SaaS Pricing Is Dying. The Founders Who See It Coming Are Already Switching Models.
By Creatives Takeover · June 9, 2026
The Future of SaaS Pricing.
The Model That Built an Industry
For most of the last decade, selling software followed a simple and reliable formula.
You built a product. You picked a tier. You charged a flat monthly fee per seat. As a company grew its team, it added seats, your revenue grew with it, and your margins stayed healthy because adding one more user cost you almost nothing to serve. It was clean, predictable, and easy for both sides to understand.
That model built some of the most valuable companies in history. Salesforce, Slack, HubSpot, Zoom. The per-seat subscription became so dominant that it stopped feeling like a business decision and started feeling like a law of nature. If you were building SaaS, this was simply how you priced it.
The law is breaking. And the data makes it increasingly hard to look away.
What the Numbers Are Already Saying
Bloomberg is forecasting that subscription-based pricing will drop from 60 percent to 30 percent of SaaS models over the next decade. IDC predicts that by 2028, pure seat-based pricing will be obsolete, with 70 percent of software vendors refactoring their pricing strategies around entirely new value metrics. Gartner projects that at least 40 percent of enterprise SaaS spend will shift to usage, agent, or outcome-based models by 2030, with seat-based revenue share declining from 21 percent to 15 percent.
These are not fringe predictions. They are mainstream forecasts from the firms that enterprise CFOs and CIOs pay attention to.
And the market is already moving. According to Chargebee's 2025 State of Subscriptions Report, 43 percent of companies now combine subscriptions with usage-based components, with adoption projected to hit 61 percent by the end of 2026. Credit-based model adoption surged 126 percent year over year across the top 500 SaaS companies in 2025. Figma, HubSpot, Airtable, and Monday.com all adopted credit-based models last year. GitHub announced in April 2026 that all Copilot plans would transition to token-based billing on June 1st.
This is not a niche shift happening at the edges of the industry. It is a structural renegotiation of how software gets paid for, and it is happening across every category simultaneously.
Why AI Made the Old Model Unsustainable
The per-seat model worked because adding one more user to a traditional SaaS product cost essentially nothing. A CRM, a project management tool, a design platform. The marginal cost of serving another seat was close to zero, which is exactly why the economics made sense for so long.
AI broke that logic in two ways.
First, the cost side. An AI feature that calls a large language model has real per-request compute costs that vary depending on the complexity of the input. The underlying infrastructure is tied to consumption, not headcount. You cannot sustainably price by seat when your costs are being driven by usage. The two things are misaligned, and that misalignment compounds the more AI you ship into your product.
Second, the value side. When one person with an AI agent can do the work that previously required five, headcount stops being a useful proxy for how much value a customer is extracting from your product. A company that replaces five seats with one AI-assisted operator is not a shrinking customer. They might actually be getting more done than ever. But under the per-seat model, your revenue just dropped by 80 percent.
That is the core tension. Enterprise buyers have started noticing. Investors are already pricing in the risk. Analysts have a name for it now: AI seat risk.
The Companies That Figured It Out First
The clearest proof that the new models work is not in the forecasts. It is in the revenue numbers of the companies that built on them early.
Snowflake charges nothing per seat. Its entire pricing structure is built around consumption: storage bills per terabyte, compute bills per second per warehouse, and there is no headcount metric anywhere in the model. The result is that as customers use more, they pay more, without any sales motion required. That built-in expansion mechanism helped Snowflake achieve a 158 percent net revenue retention rate, one of the highest ever recorded in public SaaS.
Datadog runs a variation of the same idea. Its core monitoring product charges per host, log management bills per gigabyte ingested, and application performance monitoring charges per traced request. The result is multiple usage vectors expanding revenue inside existing accounts at the same time. Datadog generated $3.43 billion in revenue in 2025, up 28 percent year over year.
Twilio charges $0.0079 per SMS sent and $0.013 per minute for voice. The model makes intuitive sense because Twilio's own costs are usage-based on the carrier side. By matching how they charge customers to how they are charged internally, they maintained healthy margins while letting developers experiment at low risk and grow into larger spend naturally.
What these three companies share is not just a pricing model. They share a billing metric that maps directly to how customers actually extract value from the product. That alignment is what makes the model durable.
The Cautionary Tale Nobody Talks About Enough
Not every company made this transition well.
Cursor, the AI coding assistant that reached $500 million in ARR faster than OpenAI, became one of the most instructive cautionary tales of 2025. In June of that year, the company replaced its fixed request-cap model with usage-based billing tied directly to API costs. The change made economic sense. Frontier AI models are expensive to run, and flat-rate pricing was becoming unsustainable as power users leaned on the most expensive models for increasingly complex tasks.
The problem was not the decision. It was the execution. Cursor made the change with minimal notice, no in-app usage meters, and no hard spending caps for customers. Developers woke up to invoices they had no way of anticipating. Some reported bills in the hundreds or thousands of dollars for a single month of usage. By July 4th, Cursor issued a public apology and offered refunds for unexpected charges incurred between mid-June and early July.
The reputational damage was real. Tools that had previously been afterthoughts, including Windsurf and Claude Code, began picking up frustrated Cursor users almost immediately. The product remained strong. But the trust that had made Cursor the default recommendation in developer communities took a significant hit.
The lesson is not that usage-based pricing is dangerous. The lesson is that the transition itself is a product launch, and it needs to be treated like one.
The Three Models Taking Over
The industry is not moving toward a single replacement for per-seat. It is moving toward three distinct models, and understanding the differences matters if you are building something right now.
Usage-based pricing charges customers for what they actually consume. API calls, storage, compute hours, messages sent, credits used. The customer pays for what they use, nothing more. The advantage is alignment: customers who extract more value pay more, and customers who barely use the product do not churn purely because of cost. The challenge is revenue predictability, which is harder to manage when billing fluctuates month to month.
Outcome-based pricing goes one step further. Customers pay for results, not usage. Zendesk was the first major SaaS incumbent to launch outcome-based pricing for AI agents in 2024, charging $1.50 per support ticket resolved by AI without human intervention. Intercom's Fin charges $0.99 per resolved customer interaction, and scaled to eight-figure ARR at a 393 percent annualized growth rate on that model. Salesforce Agentforce charges approximately $2 per conversation or lead. Gartner projected that over 30 percent of enterprise SaaS solutions would incorporate outcome-based components by 2025, and that number continues to rise.
The hybrid model is where most companies land in practice. A base platform fee that gives customers the predictability they need to budget, paired with variable usage or outcome components that scale with value. According to Metronome, 61 percent of SaaS companies now use hybrid models. It is the dominant transition state precisely because it asks neither side to fully abandon what they already understand. Most enterprise renewals in 2025 and 2026 are landing here.
How to Actually Iterate on Your Pricing Model
The most common mistake founders make is treating pricing as a launch decision rather than a product discipline. Pricing needs to be iterated, tested, and refined with the same rigor applied to the product itself. Here is how the founders who are getting this right tend to approach it.
They start by identifying the value metric before choosing a pricing structure. The single most useful question is: what is the one unit of value my customer receives from this product? Not what is easy to measure internally. What is meaningful to the buyer. Snowflake landed on compute seconds because that is when customers extract value from their data. Intercom landed on resolved conversations because that is the outcome their customers are paying to achieve. Getting the metric right is more important than getting the model right.
They instrument before they price. Outcome-based pricing requires the ability to reliably track and attribute results. That is an engineering requirement, not an afterthought. The founders who are winning at outcome pricing built the measurement layer first, often before they built the pricing page. If you cannot measure it cleanly, you cannot charge for it credibly.
They sequence the transition carefully. Most Series A companies are better served starting with simpler pricing and moving toward consumption or outcome models once they understand their customers' usage patterns. Retrofitting a per-seat model into a hybrid or outcome structure after the fact is painful, requiring renegotiated contracts, rebuilt billing infrastructure, and managed customer expectations through a change they did not sign up for. But if you are building something new today, the window to choose the right model from the start is right now.
They communicate changes like product launches. The Cursor story is the clearest possible evidence of what happens when pricing changes get treated as operational updates rather than customer moments. Direct email notification, in-app banners, grace periods on old terms, and visible usage dashboards are not nice-to-haves. They are the difference between a pricing evolution and a trust crisis.
They build spend controls into the product from the beginning. Bill shock is a pricing design failure, not a customer education problem. Spend caps, usage dashboards, and alert thresholds are engineering requirements. If your customer cannot see in real time how much of their included amount they have used, you have already made a mistake that will cost you later.
Five Lessons for Founders Building Right Now
The pricing model is a product decision, not a finance decision. How you charge communicates what you believe your product is worth and how customers experience value. Get it wrong and the rest of the funnel suffers regardless of how good the product is.
Start with the value metric, not the pricing structure. Everything follows from answering one question honestly: what is the unit of value my customer actually receives? Seat count is rarely the right answer in 2026.
Hybrid is not a compromise. It is a strategy. A base fee plus variable components gives buyers the predictability they need to approve budgets and gives you the upside that scales with usage. There is a reason it is the dominant model right now.
The transition is harder than the decision. If you are already on a per-seat model and need to shift, plan for a six-month runway, not a six-week one. Over-communicate, grandfather existing customers where you can, and give buyers time to understand the new logic before they see a new invoice.
Pure usage-based is not always the answer. Seats still make sense where access itself is the value, where usage is stable and predictable, and where AI is not meaningfully changing your cost structure. The question is not whether seat pricing is dead. The question is whether it is the right model for what you are specifically building. In 2026, for most AI-native products, it is not.
The per-seat model is not disappearing overnight. But the direction is clear, the data is consistent, and the window to build with the right model from the beginning is narrowing. The founders who understand that earliest will not have to unwind a pricing architecture later when the market forces their hand.
Software has always moved toward the model that most honestly reflects the value it delivers. For thirty years that was access. In 2026 it is outcomes. The founders who price accordingly are not following a trend. They are building for where the trust between software and its buyers is actually heading.