The H1 2026 SaaS pricing report

SaaS pricing in 2026 is splitting three ways: pure seat-based is losing share to hybrid base-plus-meter, AI products are climbing an outcome-pricing curve from tokens to per-resolution to per-agent, and roughly 60% of SaaS now runs some form of usage-based component. This report breaks down the data, the model mix, and what investors expect on your pricing page.

• SaaS pricing trends 2026: the model mix
• Seat based vs usage: why seats are losing share, but not dying
• Usage based pricing 2026: what the public market shows
• AI pricing models: pricing against token costs
• The hybrid pricing model: base fee plus meter
• The outcome pricing maturity curve
• Pricing benchmarks SaaS investors actually look at
• Why this matters for your raise

Most pricing guides written in 2026 still treat the question as "pick a model." That framing misses the actual shift. SaaS pricing in H1 2026 is no longer about choosing between three boxes; it is about which atomic unit of value you can defend to a board, a buyer, and an AI cost ledger at the same time.

The unit used to be a seat. Now it is sometimes a seat, sometimes a token, sometimes a successful outcome, sometimes a full agent that replaces a head of headcount. Treating these as the same conversation is what makes a pricing page look legacy in 2026.

This is the H1 2026 SaaS pricing report: the data, the models, the live examples, and the unit-economics floor each one assumes.

SaaS pricing trends 2026: the model mix

Roughly 60% of SaaS companies now run some form of usage-based pricing, with 46% on a hybrid usage-plus-seat structure and 15% on largely usage or pay-as-you-go, according to OpenView Partners. Pure seat-based is no longer the majority position.

The model mix matters because it changes the conversation about NRR, ARR predictability, and AI gross margin in the same breath. Here are the four models a founder will pitch in 2026, what each is for, and what each fails at.

The OpenView mix understates how fast outcome-based pricing is climbing, because most outcome-priced AI agents are sub-$10M ARR and not yet in the benchmark sample. Treat the table as a 2025 snapshot. By the end of 2026 expect the outcome row to capture share from both the hybrid and pure-usage rows above it.

Seat based vs usage: why seats are losing share, but not dying

Seat-based pricing is uneconomic for AI products and overpriced for some human end-user products at the same time, which is the actual reason it is losing share.

a16z's December 2024 enterprise note put the point bluntly: AI inference cost scales per token, so "every API call, every token processed, adds to their cost structure," which makes the per-seat unit "no longer the atomic unit of software." Once a software company has a per-token COGS line, charging a flat seat fee is a bet that the buyer will not use the product. That is not a great bet.

The counter-evidence is that the seats that survive are pricing power up. Slack, Notion, and Figma are charging more, not less. The right read: per-seat is shrinking as a share, while the surviving seat products are gaining pricing power. Use the model where consumption is stable and marginal cost is near zero. Stop using it where the user is an LLM call.

✅ Good: $39/seat for a collaboration tool with stable per-user storage cost and zero LLM inference. The unit (a human) maps cleanly to the cost (a license slot).

❌ Bad: $39/seat for an AI assistant that hits a frontier model on every keystroke. The unit (a human) does not map to the cost (variable inference per query).

Don't default to seat-based because it makes your ARR math easy on a deck. Investors in 2026 read flat seat-based pricing on an AI product as a tell that the founder has not modeled inference cost.

Do keep seat-based pricing where the buyer wants budget predictability and the marginal compute cost is genuinely flat. Internal collaboration tools, productivity apps with no LLM inference dependency, and design tools all sit cleanly here.

Usage based pricing 2026: what the public market shows

Usage based pricing in 2026 has the strongest public-market evidence behind it of any SaaS model.

Public SaaS companies that run usage-based models grew revenue 54% faster than the broader SaaS index and trade at a 50% revenue-multiple premium to peers, with seven of the nine best net-dollar-retention IPOs over the prior three years running a usage-based model. That is not a vibe. That is the comp set your Series A lead is using to value you.

NEA's investment memo on Metronome, the leading usage-based billing infrastructure player, called this shift "the next tectonic shift in software business models" and identified the hybrid form (UBP plus seat) as the modal endpoint, not pure pay-as-you-go. That is the cleanest investor articulation of why this is happening.

The trade-off founders underestimate: usage-based pricing trades ARR predictability for NRR upside. Pure pay-as-you-go gives you high NRR on power users and ARR forecast variance that makes a CFO twitch. That is why pure UBP is only 15% of SaaS and the dominant mode is hybrid.

Don't ship pure usage-based pricing on a seed-stage AI product. Your board deck will show MRR moving by double-digit percentages month over month, and that volatility will get read as churn.

Do ship a hybrid floor with a usage meter on top. The base fee buys you ARR predictability for the board deck; the meter buys you the NRR upside for the next round.

AI pricing models: pricing against token costs

AI pricing models in 2026 are not just usage-based. They are inference-cost-aware, which is a different constraint.

Google AI's Vikas Kansal, writing in Lenny's Newsletter, spells out the asymmetry: "In traditional SaaS, serving an extra free user costs essentially zero. In AI, every time a free user hits Enter, your GPUs fire, and your cash burns." That asymmetry breaks the SaaS freemium playbook and forces AI founders to gate aggressively.

Three concrete 2026 anchor prices a founder can copy:

• Intercom Fin at $0.99 per resolution. Outcome-priced AI agent, not a per-seat add-on. The customer pays for the work, not the tool.
• Gemini Advanced at $20/month with 1M token context. Consumer-tier subscription floor; the bet is that aggregated free-tier inference cost stays below the premium-tier subsidy.
• GitHub Copilot on AI Credits from June 1, 2026. The default AI dev-tool, moving from flat $19/month to per-token AI Credits. When the de-facto market reference shifts to credits, founders pricing flat-rate AI look frozen.

Sequoia and Paid CEO Manny Medina describe an AI pricing maturity curve: activity-based (count tokens or usage), workflow-based (charge per completed process), outcome-based (get paid for the result), and per-agent (replace a human FTE). The successful AI apps "print money" by picking one friction-heavy workflow and climbing the curve deliberately. Picking a token meter and never moving up the curve leaves the outcome margin on the table.

Don't price an AI product flat-rate at parity with a seat-based SaaS comp. You will lose margin every time a power user fires the model. The flat $19 GitHub Copilot price is being killed for exactly this reason.

Do model your inference cost per active user before publishing any flat tier. If the cost-to-revenue ratio looks shaky on day one, the tier is wrong.

The hybrid pricing model: base fee plus meter

Hybrid pricing won the 2025-2026 cycle. The base fee plus meter is the default architecture for any SaaS product with variable usage cost.

The hybrid is the largest single category in the 2025 mix at 46% of all SaaS companies. NEA explicitly called hybrid UBP-plus-seat "the next tectonic shift in software business models" when they invested in Metronome. The argument has shifted from "do we add a meter?" to "which meter, and how big is the bundled allowance?"

The mechanics of a defensible hybrid:

• Base fee floor. Sets the ARR contribution and pays for the platform components that cost the same whether the customer uses them or not (storage, dashboards, support).
• Meter on the variable cost line. Tokens, API calls, transactions processed, leads generated, tickets resolved. The meter must be tied to something the customer feels as value, not something the customer feels as a tax.
• Headroom on the meter. Most meters ship with an included usage bundle inside the base fee so the customer can use the product without feeling a coin-slot. The bundled allowance is the part of the meter that matters most for retention.
• One overage rate, not a tier ladder. Three meter tiers reads enterprise-grade and complicated on a pricing page. One overage rate is what the modal hybrid uses in 2026.

Don't ship a hybrid with a meter the customer cannot predict. If your meter is "API calls" but the customer cannot tell from your docs how many API calls a typical workflow makes, you have a churn problem dressed as a pricing problem.

Do publish a usage calculator next to the pricing page. Pricing-page bounces on hybrid models concentrate on calculators that do not load.

The unit used to be a seat. Now it is sometimes a seat, sometimes a token, sometimes a successful outcome, sometimes a full agent that replaces a head of headcount.

The outcome pricing maturity curve

Outcome-based pricing is the destination most AI founders should be planning toward, even if they ship a hybrid first.

Sequoia's "Services: The New Software" makes the case in dollar terms: a $50/user/month SaaS seat compared with a $1,000/month AI agent priced on a specific outcome. The TAM for the labor a CRM supports is around $1 trillion; the TAM for CRM software itself is around $60 billion. The next $1T company will be "a software company masquerading as a services firm." The pricing model is what unlocks that TAM ratio.

The Manny Medina / Sequoia maturity curve, with the founder action at each step:

1. Activity-based. Count tokens or actions, charge per unit. Founder action: ship the meter; do not optimize it yet.
2. Workflow-based. Charge per completed process (lead enriched, NDA drafted, ticket resolved). Founder action: stop counting tokens externally; charge per workflow even if you bill yourself internally per token.
3. Outcome-based. Get paid for the result (deal closed, dispute won, candidate hired). Founder action: take attribution risk. Customers will pay materially more if you stand behind the outcome.
4. Per-agent. Replace a human FTE; the agent is the price unit. Founder action: anchor against fully-loaded labor cost, not against SaaS seats.

The example economy is already visible. SignalFire's Networked SaaS write-up names AI startups handling "billions of dollars in transaction volume annually" with monetization via transaction fees, premium features, and analytics, not per-seat. The unlock is the work budget, not the tool budget.

Don't stay at activity-based pricing because it is the easiest to instrument. Activity is the smallest unit of value, which means it captures the smallest share of the budget.

Do plan a 12 to 18 month migration up the curve. Ship activity-based at launch so you can charge anything; publish workflow-based pricing as soon as you have data on what a workflow costs; start an outcome-pricing design-partner pilot in parallel.

Pricing benchmarks SaaS investors actually look at

Pricing benchmarks SaaS founders pitch with in 2026 are not the same numbers operators look at internally.

For a seed or Series A pitch, investors read your pricing model as a leading indicator on three lines: NRR, AI gross margin, and ARR forecast quality. The implicit benchmarks:

• NRR is the lead metric. Hybrid models with a healthy meter typically land above the SaaS-index median; pure usage models clear higher still on power users. If your NRR is dragging the comp set down, the meter design (not the price point) is usually the problem.
• AI gross margin pressure. If your inference cost eats too large a share of revenue, your model looks like a services business and your multiple compresses. Investors will not say a floor number out loud, but they compare your gross margin line against the SaaS index reported by High Alpha and OpenView every time.
• ARR forecast quality. Hybrid base-fee floor locks most of next-quarter ARR before the quarter starts. Pure usage leaves a larger share variable, which is why CFOs prefer hybrid for early-stage rounds.
• Price increase cadence. First Round Review's pricing canon names "never revisiting price" as one of the four ways startups fail at pricing. Annual price increases are the minimum cadence investors expect.
• Single value metric. Investors will ask, "what is the one number on your customer's side that goes up when they win?" If the answer is "users" on an AI product, you have a pricing model mismatch. If it is "tickets resolved," "leads enriched," or "dollars saved," you have a defensible value metric.

YC's Tom Blomfield tells B2B founders explicitly to anchor pricing with a written "value equation" co-created with the customer's champion, then warns against "ludicrously low numbers" like $19/mo because B2B software routinely commands tens or hundreds of thousands of dollars. The benchmark investors want to see is annual contract value (ACV) that matches the value equation, not a number that fit on a self-serve pricing page.

If your pricing model is investable but your outbound is the bottleneck, tools like Causo handle the investor-targeting and personalization layer. They do not fix a pricing model that prices an AI product at flat per-seat.

Why this matters for your raise

Pricing pages are read closely in VC diligence after the deck. The pricing model is read as a proxy for how the founder thinks about defensibility, gross margin, and net dollar retention. A pricing page that ships flat per-seat on an AI product in 2026 is read as a founder who has not modeled the unit economics; a pricing page that ships outcome-based on a non-AI tool is read as a founder reaching for hype. Pricing right is one of the few moves where a seed-stage founder can shift the valuation conversation without changing the product.

FAQ

What pricing model do SaaS startups use in 2026? Roughly 60% of SaaS companies run some form of usage-based pricing, with 46% on a hybrid base-fee-plus-meter structure and 15% on largely usage or pay-as-you-go, according to OpenView Partners. Pure seat-based is the residual minority, used mainly where the end user is a human and the marginal cost is flat. Outcome-based pricing is emerging fastest among vertical AI agents.

Is usage-based pricing winning? By the public-market data, yes. Usage-based SaaS companies scaled revenue 54% faster than the broader SaaS index and trade at a 50% revenue-multiple premium. Seven of the nine best net-dollar-retention IPOs over the prior three years ran a usage-based model. The dominant architecture is hybrid (base fee plus meter), not pure pay-as-you-go.

How are AI products priced in 2026? AI products are pricing on a maturity curve from activity-based (tokens, actions) to workflow-based (per completed process) to outcome-based (per result) to per-agent (replacing an FTE), per Sequoia and Paid. Concrete 2026 anchors: Intercom Fin at $0.99 per resolution, Gemini Advanced at $20 per month, GitHub Copilot moving to AI Credits on June 1, 2026.

Should I use seat-based or usage-based pricing? For human-end-user tools with stable consumption (collaboration apps, design tools), seat-based still works. For any product where compute cost scales with use (anything with LLM inference, transactions, API calls), ship a hybrid base-fee-plus-meter. a16z's rule of thumb is that usage-based works best when the end user is software, subscription when the end user is human.

What is outcome-based pricing in SaaS? Outcome-based pricing charges per result delivered (deal closed, ticket resolved, lead enriched), not per seat or per usage unit. Intercom's Fin at $0.99 per resolution is the canonical 2026 example. The model captures more of the buyer's work budget (labor spend) instead of the smaller tool budget (software spend), which is why Sequoia frames outcome pricing as the path to the next $1T company.

Related on the hub

• The Series A bar in H1 2026: what it takes to close — for when the playbook turns into a raise.
• How to price SaaS at seed 2026: the founder framework — Related pricing guide.
• H1 2026 B2B SaaS GTM Benchmark Report — Related gtm business model guide.
• The H1 2026 AI Product GTM Report: data, pricing, and retention — Related gtm business model guide.