Edge Computing Cost Traps — Vercel vs AWS vs VPS Compared

Edge computing is fast. It serves responses from the server closest to the user — Tokyo for Japanese visitors, São Paulo for Brazilian ones, Washington D.C. for Americans.

But speed comes with billing complexity. I run a multilingual blog (108 articles × 3 languages) on Vercel and hit the free plan's CPU limit. The cause was not understanding how edge-specific usage-based billing works.

This article compares the usage-based billing models of Vercel, AWS Lambda, and Cloudflare Workers against fixed-cost VPS and on-premise hosting. This isn't about which is better — it's about which model fits which project.

What Edge Computing Actually Is

Traditional servers sit in one location (say, Tokyo). When traffic comes from around the world, response times increase with physical distance.

Edge computing runs your code on servers distributed globally (edge nodes). Like CDNs cache static files, edge platforms execute dynamic logic close to users.

Major platforms:

What they share is usage-based billing. You pay for what you use. If nothing runs, you pay $0. The definition of "what you use" differs across platforms, and that's where costs surprise you.

How Usage-Based Billing Works

Vercel (Functions / Fluid Compute)

Vercel Pro costs $20/month as a platform fee, plus a $20 monthly usage credit. Anything beyond the credit is billed on-demand.

Active CPU billing: You're only charged while your code is executing. I/O wait (database queries, external API calls) doesn't count. However, Provisioned Memory is billed for the entire instance lifetime.

Hobby plan (free) caps Active CPU at 4 hours/month. Exceeding it pauses your project.

AWS Lambda

AWS Lambda bills on two axes.

Choosing ARM (Graviton2) saves about 20% ($0.0000133334/GB-sec).

Lambda's model: Memory allocation (128MB–10,240MB) proportionally increases CPU power. More memory means faster execution per request, and since pricing is per GB-second, shorter execution can actually reduce cost.

Cloudflare Workers

Cloudflare Workers has the simplest pricing.

Workers' advantage: Bandwidth (data transfer) is free. You're billed only for request count and CPU time, regardless of response size. Strong for high-volume lightweight operations.

How Fixed-Cost Models Work

VPS (Virtual Private Server)

VPS hosting charges a flat monthly fee for server resources.

VPS pricing model: No per-request, CPU-time, or bandwidth surcharges. Whether you serve 1 million or 10 million requests, the monthly cost stays the same. If the server can't handle the load, you upgrade to a larger plan.

On-Premise

Self-managed servers in your own facility. High upfront costs (hardware, rack space, power, networking) but running costs are limited to electricity and connectivity. At massive scale, this offers the best cost efficiency.

The tradeoff: you handle failovers, security, and hardware lifecycle entirely in-house.

Usage-Based Billing Traps

Trap 1: SSR on Non-Existent URLs

This happened on a site migrated from WordPress. Bots crawling old URLs triggered Next.js SSR (server-side rendering) attempts. The result was a 404, but the CPU time for SSR was already consumed and billed.

On a VPS, returning a 404 costs nothing extra. On edge platforms, each request adds to your CPU bill.

Fix: Set dynamicParams = false to prevent SSR for URLs not in the static params list.

Trap 2: Middleware Running on Every Request

i18n libraries like next-intl use Middleware to detect locale on every request. Static files (CSS, JS, images) are excluded, but every page access passes through Middleware.

On my site, 14.1% of CPU consumption was Middleware — even though the pages themselves were statically generated.

Trap 3: Lightweight Processing × Every Page × Every Visitor

I added a like button that made an API Route call (Redis GET, a few milliseconds) on every article page. The per-request cost was negligible, but 324 pages × every visitor added up.

On a VPS, the same processing completes within the server at zero additional cost.

Which Model to Choose

Rules of Thumb

Edge is the right choice when:

• Your site is multilingual and targets users worldwide
• Google rankings (Core Web Vitals) directly drive revenue
• Your team is small and can't spend time on infrastructure

VPS is the right choice when:

• Your site serves one language in one region
• Traffic is unpredictable (viral potential)
• You need a predictable monthly bill

On-premise is the right choice when:

• Traffic is massive (tens of millions of monthly pageviews) and cost optimization is critical
• Security requirements prevent data from leaving your facility
• You have a dedicated infrastructure team

Wrapping Up

Edge computing delivers "fast" and "cheap" together — while traffic is low. As it scales, usage-based billing takes effect.

Fixed-cost models (VPS, on-premise) offer predictability. Request volume doesn't change the monthly bill. But global delivery and infrastructure management are separate costs.

The right answer depends on the project. The important thing is understanding the billing model differences before choosing. Picking edge hosting just because "it starts free" can lead to an unexpected invoice at the end of the month.

For a detailed guide on controlling Vercel's usage-based billing, see the Spend Management configuration guide.