Pricing model debates in SaaS used to be mostly philosophical — per-seat feels fair, usage-based feels aligned with value. In 2026 that debate has largely been settled by AI infrastructure costs, not philosophy: when every inference call has a real marginal cost, flat per-seat pricing stops making financial sense the moment usage varies 10x between customers.
The adoption numbers
Usage-based pricing adoption in SaaS climbed from 27% in 2023 to 38% in 2026 (ValueAdd VC). A January 2025 survey of 100 SaaS companies by Metronome and Greyhound found 85% had adopted usage-based pricing in some form (Orb) — the gap between these figures reflects "any usage component at all" versus "primarily usage-based," which matters for interpreting the trend correctly: usage-based pricing as a pure model is still a minority approach, but usage-based pricing as a component is now closer to the norm.
The size effect is sharp: 77% of the largest software companies have some level of usage-based pricing, and 40% of companies with ARR above $50M include consumption- or outcome-based revenue in their ARR, versus only 20–27% in smaller ARR bands (Orb). Usage-based pricing correlates with company maturity — it requires metering infrastructure and billing sophistication that smaller companies often haven't built yet.
Seat-based pricing is actively shrinking
The more striking number is the decline of pure seat-based pricing: it fell from 21% to 15% of SaaS companies in just twelve months, while hybrid models surged from 27% to 41% over the same period (ValueAdd VC). Hybrid pricing (base fee plus metered usage or outcome component) now covers 43% of SaaS companies and is projected to reach 61% by end of 2026 (ValueAdd VC).
| Model | 12 months ago | Now (2026) | Projected end-2026 |
|---|---|---|---|
| Pure seat-based | 21% | 15% | — |
| Hybrid (base + usage) | 27% | 41–43% | ~61% |
| Usage-based (any form) | — | 38–85%* | — |
*Range reflects "primarily usage-based" (38%) vs. "any usage component present" (85%). Sources: ValueAdd VC, Orb
Hybrid isn't a compromise position gaining ground reluctantly — it's the model that's actively winning against both pure alternatives.
Why AI cost structure forces the shift
The underlying driver is structural, not stylistic. Traditional SaaS has near-zero marginal cost per user — an extra login doesn't cost the vendor meaningfully more. AI-native SaaS breaks that assumption: every inference call costs money, token generation has a per-unit price, and heavy users can cost 10x what light users cost the vendor to serve (Fungies).
Flat per-seat pricing under this cost structure means either underpricing power users (losing money on your best customers) or overpricing light users (pushing away exactly the self-serve signups a PLG motion depends on). Hybrid pricing — a base subscription covering access and baseline usage, with metered overages for heavy consumption — resolves this by aligning price with the actual cost driver while still giving procurement teams a predictable base invoice (Fungies).
Note
The Cursor invoice: what happens when the architecture has no guardrail
In July 2025, a single Cursor developer generated a $7,225 invoice in one day after a team member exhausted 500 requests under an annually-billed plan (ValueAdd VC). This became a widely cited case study in AI SaaS pricing risk — not because the billing itself was technically incorrect, but because the pricing architecture allowed a single user's behavior to generate a five-figure invoice with no warning system intervening before it happened.
The lesson isn't "don't do usage-based pricing" — it's that usage-based and hybrid models need active spend controls as a first-class product feature, not an afterthought:
- Real-time usage dashboards visible to the customer, not just post-hoc invoices
- Configurable spend caps or alerts before a threshold is crossed
- Rate limiting or approval gates for anomalous single-session usage spikes
A pricing model that's directionally correct (align cost to value) can still produce a customer-relationship-ending incident if the guardrails around it are missing.
Revenue and growth outcomes
Companies primarily using consumption-based pricing models grew revenue roughly 8 percentage points faster on average than companies that don't (Orb). Concrete examples: Snowflake at 34% revenue growth with 126% net revenue retention, and Datadog at 32% growth — both well ahead of the roughly 14% median growth rate for public SaaS companies overall (Orb).
This correlation likely runs both directions — usage-based pricing enables faster expansion revenue capture as customers grow their usage organically, and companies with strong product-market fit (who grow faster generally) are also more likely to have the metering sophistication to run usage-based pricing well. It's not purely causal, but the pattern is consistent enough across companies to take seriously.
The predictability tradeoff, and how vendors are resolving it
The honest downside of usage-based pricing is real: it aligns revenue with delivered value and lowers adoption friction, but makes revenue less predictable for the vendor's own forecasting, and makes the customer's monthly invoice less predictable for their budgeting (Orb).
The 2026 convergence point is a committed-use floor with usage-based overage: the customer commits to (and is billed for) a baseline amount of usage upfront, and only usage beyond that floor is metered (Orb). This gives the vendor a predictable revenue floor and the customer a predictable base invoice, while still capturing upside from heavy usage — effectively the same structure as the "hybrid" model described above, arrived at from the revenue-predictability angle rather than the cost-structure angle.
Decision framework: when each model fits
Pure seat-based: Low marginal cost per user, usage doesn't vary
much between customers, procurement wants
simple per-head budgeting.
Pure usage-based: High, variable marginal cost per unit of use
(AI inference, storage, compute). Value scales
directly with consumption. Self-serve motion.
Hybrid (base+usage): AI-native product with real marginal costs,
but selling into procurement-driven buyers who
need budget certainty. The 2026 default for
AI SaaS.
Actionable takeaway
- Default to hybrid, not pure usage-based, if you're building an AI-native product selling to businesses — it's the model gaining share fastest (27% → 41% in a year) precisely because it balances vendor cost-alignment with buyer budget predictability.
- Ship spend guardrails alongside usage pricing, not after a Cursor-style incident forces the issue — real-time usage visibility and configurable caps are now expected, not optional.
- Use a committed-use floor if predictable revenue matters to your forecasting — bill a baseline upfront, meter only the overage.
- Don't assume usage-based pricing alone drives growth — the 8-point growth advantage correlates with consumption pricing but is likely compounded by the product-market fit that made consumption pricing viable in the first place.
- Revisit pricing architecture as you scale — the data shows usage-based sophistication rises sharply with company size (20–27% at smaller ARR bands vs. 40%+ above $50M ARR), suggesting it's something to grow into deliberately, not necessarily launch with on day one.
Sources: ValueAdd VC — Usage-Based SaaS Pricing in 2026, ValueAdd VC — AI Product Pricing Strategy in 2026, Orb — 40 SaaS Pricing Statistics, Fungies — AI SaaS Pricing Models in 2026, Zylos Research — Hybrid Pricing Architectures for AI Agent Platforms
Get new posts as they publish
No spam — just the next post, straight to your inbox.