Vercel

Vercel sells a single promise to founders: connect a git repo, get a global edge URL in minutes, and never log into a server again. In 2026 that promise now extends to AI workloads via the AI Gateway, warmer functions via Fluid Compute, and a Next.js 16.2 build pipeline.

Why Next.js teams default to Vercel

Vercel employs the maintainers of Next.js, so framework features land on the platform first and break the least. **App Router, server actions, partial prerendering, and the Data Cache** all arrive with documented Vercel runtime behavior on the same day they ship in the framework, which is rarely true on third-party hosts.

The day-to-day workflow is what closes the deal. Every pull request gets a unique **preview URL** that designers and PMs can click. ISR refreshes content on a schedule without a full rebuild. **Speed Insights** captures Core Web Vitals from real users and pins regressions to the deploy that introduced them.

• Native Next.js — Turbopack production builds, Server Fast Refresh, and a 4x faster dev server in 16.2 (nextjs.org/blog/next-16-1).
• Preview URLs per PR — production-equivalent deploys without a staging cluster, with optional password protection on Pro.
• Edge runtime + ISR — pages serve from the CDN until invalidated; on-demand revalidation handles freshness.
• Web Analytics + Speed Insights — privacy-friendly traffic and performance data baked into the platform.
• Image Optimization API — automatic resizing, format selection, and lazy loading at the edge.

Pricing in plain English (and the bandwidth gotcha)

Three public tiers, one wrinkle. Hobby is free for personal projects only. Pro is per-seat plus usage. Enterprise is negotiated, with SAML SSO, SCIM, audit logs, and contractual SLAs as the procurement carrots.

Plan	Price	What you actually get
Hobby	$0	100 GB bandwidth, 6,000 build minutes, 360 GB-hours memory, single seat. Fair Use forbids commercial or revenue-generating projects on Hobby.
Pro	$20 per developer / month	1 TB bandwidth, 6,000 build minutes, 1,000 GB-hours memory, 10M edge requests, $20 monthly usage credit, unlimited free viewer seats, password-protected previews. Overages: $0.15 per extra GB bandwidth, $0.0106 per GB-hour memory.
Enterprise	Custom	SAML SSO, SCIM, 99.99% SLA, managed WAF rulesets, multi-region compute, up to 1,000 firewall rules, audit logs, dedicated support, ZDR contracts.

The wrinkle is the **bandwidth gotcha**. All egress counts toward the 1 TB allowance, including bot traffic and malicious requests. Founders should set a hard spend cap on day one, before the first marketing push, full stop.

Founder credits are real. **Vercel for Startups** offers up to $30,000 in platform credits for eligible early-stage teams, with larger tranches available when teams come in through accelerator partners. The application asks for stage, funding, and intended workload — verified founder pricing via SaaSTweaks confirms the credit covers Pro seats and usage in year one for most teams (vercel.com/startups/credits).

What shipped in 2025-2026: Fluid Compute, AI Gateway

**Fluid Compute** is the headline runtime change. Functions stay warm, handle concurrent requests on a single instance, and bill **Active CPU at $0.128 per hour** plus provisioned memory at **$0.0106 per GB-hour**. Cold starts dropped meaningfully versus the legacy serverless model, and the pricing now sits closer to a long-running container than a per-invocation function (vercel.com/blog/introducing-active-cpu-pricing-for-fluid-compute).

**AI Gateway** is the second pillar. It is a single endpoint that fronts hundreds of language models across Anthropic, OpenAI, Google, and others. Every team gets **$5 of free credits per month** that refresh on a 30-day cadence.

Tokens beyond that bill at the underlying provider rate with **zero markup**, including bring-your-own-key. Zero Data Retention now covers the major providers as of April 2026 (vercel.com/blog/ai-gateway, vercel.com/docs/ai-gateway/pricing).

Vercel vs Netlify vs Cloudflare Pages vs Render

The four hosts cover overlapping but distinct sweet spots. Vercel chases Next.js and AI workloads. Netlify chases JAMstack breadth with framework-agnostic tooling. Cloudflare Pages chases bandwidth and global edge reach. Render chases backend-heavy teams who want managed Postgres alongside a web service.

Dimension	Vercel	Netlify	Cloudflare Pages	Render
Cheapest paid plan	$20 / dev / mo	$19 / mo Personal, then $20 / seat Pro	$5 / mo Workers Paid	$7 / mo per service
Free tier bandwidth	100 GB	~20 GB equivalent	Unlimited	100 GB egress per service
Edge functions	Yes (Fluid Compute)	Yes (Edge Functions)	Yes (Workers, 300+ POPs)	No native edge runtime
Best for	Next.js teams shipping fast	JAMstack + framework variety	High-bandwidth global static + Workers	Backend services + managed Postgres

The smaller view that founders care about is **bandwidth and function execution per dollar**, because that is where invoices go sideways.

At a glance	Vercel Pro	Netlify Pro	Cloudflare Workers Paid
Bandwidth overage	$0.15 / GB	$0.55 / GB equivalent	Effectively free for static
Function memory	$0.0106 / GB-hr	Per-invocation tiers	$0.30 / million requests
Edge POPs	~40 regions	~30 regions	300+

Watch a Vercel deploy in real time

The clearest single demo of the platform is the official Vercel Product Walkthrough on the @VercelHQ channel, hosted by the VP of Developer Experience. It covers a full git connection, preview deploy, edge function, and analytics view in one continuous run.

The viral-traffic bill problem (and how to avoid it)

Vercel charges for **all egress**, including traffic from bots and from active attacks. A widely covered 2025 incident saw a side project hit with a roughly **$23,000 invoice** after a sustained DDoS, because every malicious byte counted toward the bandwidth meter (techcrunch.com).

• Spend caps in team billing settings are the single most useful guardrail.
• Managed WAF Rulesets on Pro filter common attack patterns before functions run.
• ISR + Data Cache keep functions idle on most requests.
• External CDN for heavy assets (Cloudflare R2, Bunny) sidesteps the Vercel bandwidth meter for downloads.

"Zero-config preview URLs and the Speed Insights overlay still make this the fastest team workflow on the internet." — G2 verified Pro user, 2026

"A side project went viral, bots flooded the homepage, and a $500 invoice landed before the spend cap kicked in. Set the cap before the launch, not after." — Reddit thread, March 2026

The avoidable mistakes have a short list, and every founder should walk through it before flipping DNS:

• Set a hard spend cap in team billing settings the same day Pro is activated.
• Enable Managed WAF Rulesets on Pro and add IP blocks once Enterprise is in play.
• Cache aggressively with ISR and the Data Cache so functions do not run on every request.
• Watch the Speed Insights tab for traffic spikes that look like crawler floods rather than humans.
• Move giant assets to R2 or a static CDN if the site ships big videos or downloads.

Day one with Vercel: from git push to live URL

The first session is short. Connect GitHub or GitLab, pick a Next.js repo, accept the framework preset, and watch a green deploy land at a `*.vercel.app` URL within minutes. Custom domains and SSL provisioning happen the same afternoon.

The second session is the team setup. Invite engineers as developer seats, add the read-only billing seat, wire Slack notifications for failed builds, and gate previews behind a password if marketing wants closed reviews. By end of week one most teams have CI green-lit, preview URLs landing in PR comments, Speed Insights collecting field data, and the spend cap configured.