Dora NodaIn nine days in April 2026, two of the biggest developer PaaS platforms killed the same pricing lever — and walked away from years of charging you per teammate.
On April 14, Netlify CEO Mathias Biilmann published "The end of seats: pricing Netlify for 3 billion builders," making Pro a flat $20/month with unlimited seats — every developer, PM, and AI agent on the team included. Before that, Pro cost $20 for one seat and another $20 for every additional member or Git contributor. Nine days later, on April 23, Render announced its new workspace plans: Hobby $0, Pro $25, Scale $499 — all flat monthly fees with unlimited members, replacing the old $19/member/month Organization plan. Legacy Render workspaces auto-migrate on August 1.
Two vendors, same month, same shape: per-seat to flat-fee. That is not a coincidence worth noting in passing. It is a category telling you where team-size-based pricing broke as a lever — and why a self-hosted fleet on machines you own was never exposed to that dimension in the first place.
This post puts the math on one page: what each plan actually changed, what the same team costs at 1, 5, 20, and 50 seats before and after, why per-seat pricing cracked under AI agents and plain seat-tax fatigue, and what the self-hosted alternative still does — and does not — save you from.
1. What actually changed — the before/after on one table
Both moves were narrower than they sound. Neither vendor made hosting free. Both kept metered usage (bandwidth, compute, credits) on top of the flat workspace fee. What changed is the team-size line item: whether headcount itself is a multiplier on the bill.
| Platform | Before April 2026 | After April 2026 | What moved | What did not move |
|---|---|---|---|---|
| Netlify Pro | $20 for 1 seat + $20 per extra seat / Git contributor | Flat $20/mo, unlimited seats (published April 14, 2026) | Seat count no longer multiplies the bill | Usage still via credits: bandwidth ~10 cr/GB, compute, web requests, AI inference, deploys. Credit rates for compute/bandwidth/requests were also updated April 14 (bandwidth/compute higher, forms now free). Legacy pre-September-2025 plans keep old pricing unless migrated. |
| Render Workspaces | Organization plan $29 per member per month (billed per seat) | Hobby $0 / Pro $25 / Scale $499 flat (announced April 23, 2026) with unlimited members + 5 / 25 / 1,000 GB included bandwidth at $0.15/GB overage; 15 custom domains free on Pro then $0.25/domain/mo | Seat count no longer multiplies the workspace fee | Per-service compute still metered (Starter $7/mo for 512 MB/0.5 vCPU, Standard $25/mo up to 2 GB). Bandwidth overage and domain fees still scale with usage. |
| Vercel Pro (still per-seat for contrast) | $20 per seat per month (each seat includes $20 usage credit), Pro team still billed per developer seat as of mid-2026 | No change announced — still per-seat | — | Remains the holdout among the three; bandwidth ~1 TB included then $0.15/GB. |
| Railway | Hobby $5 / Pro $20 per seat + usage ($10/GB RAM/mo, $20/vCPU/mo — meter varies by doc version) | Pro workspace still tied to seat count via subscription; no flat-fee announcement tracked as of August 2026 | — | Usage metered per-second on top of the subscription. |
Sources: Netlify's April 14 "The end of seats" post and its April 14 2026 credit-plan changelog; Render's "Better pricing for fast-growing teams" (April 2026), its new workspace-plans docs, and its changelog (updated plans for Render workspaces, legacy migration Aug 1, 2026); comparison trackers (Northflank 2026 guides, GetDeploying, FluxPrice, AgentDeals) cross-checking the same numbers. Vercel and Railway rows reflect mid-2026 comparison maintainors' snapshots — the point is the divergence, not a single doc URL.
Why billing still moves even though seats do not: Netlify moved to credit-based billing in September 2025 first ("one pool of credits for deploys, bandwidth, compute, requests" — Netlify changelog, Northflank's 2026 guide). The April 2026 update then fixed the seat-tax part but also repriced the credit rates. A 10-person team that goes from $200/mo in seats to $20/mo flat still pays bandwidth and compute per credit. The seat line is dead; the usage lines are not.
2. The math in one glance — what the same team costs before and after
Fix everything except team size. One Pro workspace, no change in services. Workspace fee only — compute/bandwidth extra in every row, so this isolates the seat-tax delta that actually moved in April.
| Team size | Netlify Pro before (per-seat) | Netlify Pro after (flat) | Saved by flat fee | Render Organization before (per-seat) | Render Pro/Scale after (flat) | Saved by flat fee | Vercel Pro today (still per-seat, for contrast) |
|---|---|---|---|---|---|---|---|
| 1 | $20 | $20 | $0 — solo dev feels nothing | $29 (1 × $29) | $25 (Pro) | $4 — slight win | $20 |
| 5 | $100 (5 × $20) | $20 | $80/mo, $960/yr | $145 (5 × $29) | $25 | $120/mo, $1,440/yr | $100 |
| 10 | $200 | $20 | $180/mo | $290 | $25 | $265/mo | $200 |
| 20 | $400 | $20 | $380/mo, $4,560/yr | $580 | $25 | $555/mo | $400 |
| 50 | $1,000 | $20 | $980/mo | $1,450 | $499 (Scale tier) | $951/mo | $1,000 |
Read it two ways. At the team level, the flat-fee switch is a cliff: a 10-person Netlify team went from $200/mo to $20/mo on the workspace line overnight — a 90% cut that has nothing to do with how much bandwidth they push or how many functions they run. At the per-person level, the old model charged $20–$29/head for the privilege of deploying. The new one charges $0.40–$2.50/head at 10–50 people. That is the seat tax in a single number.
The breakpoint is trivial — every team larger than one person saves money, and the savings grow linearly with headcount while the new price stays flat. No sensitivity analysis needed: there is no workload where the old per-seat line was cheaper past two seats, because it was a pure headcount multiplier with no usage included to offset it.
Include compute and the ordering still holds. A two-service stack (API + worker + Postgres) at ~$80–$95/mo in compute looks similar on either side of the date. The difference is whether a 20-person team pays ~$460–$495 (compute + flat fee) or ~$840–$1,100 (compute + per-seat fee) for the identical deploy and traffic. Seats, not servers, drove the variance — precisely the line both vendors erased.
3. Why per-seat pricing broke — three pressures in one month
Nine-day convergence does not happen because two CEOs woke up with the same idea. It happens because the same three pressures hit every seat-taxed PaaS at once.
Seat-tax fatigue is not new, but it became measurable
Per-seat pricing worked when the PaaS buyer was a small frontend team (3–5 committers) and the seat count tracked value. By 2024–2025 that stopped matching reality:
- Every
git pushby a PM, designer, or contractor could become a billable "Git contributor" on Netlify. Teams learned to police who merged PRs to avoid triggering a $20 seat — tax behavior, not product behavior. - Comparison audits in 2025–2026 routinely flagged "every repo committer is charged as a full Pro seat" as Netlify's top gotcha, and "$19/member/month" as Render's. Both entries disappeared from gotcha lists the week of the announcements.
- Surveys and pricing-analysis trackers (FlexPrice, AgentDeals, Northflank) started modeling the team-size column separately from the usage column — the act of modeling it betrayed that it had become a standalone cost center readers shopped on.
When the lever you price on becomes the thing customers optimize to avoid, the lever is brittle. Netlify and Render both said so in their own framing: Render modeled "75% of paying customers see costs decrease or stay flat" under flat fees; Netlify reached out only to the ~2% of high-bandwidth/compute-heavy users whose usage outpaced the seat savings.
AI agents do not sit in seats
This is the industry-wide forcing function, not a niche complaint. IDC's 2026 framing — "by 2028, pure seat-based pricing will be obsolete as AI agents replace manual repetitive tasks" — landed on the same premise every SaaS pricing essay made in 2025–2026: an agent doesn't log in, doesn't occupy a seat, and may do the work of ten humans while billing for zero. AWS's agentic-pricing paper, FourWeekMBA's "Seat Tax" analysis, and Stacktree's June 2026 audit ("The end of per-seat pricing: AI agents have no seats") all formalized it:
- If an agent generates 500 deploys, 2M tokens, and continuous preview environments without a human clicking, seat count undercounts value while credit/bandwidth consumption overcounts it — or vice versa, depending on which side you monetize.
- Netlify made this explicit: Pro's flat $20 now covers "every developer, internal builder, PM, marketer, and AI agent on the team … without per-seat costs" (via its distribution partners and changelog mirrors). That line is not marketing poetry; it is a billing architecture decision pegged to the agent era.
- Railway and Vercel, still per-seat, face the one question a seat-tax vendor cannot answer cleanly: does your build agent count as a seat? Flat-fee vendors do not have to answer it at all.
The April 14 timing is telling. Netlify titled the post "for 3 billion builders" — a direct nod to a world where builders include agents. Flat-fee team pricing aligns the meter you charge on (usage, when it applies) with the actor you serve (humans + agents). Per-seat pricing misaligns both.
Growth made the penalty visible
A vendor prices on seats when it wants revenue to scale with adoption. That works until adoption is the thing customers resent being taxed. Concretely:
- A startup at 5 people pays $100/mo in seats. At 20 — still a small company — it pays $400–$580/mo before a single extra gigabyte or function. The seat bill quadruples while engineering output roughly doubles.
- Finance teams started reporting seat-count volatility as the biggest swing in hosting spend — not because the bill was high, but because hiring one person moved it. That makes headcount planning a procurement decision.
Both announcements explicitly sold predictability. Render's blog: "Better pricing for fast-growing teams." Netlify's subhead: "build without barriers." The marketing is different; the promise is identical: your next hire will not change your hosting invoice.
4. Why a self-hosted fleet was never in this category — and what it still does not protect you from
Here is the simplest table in the post, and the one that makes the bex comparison honest.
| Question | Hosted PaaS (per-seat era) | Hosted PaaS (flat-fee era) | Self-hosted PaaS on owned hardware (bex / Cluster API) |
|---|---|---|---|
| Does adding a teammate change the hosting bill? | Yes — one more seat, one more $19–$20/mo | No — flat workspace fee, unlimited members | No — and never did. There is no workspace tier, no seat meter, no Git-contributor gate. The box does not know how many humans pushed to it. |
| Does adding a teammate change any infra cost? | No — seat is pure vendor margin | No | Only if you genuinely need more RBAC or audit tooling — usually no. |
| What still scales the bill when you grow? | Services, bandwidth, build minutes, credits, domains | Same usage lines | Hardware, bandwidth (but Hetzner includes ~20 TB per instance), power, ops — and since 2026, rising DRAM/NVMe prices (Hetzner's April and June repricings: dedicated AX42 from €47.30 → €57.30, cloud 30–176% depending on family). |
| When does the vendor reprice you unilaterally? | Whenever it announces — seats, credits, overage | Whenever it updates credit/domain rates | When the hardware vendor reprices the substrate, not when a PaaS revs its plan. But it still reprices — see Hetzner's two 2026 hikes. |
| Who absorbs the overage? | You pay overage to the PaaS | You pay overage to the PaaS | You absorb capacity planning — idle hardware is your over-provisioning, not a vendor surcharge. |
The self-hosted point is not "owning hardware is cheaper than a flat $20" — at one price comparison it is not. The point is exposure: the self-hosted fleet was never exposed to a billable dimension (seat count) that two well-funded, sophisticated competitors simultaneously decided was unsustainable as a lever. If the two vendors that harvested the most margin from that dimension walked away from it in the same month, the dimension was not pricing power — it was risk that had to be retired.
That does not mean owned hardware is risk-free. Any 2026 fleet that treated "a €7.99 CPX22" as a fixed constant had that assumption repriced twice by its own vendor (see this list's Hetzner hikes coverage). Cluster API lets you move a node pool between providers (Hetzner, UpCloud, any CAPH-compatible EU target), but the DRAM-supply shock of 2026 lifted prices at every vendor that buys RAM on the same market. Flat-fee PaaS removes one unilateral repricing risk; owning hardware removes a different one. Neither removes all of them.
The durable advantage of the self-hosted model is narrower and more boring — and therefore more honest: no vendor can add, or reprice, a billable dimension that maps to a headcount or workspace tier. They can still reprice the commodity they all buy (RAM, SSD, egress) because they buy it from the same silicon market you do. But they cannot invent a "per-Agent seat" tomorrow and put your CI bot on it. That inventable dimension is what Netlify and Render just agreed, within nine days, was not defensible — by retiring it themselves.
5. What to actually do with this — three checks before your next invoice
If you are on Netlify or Render today: nothing urgent. Verify which plan you are on. Netlify accounts created before September 2025 keep legacy pricing unless migrated — check netlify.com/pricing vs your dashboard. Render workspaces on legacy Organization plans auto-switch to Pro on August 1, 2026 unless you opt in earlier. In both cases, audit the non-seat lines that survived the repricing: credit/bandwidth rates (Netlify) and bandwidth included vs overage plus domain fees (Render). A team that saves $380/mo on seats but ignores a doubled bandwidth credit rate may give part of it back on traffic.
If you are on Vercel or Railway today: model the seat line explicitly. At 10–20 people, per-seat Pro is a $200–$400/mo workspace fee before compute; flat-fee peers charge $20–$25. That delta pays for a CPX22-class box with money left over. The question is not "which PaaS is cheapest per service?" but "at what team size does the seat tax exceed any usage delta?" For most teams beyond five, the answer is already obvious.
If you are evaluating self-hosting: price the two costs a hosted PaaS cannot reprice at will, and the one it can. On an owned box, team size and workspace tier are genuinely zero — no seat meter to retire later, no Scale-plan gate at $499. What moves is the hardware you buy (and in 2026 it moved 30–170% on DRAM-sensitive families). Keep that sensitivity in the model you compare against — daily at hetzner.com/pricing, not frozen at a January price — and the flat-fee convergence above stops being surprising. The vendors that could price on headcount chose to price on hardware and traffic instead, because hardware and traffic are what the fleet actually has to provision.
Netlify pushed the first domino on April 14. Render pushed the second on April 23. The next domino is the PaaS still charging per seat — and whether its customers decide to wait for it to fall or to step off the seat-taxed fleet entirely.
Bex.co is the open-source, AI-native Render alternative — push a git repo, get a running HTTPS service on machines you own. No per-seat workspace fees to walk back later: Render-compatible API, bex.yml, and a Cluster API fleet behind a simple git-push surface. Star the repo on GitHub or deploy your first app today.
Sources
- Netlify — "The end of seats: pricing Netlify for 3 billion builders" (Biilmann, April 14, 2026) and Changelog "Pricing updates for Credit-based plans" (April 14, 2026); credit-based plans launch (September 2025).
- Northflank — "Vercel vs Netlify: which deployment platform should you use in 2026?" (credit-model summary, April-2026 seat removal note).
- Render — "Better pricing for fast-growing teams" (April 23, 2026); Docs "New Workspace Plans" (Hobby $0 / Pro $25 / Scale $499, unlimited members, legacy $29/member, auto-migration Aug 1 2026; bandwidth/domain rules); Changelog "Updated plans for Render workspaces."
- Render docs mirrors & comparators — GetDeploying (Render pricing review, flat-fee workspace + compute proration); Northflank "Railway vs Render" (Render removed per-seat in 2026, no per-member charges; Hobby vs Pro distinction).
- Pricing-analysis aggregators (mid-2026 snapshots) — FlexPrice "Complete Guide to Netlify Pricing and Plans 2026"; AgentDeals / Stacktree per-seat audits (Vercel $20/seat, Netlify $19/seat before April, Railway $20/seat Pro); PriceTimeline (Netlify plan history).
- Agent-era per-seat failure — IDC via CIO.com ("By 2028, pure seat-based pricing will be obsolete"); AWS "Why SaaS companies must transform for agentic AI"; FourWeekMBA "Seat Tax" / Slashdot "Software isn't dead but its cosy business model might be"; Stacktree "The end of per-seat pricing: AI agents have no seats" (June 2026).
- Context on hosting repricing wave — Techsy / Railway vs Render vs Fly.io benchmarks (2026); Hetzner April/June 2026 repricings and DRAM-supply commentary (secondary context for owned-hardware sensitivity — see this list's prior Hetzner coverage).