African region will help more than optimizing code.

What to do next

1. Audit your current or planned tech stack: Read tech stack choices for Nigerian startups in 2026 to ensure your framework and database choices align with your hosting model.
2. If you're choosing a database: Compare managed options in Supabase vs Neon vs self-hosted Postgres for African startups. Your database choice constrains your hosting options.
3. If your users are on slow connections: Learn designing for low-bandwidth. Hosting in the right region is only half the battle; your app architecture matters too.

Start with the hosting model that matches your stage and your constraints. You can migrate later. The goal is to ship fast, keep users happy, and not wake up at 3 a.m. managing servers.