Here is the trap built into the old SaaS contract. A customer deploys your AI tool, automates the work of its support team, and cuts that team from 50 agents to five. The output is the same. Your revenue just fell 90% — because you were never charging for the work. You were charging for the chairs. When one AI agent can do the work of ten support reps, per-seat pricing punishes the customer for becoming efficient, and punishes the vendor for delivering the efficiency. That is why the per-seat subscription is in structural decline. For a generation, SaaS charged per user per month and nobody asked questions. In 2026, the companies winning on net revenue retention share a common trait — they stopped counting seats and started metering value.
The reckoning did not start with AI. It began in the 2022–2024 correction, when buyers started auditing unused seats and negotiating hard on every renewal. AI-native tools then accelerated the break in 2025 and 2026 by introducing consumption patterns that flat per-seat billing simply cannot describe. The churn shows up in pricing velocity: among the top 500 SaaS and AI companies with transparent pricing, there were over 1,800 pricing changes in 2025 alone — 3.6 per company. An industry does not rewrite its price list three and a half times a year because it is fine-tuning. It does it because the old unit stopped matching reality.
Three in five now meter something
The adoption numbers have crossed the point of argument. OpenView's research puts more than 60% of SaaS companies offering some form of usage-based billing, up from 27% in 2018. A January 2025 survey by Metronome and Greyhound Capital, covering 100 SaaS companies, found 85% already had usage-based pricing in place. Different samples, same direction — and barely a quarter of the industry metered anything at all in 2018.
Demand is pulling from the buyer side as well. IDC finds usage-based options now capture 42% of purchasing preference, prepaid and post-paid variants combined, overtaking traditional subscriptions at 38%. And 59% of SaaS businesses expect usage-based approaches to grow as a share of revenue, up 18% from 2023. When buyers prefer the meter and vendors expect the meter to grow, the negotiation is no longer about whether — only about the shape of the blend.
Hybrid is where the market is landing. Chargebee's 2025 State of Subscriptions report puts hybrid adoption at 43% of companies today, projected to reach 61% by the end of 2026. The reason is performance: hybrid firms report 38% higher revenue growth and 38% higher net revenue retention than pure-subscription peers, and high-growth SaaS companies — growing at 40% year-on-year and up — that run hybrid models post a 21% median growth rate, ahead of both pure-subscription and pure-consumption rivals. The retention gap deserves emphasis, because retention is where SaaS economics compound; it is the same compounding that governs the customer record in first-party data strategy and identity resolution, where value accumulates in the accounts you keep rather than the ones you win.
The forecasts extend the line. Gartner predicts 70% of businesses will prefer usage-based pricing over per-seat by 2026, and that 70% of vendors will refactor their pricing by 2028. Stripo's 2026 pricing research adds the sharpest variable: 73% of SaaS vendors now charge separately for AI features. AI monetization is the fastest-moving part of the market, and it is dragging the rest of the pricing architecture with it — because AI costs scale with use, a flat seat price for an AI feature is a margin accident waiting for a heavy user.
Seats, credits, or both
There is no single successor model. Three are competing, and the winners mostly blend them.
Pure usage is the veteran: pay per action — API calls, storage, compute minutes. Twilio, Stripe and most AI APIs have worked this way for a decade. Snowflake is the extreme case, charging storage per terabyte and compute per second with no per-seat metric anywhere in the contract, and it has reached a 158% net revenue retention rate, among the highest in public SaaS. When the meter is the product's natural unit, bill and value stay aligned almost by construction — customers expand because using more is worth more.
Credits are the insurgent. Customers prepay for bundles and spend them down on actions, which gives finance a subscription-like number to approve while usage flexes underneath. The model is up 126% year over year, and 79 of the top 500 SaaS companies now offer it, up from 35 a year earlier. Figma, HubSpot and Salesforce all adopted credit models in 2025; Box gives 20 AI credits per user per month. Credits are best understood as a bridge technology — a way to meter without frightening procurement.
Hybrid is the heavyweight: a base subscription, with usage-based overage or add-ons on top. Datadog runs it in layers — core monitoring per host, log management per gigabyte ingested, APM per traced request — so every product line opens another usage vector inside accounts it already holds. The company generated $3.43 billion in revenue in 2025, up 28%, per industry analysis of its hybrid model. The pattern across all three models is the real finding from 2026's pricing research: the pricing decision has become a retention decision. Vendors increasingly align price with how customers derive value — usage, outcomes, or a blend — instead of counting seats.
Customer success, rewritten around the meter
Change the pricing and you change the jobs downstream of it, starting with customer success. Under per-seat contracts, the CSM's world was the renewal: defend the seats, prevent the downgrade, keep the logo. Under consumption, the job becomes consumption health: is usage growing, is it healthy growth, and does the customer understand the value behind the meter they are paying?
Four practical shifts follow. Onboarding stops chasing seat fill and starts chasing activation depth; time-to-first-value-action replaces time-to-full-deployment as the number that matters. Expansion turns product-led, because the usage data itself identifies which accounts are ready for more — the sales motion follows the meter instead of the calendar. Churn signals arrive earlier, since a drop in consumption precedes cancellation by weeks or months. And the renewal conversation changes character entirely: when the bill reflects actual consumption, the discussion is about ROI, not headcount. Increasingly, the teams doing this work are distributed themselves — vendors hire across borders without setting up local entities to staff them.
The catch is variance. Usage-based suppliers show significantly higher spend variance than fixed-cost alternatives, and average enterprise software spend is forecast to rise at least 40% by 2027, with generative AI as the primary accelerant. That is why guardrails have become table stakes rather than features: spend alerts, caps, committed-use discounts — the machinery that keeps a metered relationship from becoming an ambush.
The risks are real, and so is the test discipline
None of this is a free upgrade. Revenue gets less predictable. Forecasting gets harder. Bill shock is a genuine churn driver. And the AI cost curve hides a landmine: scaling AI from pilots to production has routinely revealed cost underestimation of 500–1,000%. A pilot that looks cheap at small scale can return a production bill five to ten times the estimate once usage compounds — more users, more calls per user, AI features switched on by default. The customer does not experience that as a pricing model working as designed. They experience it as a betrayal, and betrayed customers churn at renewal whatever the contract says. Spend alerts, caps and committed-use discounts are not generosity. They are how a vendor proves the meter is honest before the invoice has to argue the point.
Pricing power has hard limits too. IT budgets are growing roughly 2.8% a year while SaaS prices have hiked 9–25%. Every renewal is becoming a negotiation conducted inside that squeeze. It lands hardest on lean teams — the ones running the bootstrapped playbook cannot absorb a botched repricing the way a funded giant can — which is one more reason the durable answer emerging from the data is hybrid rather than pure consumption.
With 1,800-plus pricing changes among the top 500 in a single year, pricing has become a continuous experiment, and the best-run experiments share a discipline. Grandfather first, optimize second: existing customers keep their plans while new cohorts trial the new model, trading short-term expansion for the trust that lets a usage model survive its first bill-shock incident. Meter before you monetize: leading teams instrument usage two to three quarters before attaching a price to it, building the consumption baselines that make fair pricing possible. And package the escape hatches: committed-use discounts, spend caps, annual prepay options, the machinery that converts unpredictable consumption into forecastable revenue.
The test of all this arrives on a schedule. Chargebee's projection — hybrid adoption reaching 61% by the end of 2026 — is close enough to check, and Gartner's expectation that 70% of vendors will refactor pricing by 2028 gives the experiment a second checkpoint. If the hybrid share stalls while bill-shock churn climbs, the market will have learned that metering has a ceiling. If it lands on target, the per-seat era will be over in everything but legacy contracts.
One constituency deserves a mention before the forecasts get graded: the customers themselves, whose preference data — 42% for usage-based options against 38% for traditional subscriptions — is easy to misread as enthusiasm for meters. It is better read as enthusiasm for fairness. Buyers watched the seat model charge them for shelfware through a correction, then watched AI collapse the relationship between headcount and output. Their demand is not for complexity; it is for a bill that argues its own case.
Retention math lands differently depending on how the company was built. For bootstrapped micro-SaaS founders running the lean playbook to $10M ARR, every point of NRR is survival rather than valuation dressing. The customer-success teams protecting that number are also increasingly distributed, and the best of them operate like the asynchronous remote enterprise — health checks and expansion plays documented in writing, not buried in call recordings nobody replays.
Sources
- Stripo Research — "SaaS Pricing Trends 2026" (per-seat decline; 42% usage-based option; Metronome/Greyhound 85%; 73% charge for AI): research.stripo.email
- GitHub / bbinto podcastit — "Hybrid Pricing in SaaS 2026" transcript (OpenView 60%; Chargebee 43%→61%; 38% growth/NRR uplift; Snowflake 158% NRR; Datadog $3.43B): github.com
- PPL.ug — "The Future of SaaS: Trends Every Startup Should Watch in 2026" (IDC 42% vs 38%; 59% expect usage growth): ppl.ug
- Superframeworks — "The End of Per-Seat Pricing" (1,800 pricing changes; credit model +126%; Gartner 70%): superframeworks.com
- GitHub / danbonfim pricing-monetization-ops — sourced benchmarks (77% of largest software cos use consumption; $6.5B usage-billing market 2026; hybrid 43%): github.com
