Vercel pricing

About

Vercel is the frontend cloud platform built around the Next.js framework. Founded in 2015 by Guillermo Rauch as ZEIT, the company rebranded to Vercel in April 2020 alongside the launch of Next.js 9.4. Vercel raised a $250M Series E at a $3.25B valuation in 2024 and crossed an estimated $200M+ ARR by mid-2025.

The platform is the deployment target for a majority of Next.js production apps and competes with Netlify, Cloudflare Pages, AWS Amplify, and (at the enterprise tier) directly with hyperscaler PaaS offerings.

Pricing is one of Vercel’s most-discussed product surfaces — partly because it’s structurally complex (eight billing dimensions on Pro), partly because real bills routinely exceed the $20/seat headline by 5–15x.

Pricing summary : Hybrid seats plus eight metering dimensions

Vercel uses a hybrid model with two dimensions at the plan level, but eight metering dimensions within the Pro tier:

1. Platform access (seat-based): $20 per developer/month on Pro.
2. Infrastructure consumption (multi-dimensional metering): every Pro team gets a $20 monthly usage credit that absorbs overages across bandwidth, edge requests, function invocations, CPU, memory, image optimization, and builds.

What makes this different: Most PaaS competitors (Netlify, Cloudflare Pages) bundle infrastructure into seat fees or meter on 2–4 dimensions. Vercel exposes 8 dimensions, which gives precise alignment to infra costs but creates real bill-prediction difficulty for users.

On top of the core platform, Vercel’s AI surface is billed per token: v0 (separately-billed product) and, as of the July 2026 pricing page, a new Vercel Agent BETA SKU metered at $0.25 per million tokens plus pass-through model cost.

See the three-plan grid at the top of this page for the core platform tiers. v0 is a separately-billed product — see the v0 sub-section under Pricing by product for its own plan grid, and the Vercel Agent sub-section for the new token-metered AI assistant.

Pricing by product

Core platform (Hobby, Pro, Enterprise)

Sales motions across products: self-serve PLG for Hobby and Pro; sales-led for Enterprise.

Pro plan: every billing dimension

The $20 monthly usage credit absorbs overages in a fixed priority order: bandwidth first, then edge requests, then function invocations. Once the $20 credit is depleted, additional usage bills at the rates above.

Spend management: every team can set an on-demand usage budget (customizable spend limit). Email, web, and SMS notifications fire as the limit is approached, and Vercel can stop usage before you exceed your budget.

Fluid Compute + Active CPU pricing

Vercel’s flagship pricing innovation. Released February 2025 with time-based billing; switched to Active CPU billing in June 2025.

The mechanic: Fluid Compute lets a single function instance handle multiple concurrent requests. Active CPU billing only meters CPU-active milliseconds — the time your code is actually executing, not the time it spends waiting for I/O.

Example impact:

• AI inference call: 50ms compute + 950ms model wait → billed for 50ms (95% savings)
• Database query: 20ms compute + 480ms query wait → billed for 20ms (96% savings)
• Pure-compute function: 100ms compute, no I/O → billed for 100ms (no change)

Vercel publicly cites up to 90% savings for I/O-heavy workloads. This is the first Active CPU billing model from a major PaaS provider.

v0 (AI UI generator) — separately billed

v0 is Vercel’s AI UI generator, billed as a distinct product:

Per-model token pricing (Jun 2026):

Zero markup policy: Vercel’s AI Gateway passes through token costs without markup, including for bring-your-own-key. This is notably distinct from competitors that bundle inference into seat fees with implicit markup.

AI Gateway — separately metered

AI Gateway is pay-as-you-go with zero markup on provider list rates. Every team gets a free tier ($5/month of included credits, Free-Tier-eligible models, no BYOK); the paid tier is purchased AI Gateway Credits with all models, BYOK, and higher rate limits. Some capabilities carry add-on surcharges deducted from the credit balance:

Per-request only provider filtering and per-request ZDR are no additional cost.

Vercel Agent (BETA) — token-metered AI assistant

As of the July 2026 pricing page, Vercel Agent is listed under the AI section as a BETA product with its own line-item pricing on Pro, distinct from v0 and AI Gateway. It is billed per token, not per action: a flat AI Gateway-style rate plus the underlying model cost passed through.

The $0.25 / million tokens is Vercel’s own metering fee on top of the pass-through model token cost — consistent with the zero-markup AI Gateway stance (Vercel charges a thin per-token platform fee and passes the model cost through at list). Code Review and Investigations are the two named Agent capabilities available on Pro and Enterprise.

New BETA infrastructure SKUs (July 2026 pricing page)

The July 2026 pricing page adds several BETA-tagged usage SKUs alongside the established metering dimensions:

These sit under the broader “Vercel Compute” and “AI” groupings on the pricing page and are billed on the same pay-as-you-go, spend-managed model as the core dimensions.

Hidden costs : Why the $20-per-seat headline becomes $400-plus

The $20/seat headline understates real bills by 5–15x at moderate traffic. Two real-world examples:

Medium-traffic Next.js app (500k MAU)

A “$20 plan” that actually bills $481/month. The Active CPU savings (~95% on AI inference routes) materially shifted this number — the same workload billed wall-clock would be ~$900. See our Vercel pricing calculator to model your own usage.

Small SaaS app (50k MAU)

For small apps Vercel stays cheap. The cliff hits between 250k and 1M MAU — exactly where Pro→Enterprise pricing conversations get hard. Bill predictability in usage-based pricing is the unsolved problem here.

Estimate your own bill: Vercel pricing calculator — model traffic, function workload, and Fluid Compute savings interactively.

Pricing evolution : From wall-clock to Active CPU billing

Cadence

Tracked range: 2020 Q2 (Vercel rebrand) – 2026 Q3. Quarters not listed above were verified stable (0 changes, 0 additions).

Notable changes

• 2020-04 — ZEIT rebrands to Vercel alongside Next.js 9.4 launch. Pro at $20/seat with bundled allocations.
• 2024-04 — Granular per-dimension metering replaces coarse bundled allocations. Now bills separately on bandwidth, edge requests, function invocations, CPU hours, memory GB-hours, image optimization, build minutes. Headline price unchanged.
• 2025-02-12 — Fluid Compute execution model launches. Single function instance handles multiple concurrent requests. Initial billing still time-based.
• 2025-06-12 — Active CPU pricing for Fluid Compute. Only CPU-active milliseconds are billed; I/O wait time costs $0. Up to 90% reduction on I/O-heavy workloads (AI inference, database queries).
• 2025-09-15 — Pro plan adds a $20 flexible spending credit that absorbs overages in priority order (bandwidth → edge requests → function invocations). Several Enterprise-only features pushed down to Pro.
• 2026-03 — v0 launches three model tiers (Mini, Pro, Max) with explicit per-million-token rates. Zero-markup AI Gateway policy formalised — token costs pass through at AI Gateway cost, including bring-your-own-key.
• 2026-06-30 — v0’s flagship v0 Max Fast model is cut ~3x, from $30/$150 to $10/$50 per 1M input/output tokens (cache write $12.50, cache read $1). Because v0 passes AI Gateway tokens through at cost, this is the zero-markup policy transmitting an underlying model-price cut straight to buyers rather than a margin decision. Same day, the legacy Ultra/Premium plan was retired, collapsing the v0 lineup to Free ($0) / Team ($30) / Business ($100) / Enterprise — a simplification that removes a tier and pushes the prior Premium audience onto Team. Core platform pricing (Pro $20/seat, eight dimensions, all overage rates) is unchanged.
• 2026-07 — Vercel Agent appears on the main pricing page as a BETA AI SKU, billed $0.25 per million tokens plus pass-through token cost on Pro (Custom on Enterprise), with Code Review and Investigations as its named capabilities. Alongside it, several BETA usage SKUs are surfaced — Vercel Services ($0.50 per 1M requests), Vercel Queues ($0.60 per 1M API operations), and Vercel Container Registry Image Storage ($0.10 per GB-month). The Agent’s per-token metering fee mirrors v0’s zero-markup, token-based model rather than a per-action charge. Core platform pricing (Pro $20/seat, eight dimensions, all overage rates) is unchanged.

What’s unique : Active CPU, zero-markup AI billing extended from v0 to an agentic SKU

1. Eight metering dimensions on one plan. No other PaaS exposes this many axes. Aligns billing precisely to infrastructure cost — and creates the prediction-difficulty complaint that defines Vercel’s pricing reputation.

2. Active CPU is a category-first. Charging $0 for I/O wait is unusual enough that no major competitor offered it as of mid-2026. The savings are real, but they require a Fluid Compute architecture.

3. Priority-ordered credit absorption. The $20 credit absorbs overages bandwidth → edge requests → functions in a fixed order. This is the kind of unglamorous mechanic that materially affects budget planning, and Vercel publishes the order clearly.

4. Zero-markup AI Gateway, now extended to an agentic SKU. v0’s published commitment to pass through AI Gateway tokens at cost (including BYOK) is rare. Most AI-feature-bundled SaaS marks up inference 1.5–3x. The mechanic showed its teeth on 2026-06-30, when v0 Max Fast tokens fell ~3x ($30/$150 → $10/$50 per 1M) — a cut that flows straight to buyers precisely because there’s no margin layer to protect. As of 2026-07-06 Vercel carried the same shape into an agentic product: Vercel Agent (BETA) is billed $0.25 per million tokens plus pass-through model cost — a thin per-token platform fee, not a per-action or per-seat charge. That’s the notable structural choice: the industry norm for coding agents is per-seat or per-action, and Vercel instead priced its Code Review / Investigations agent on the same zero-markup, token-metered rails as the rest of its AI surface.

5. Separate v0 billing — and now a separately-metered agent. v0 is its own product with its own plans — Vercel resisted the temptation to bundle AI UI generation into Pro. The 2026-07-06 Vercel Agent launch repeats the pattern: rather than folding agentic Code Review / Investigations into the $20 seat, Vercel gave it its own token-metered line item. This preserves clean unit economics across the core platform, v0, and now Agent.

Strengths & weaknesses

Billing UX : Spending caps, multi-channel alerts, priority-ordered credits

• Real-time usage dashboard with per-dimension breakdown.
• Customizable on-demand spend limit on every team — set a budget and Vercel stops usage before you exceed it.
• Multi-channel alerts (email, web push, SMS) as caps approach.
• Per-project spending visibility for teams.
• Hard spending cap option blocks deploys/serves once hit (rare for production).
• Pro plan credit priority order published in docs — bandwidth → edge requests → functions.

Strategic wins : The category-defining moves and the principles behind them

1. Active CPU billing is a real customer-friendly innovation

Charging $0 for I/O wait is the kind of pricing move that should be obvious in hindsight but wasn’t. Vercel ate margin (the wait time was previously billed) in exchange for category-defining trust. Every PaaS that follows now has to defend why they’re billing wall-clock.

2. Zero-markup AI Gateway is a defensible moat

For a falling-commodity input like AI tokens, marking up the underlying cost is a race to the bottom. Vercel passing through at cost (including BYOK) means they compete on UX, latency, and developer experience — not arbitrage. Compare to competitors with implicit 2-5x token markups. The 2026-06-30 v0 Max Fast cut (~3x, to $10/$50 per 1M) is the proof of work: a vendor that markups inference has to choose whether to pass a model-price drop along; a zero-markup vendor has no such decision, so the cut reaches buyers automatically. That makes the pass-through credible rather than aspirational.

The 2026-07-06 launch of Vercel Agent (BETA) shows the strategy scaling from a single-product policy into a platform stance. Agent is billed $0.25 per million tokens plus pass-through model cost — a thin metering fee on the same rails as v0 and AI Gateway. Where most coding-agent vendors reach for per-seat or per-action pricing (which quietly re-introduces margin on inference), Vercel priced its agent per token at cost-plus-thin-fee. It keeps the moat coherent as the AI surface widens: buyers can reason about Agent, v0, and Gateway spend with one mental model instead of three pricing philosophies.

3. Priority-ordered credit absorption is honest

Publishing the order (bandwidth → edge requests → function invocations) for how the $20 credit absorbs overages is a small UX detail that builds trust. Most vendors leave this opaque.

4. Separate v0 billing prevents bundling rot

Vercel resisted the obvious move of bundling v0 into Pro. That would have boosted Pro ARPU short-term but destroyed the unit economics on both products. Maintaining clean per-product P&L preserves optionality. The 2026-07-06 Vercel Agent launch held the same line: rather than sweetening the $20 seat with “free” agentic Code Review, Agent got its own token-metered line item — so its inference cost is never cross-subsidised by, or hidden inside, the platform fee.

Areas to improve : Calculator, mid-tier, dimension grouping, spend projections

1. Lead with TCO scenarios, not plan cards

The pricing page shows the 3 plans without surfacing the real bill for archetypal users. A “medium-traffic Next.js app” archetype showing “$20 base + ~$400 typical usage = ~$420/month” would set expectations honestly. Today the $20-vs-$481 gap is the source of every “my $20 plan cost me $286” blog post.

2. Build a public Pro/Enterprise calculator

Vercel ships in-product spend dashboards but no public calculator that lets prospects model their bill before signing up. Every other multi-dimensional PaaS in 2026 (Cloudflare, Railway, Render) ships one. Adding a comparable calculator on the pricing page itself would cut the time-to-clarity for buyers from days to minutes.

3. Add a mid-tier between Pro and Enterprise

The gap is wide: Pro is $20/seat + usage; Enterprise is a custom annual contract (Vercel doesn’t publish a floor). Companies between these — say, $500/month spend with SSO and committed-use discount needs — have no good fit. A “Team Plus” tier at roughly $200/team/mo with SSO + a committed-use discount + audit logs would unlock the missing market segment.

4. Reduce dimension count or improve grouping

Eight billing dimensions is too many to reason about. Group them into 3 logical buckets — Delivery (bandwidth + edge requests), Compute (functions + CPU + memory), Assets (images + storage + builds) — with a single overage credit per bucket. The total bill stays the same; cognitive load drops materially.

5. Spend projections, not just dashboards

The current dashboard shows what you’ve spent. It doesn’t project what you’ll spend at current trajectory. A “your projected month-end bill: $X (currently $Y)” projection — emailed weekly — would prevent ~80% of bill-shock churn cases.

Key takeaways

1. Multi-dimensional metering accurately reflects cost — and confuses customers. Pick a number of dimensions a finance person can reconstruct in a spreadsheet (3, maybe 4). Beyond that, you need a calculator and projection tooling AS A PRODUCT FEATURE, not a footer link.
2. Active CPU is now the bar. Charging for I/O wait time on serverless is increasingly defensive. Vercel changed the category.
3. Zero markup is defensible when the underlying input is a falling commodity — and it scales across products. AI tokens, bandwidth, storage — all candidates. Don’t lock yourself into a margin you can’t defend. Vercel’s July 2026 Agent SKU ($0.25/1M tokens + pass-through) shows the discipline generalising: an agentic product can be priced per token at cost-plus-thin-fee instead of the per-seat/per-action markup most coding agents default to.
4. Publish your credit absorption rules. Small UX detail, outsized trust impact. Customers don’t trust math they can’t reproduce.
5. Resist the bundling temptation. v0 separate from core Vercel preserves unit economics on both. Every “we’ll bundle this for revenue acceleration” move usually destroys two P&Ls.

UBP implications

1. Multi-dimensional metering needs really good prediction tooling. Vercel’s 8 dimensions are accurate to cost but unfriendly to budgeters. The lesson: if you bill on more than 3 dimensions, you must ship a top-tier cost calculator AND in-product spend projections.
2. Active CPU is a viable “fairness” innovation. Charging only for active compute aligns customer cost with vendor cost very tightly. If your product has natural I/O wait, this is a powerful pricing differentiator.
3. Zero-markup gateways are a defensible position — and a repeatable template for new AI SKUs. When the underlying commodity (AI tokens) has falling per-unit cost, marking it up locks you out of price competition. Vercel’s pass-through gives v0 room to compete on UX while leaving margin to the platform layer. The July 2026 Vercel Agent SKU reuses the exact template — thin per-token fee ($0.25/1M) over pass-through model cost — so a new agentic product ships with a pricing story buyers already trust, instead of re-litigating markup.
4. Credit pools with priority order are operationally legible. Telling users exactly how a credit will be consumed is friendlier than opaque absorption. This is a small UX detail with outsized trust impact.

Sources

• Vercel official pricing page (accessed 2026-05-23)
• Vercel functions usage and pricing docs (accessed 2026-05-23)
• Active CPU pricing changelog (accessed 2026-05-23)
• Vercel AI Gateway pricing docs (accessed 2026-05-23)
• v0 pricing page (accessed 2026-05-23)
• Vercel changelog (accessed 2026-05-23)
• Vercel pricing docs (billable metrics) (accessed 2026-05-30)
• Vercel official pricing page (accessed 2026-06-30)
• v0 plans and pricing (accessed 2026-06-30)
• Vercel AI Gateway pricing docs (accessed 2026-06-30)
• Vercel Sandbox pricing and limits (accessed 2026-06-30)
• Vercel Functions (Fluid compute) pricing (accessed 2026-06-30)
• Vercel build pricing — Managing Builds docs (accessed 2026-06-30)
• Vercel official pricing page (accessed 2026-07-06)
• Vercel AI Gateway pricing docs (accessed 2026-07-06)
• Vercel Functions (Fluid compute) pricing (accessed 2026-07-06)
• v0 plans and pricing (accessed 2026-07-06)

Bottom line

Vercel’s pricing is the most accurate reflection of frontend-cloud infrastructure cost in 2026 — and the hardest to predict before you sign up. Active CPU billing and zero-markup AI Gateway are category-defining wins. Eight billing dimensions and the missing mid-tier are unforced errors. The platform’s job over the next year is to ship the prediction tooling that makes the accuracy tractable for budgeters.

Want to compare Vercel against other frontend cloud and PaaS pricing? Browse the pricing blueprint.

Pricing timeline : Major events on a vertical axis

Each milestone below corresponds to a public pricing change, product launch, or material adjustment. Major events use a filled marker; minor adjustments use a faded one.

Vercel Agent (BETA) + New BETA Usage SKUs

Jul 2026

The main pricing page adds Vercel Agent as a BETA AI SKU billed $0.25 per million tokens plus pass-through token cost on Pro (Custom on Enterprise), with Code Review and Investigations capabilities. New BETA usage SKUs also appear: Vercel Services ($0.50/1M requests), Vercel Queues ($0.60/1M API operations) and Vercel Container Registry Image Storage ($0.10/GB-month). Core platform pricing (Pro $20/seat, eight metering dimensions) is unchanged.

v0 Max Fast Token Cut + Ultra Plan Retired

Jun 2026

v0's top model, v0 Max Fast, drops to $10/1M input and $50/1M output tokens (from $30/$150) — a roughly 3x reduction. The v0 plan lineup is Free ($0), Team ($30/user), Business ($100/user) and Enterprise, with the legacy Ultra/Premium tier retired (a "What happened to the Ultra plan?" FAQ now redirects users to the current plans). Core platform pricing — Pro at $20/seat plus eight metering dimensions — is unchanged.

v0 Plans & Pricing Formalized

Aug 2025

v0 ships a dedicated "Plans and Pricing" page — Free ($0), Premium ($20) and Team ($30) tiers with a per-model token usage table, billing v0 usage separately from the core platform.

Active CPU Billing + AI Gateway Pricing

Jun 2025

Functions docs adopt "Active CPU" billing for Fluid Compute — I/O wait time is free, only CPU-active time is charged (up to ~90% savings for I/O-heavy workloads). The AI Gateway pricing page launches with AI Credits and a zero-markup token pass-through.

Granular Metering + Pro Add-ons

Apr 2024

Main pricing page moves to per-dimension metering (bandwidth, edge requests, function execution, image optimization source images, build execution) and introduces a "Customize your plan with Pro add-ons" section. Pro stays $20/user/month plus usage.

Dedicated Functions Usage & Pricing Docs

Feb 2024

A standalone "Usage & Pricing for Functions" docs page documents Hobby, Pro and Enterprise function limits and overage rates separately from the marketing pricing page — formalizing per-dimension function billing.

Pricing Redesign + New Metered Dimensions

Jun 2023

Pro held at $20/user/month but the plan-comparison grid expanded: Edge Functions now metered in execution units (1M included on Pro), plus new rows for Data Cache, Speed Insights, Web Analytics, and KV/Postgres/Blob storage. Web Analytics Plus introduced as a $50/month add-on.

Three-Tier Bundled-Allocation Pricing

Jan 2022

Hobby (free, non-commercial), Pro at $20/mo per member, and Enterprise (custom). Pro bundled 1TB bandwidth, 1,000 GB-hours of Serverless Function Execution, 500K Edge Function invocations/day and a 15s execution timeout — coarse per-tier allocations rather than per-dimension metering.

Rebrand to Vercel + Team Pricing

Apr 2020

Company renames to Vercel. Introduces Pro plan at $20/seat with bundled bandwidth and serverless allocations.

ZEIT Now Launches (Vercel's Predecessor)

Apr 2017

Founded by Guillermo Rauch as ZEIT, launched 'Now' as serverless deployment platform with per-second billing for compute.