AI Credits Are Eating SaaS Pricing: Why 62% of Companies Will Use Token-Based Billing by 2027

AI Credits Are Eating SaaS Pricing: Why 62% of Companies Will Use Token-Based Billing by 2027

TL;DR: 62% of B2B software companies will use token-based billing by 2027. AI credit adoption grew 126% YoY. Hybrid pricing (seats + consumption) is now the #1 model at 37%. Median AI margins are 50% vs 70-80% for SaaS. This piece covers the data, the big-mover pricing shifts, and three startups built for the credits era.

AI credit adoption grew 126% year-over-year in 2025. Another 33% of software companies plan to introduce credits within the next 6-12 months. Combined, that puts nearly two-thirds of B2B software on token-based billing by late 2027.

This is not a trend you can afford to watch from the sidelines. Whether you are a founder pricing your first product or a CFO trying to forecast next quarter's AI spend, the shift from seats to credits changes the math for everyone.

I spent time with the Growth Unhinged 2026 State of B2B Monetization report (230 companies surveyed, from $1M to $150M+ ARR) and cross-referenced it with real pricing moves from Microsoft, HubSpot, Anthropic, and SAP. What follows is what the data actually says, what it means for founders, and which startup ideas are positioned to win.

What Do 230 Companies Reveal About AI Credit Adoption?

Let us start with the hard data, because the headlines understate how fast this is moving.

29% of B2B software companies already use AI credits. That is up from roughly 13% a year ago, which is where the 126% growth figure comes from. Another 33% plan to introduce credits within 6-12 months. Add them together and you get 62% of companies either on credits or about to be. Among companies with $50M+ ARR, the number climbs to about 50% planning credit launches this year.

Hybrid pricing is now the most popular model. 37% of companies combine two or more pricing models, typically a per-seat subscription with AI consumption layered on top. That is up from 25% just 12 months earlier. Pure per-seat pricing dropped from 21% to 15% of companies in the same period.

Median target AI gross margin sits around 50%, compared with 70-80%+ for traditional SaaS. Only 12% of AI companies aim for SaaS-like margins of 80%+. A similar number target margins of 20% or below. This margin gap is the structural reason credits exist. When your cost to serve scales with usage, flat per-seat pricing either eats your margins or caps your heaviest users.

75% of companies changed pricing or packaging in the last year. The largest companies are making the most changes. It is nearly impossible to find a $50M+ ARR company not rethinking pricing for AI.

Satya Nadella framed it plainly: seats are now "just entitlement to some consumption."

How Are Microsoft, HubSpot, Anthropic, and SAP Changing Their Pricing?

The shift from seats to credits is not theoretical. The biggest software companies on the planet are making real pricing changes right now.

Microsoft is moving all GitHub Copilot subscribers to token-based billing starting June 1, 2026. Copilot Business customers will pay $19/user/month and receive $30 in pooled AI credits during a promotional period (June through August). After that, credits drop to match the subscription cost. Copilot Enterprise customers move to $39/user/month with $70 in pooled credits. Microsoft also suspended new sign-ups for individual and student accounts and pulled Anthropic's Opus models from the $10/month plan. These are not pilot programs. This is the direction.

HubSpot introduced outcome-based pricing for its Breeze AI agents in April 2026. Instead of charging per seat for AI features, HubSpot charges per outcome — per conversation resolved, per lead qualified. This is one of the first major SaaS companies to price AI value rather than AI access.

Anthropic lowered Enterprise seat prices and shifted aggressively to usage-based pricing. The move signals that even model providers see consumption billing as the future, not seat licenses.

SAP announced a shift toward AI consumption pricing for its enterprise suite. When SAP moves this way, the enterprise software market follows.

Clay introduced dual-track monetization in March, separating platform value from token cost. Customers pay for the product on one track and pay for the AI compute on another. It is the clearest articulation of the hybrid model so far.

The pattern is consistent. Credits and consumption are replacing seats. The only debate left is how quickly and how transparently.

Why Are AI Credits More Than Just a Vendor Cash Grab?

The default reaction to AI credits is suspicion, and for good reason. Credits make budgeting harder. They introduce unpredictability. When a vendor controls what a "credit" is worth and how quickly it gets consumed, the buyer loses the ability to forecast costs.

Kyle Poyar, who runs the Growth Unhinged survey, wrote it directly: "AI credits are great for vendors. They can become a nightmare for customers."

But there is a real structural reason credits exist, and it starts with inference costs.

Traditional SaaS has near-zero marginal cost per user. Add a seat, add a license key, done. AI products do not work that way. Every API call costs money. A complex agent task that runs for 30 minutes can consume more tokens than a hundred simple queries. A power user who leaves their AI assistant running all day generates costs that dwarf a casual user who checks in twice a week.

Per-seat pricing cannot handle this spread. If you price high enough to cover your heaviest user, everyone else subsidizes them. If you price low enough to attract casual users, your power users burn through your margins. Credits solve this by tying cost to consumption.

The problem is that credits become opaque fast. Microsoft calls them "GitHub AI Credits." Anthropic calls them "tokens." HubSpot calls them "outcomes." There is no standard unit, no standard conversion rate, and no way for a buyer to compare what they are getting from one vendor versus another.

This opacity creates a new market opportunity, which I will get to shortly.

Which Three Startup Ideas Are Built for the Credits Era?

Startups that understand the credits shift can build products that solve the problems it creates. Three ideas from the Best AI Startup Ideas database map directly to the gaps credits open up.

1. AttributionEngine AI — The Dual-Model Pricing Playbook

AttributionEngine AI already demonstrates the pricing model that 37% of companies are adopting: hybrid. Its model offers per-conversation or per-resolution billing. Customers choose predictability (a fixed per-conversation cost) or accountability (pay only when the AI delivers a resolved, measurable outcome).

The Growth Unhinged data confirms this is where pricing is heading. 37% hybrid adoption. Companies like Decagon and Salesforce now offer multiple pricing models simultaneously. The 37% figure is not a ceiling. It is the current snapshot of a curve that is still bending upward.

The takeaway for founders: if you are building an AI product and you have not thought about whether your pricing model matches your cost structure, you are probably leaving money on the table or setting yourself up for margin compression. Read How to Price an AI Startup When Inference Costs Are a Moving Target for the full framework.

2. PriceScope AI — The CPI Index for AI Credits

PriceScope AI started as a crowdsourced pricing intelligence platform for SaaS contracts. The credits era makes it more relevant, not less.

When every vendor prices in proprietary credits with no standard conversion rate, buyers need benchmarking data. How much should 1,000 GitHub AI Credits cost relative to 1,000 Anthropic tokens? What is the average per-seat-plus-credits pricing for a marketing automation platform? PriceScope becomes the CPI index for AI credits — the place where procurement teams go to find out what a credit actually costs across vendors.

This is not a hypothetical. Companies overpay by 20-40% on SaaS renewals because they lack data. Credits make the opacity worse, not better. A crowdsourced, verified pricing database is the only structure that can create transparency in a market where vendors have every incentive to keep pricing obscure. See Business Model Debt: Why SaaS Companies Are Dying From the Inside in 2026 for why opaque pricing is a symptom of deeper pricing model problems.

3. CFO Narrator AI — Automating the Cost Narrative

CFO Narrator AI automates financial narrative reporting. In a credits world, this product solves a specific pain point.

AI credit spend is volatile. A team that spends $5,000 on credits one month can easily spend $12,000 the next. The CFO has to explain that variance to the board. "We spent $7,000 more on AI credits" is not a narrative. It is a number that raises questions without answering them.

CFO Narrator turns volatile credit consumption data into coherent cost-justification narratives. It connects the spike in token usage to specific projects, teams, or outcomes. It gives finance teams the language to explain AI spending to boards that still think in terms of per-seat licenses.

This matters now more than ever. 70% of respondents in the Growth Unhinged survey said AI spend comes out of existing technology budgets. That means credit cost spikes are competing with SaaS renewals, infrastructure costs, and headcount for the same pool of money. If the finance team cannot justify the spend, it gets cut, along with the AI tools that depend on it.

Why Is the AI Margin Gap Only 50% vs 70-80% for Traditional SaaS?

The median target AI gross margin in the Growth Unhinged survey is roughly 50%. Traditional SaaS targets 70-80%+.

This is not a rounding error. It is a structural difference that changes how you build, price, and fund an AI company.

Only 12% of AI companies aim for SaaS-like margins. The rest have accepted that inference costs create a permanent drag. Credits are the pricing mechanism that lets them close the gap. By charging heavy users more and casual users less, credits let companies maintain a 50% margin at the aggregate level while still offering an attractive entry point for new customers.

But credits also create a new risk. If a customer's credit consumption spikes unexpectedly, they leave. The Growth Unhinged data shows that usage-based and outcome-based models struggle with revenue forecasting. Hybrid models are "generally the happiest" but still hard to explain to customers.

The founders who figure out transparent, predictable credit pricing — where buyers understand what they are paying for and can forecast it — will have a real competitive advantage. Not because credits are better for buyers. They are not always. But because credits are what the market is migrating toward, and transparency is the feature that makes them tolerable.

What Does the Credits Shift Mean for Founders Right Now?

Three things are clear from the data.

First, if you are building an AI product, price it based on consumption from day one. Not per seat. Not flat fee. Consumption, with a seat-based floor if you need predictable revenue. The Growth Unhinged survey shows 37% of companies already use hybrid models. Among companies with over $50M ARR, nearly half plan to introduce credits this year. The direction is set. Pricing your product per seat in 2026 is pricing it for 2023.

Second, build for the transparency gap. Every vendor is moving to credits. None of them want to explain what a credit costs. This is an opening for startups that help buyers understand, compare, and forecast AI spending. PriceScope AI fills this gap on the procurement side. CFO Narrator fills it on the reporting side. Both are bets that the opacity of credits will get worse before it gets better.

Third, the margin gap between AI and SaaS is real and permanent. 50% gross margins are the new normal for AI products. If your business model requires 70%+ margins to work, you need to build a data moat or a workflow moat that creates pricing power independent of compute costs. AI Unit Economics: Why Your Startup Burns Cash Despite Growing Revenue covers this in detail.

---

Explore the startup ideas mentioned in this article:

• AttributionEngine AI — Marketing Attribution Analyst — The hybrid pricing playbook in action
• PriceScope AI — Vendor Pricing Intelligence Tracker — The CPI index for AI credits
• CFO Narrator AI — Financial Report Automation — Turning volatile credit spend into board-ready narratives

Read more:

• How to Price an AI Startup When Inference Costs Are a Moving Target
• Business Model Debt: Why SaaS Companies Are Dying From the Inside in 2026
• AI Unit Economics: Why Your Startup Burns Cash Despite Growing Revenue

---

FAQ

What are AI credits in SaaS pricing? AI credits are a consumption-based billing unit where customers purchase a pool of tokens or credits that are consumed based on actual AI usage. Credits replace or supplement per-seat pricing by tying cost to the volume of AI compute, conversations, or outcomes a customer uses.

Why are SaaS companies switching to token-based billing? Inference costs scale with usage, not seats. A power user who runs complex AI tasks can cost 10-100x more to serve than a casual user. Per-seat pricing cannot capture this spread without either overcharging light users or undercharging heavy ones. Credits align revenue with cost.

How fast is AI credit adoption growing? 126% year-over-year in 2025, from roughly 13% to 29% of B2B software companies. Another 33% plan to introduce credits within 6-12 months, putting nearly 62% of companies on credit-based billing by late 2027 according to the Growth Unhinged 2026 survey of 230 companies.

What is the median AI gross margin? Approximately 50%, compared to 70-80%+ for traditional SaaS. Only 12% of AI companies target SaaS-like margins of 80%+. This margin gap is the structural driver behind the shift to consumption-based pricing.

Is hybrid pricing the best model for AI startups? The data suggests it is the most popular option, with 37% adoption and growing. Hybrid models (typically a seat-based subscription plus consumption credits) give companies the predictability of recurring revenue while covering variable inference costs. The key challenge is explaining the model to customers.

What Is the Bottom Line for SaaS Pricing?

The shift from seats to credits is the single biggest pricing change in SaaS since the move from perpetual licenses to subscriptions. The data shows it is accelerating, not slowing down. The biggest companies are leading it. The median AI margin of 50% makes it structurally necessary.

For founders, the question is not whether to adopt credits. It is how to adopt them transparently enough that customers do not flee. The winners in this transition will be the companies that make credit-based pricing understandable, forecastable, and fair. The losers will be the ones that use credits to hide margin compression behind opaque pricing units.

Build for transparency. Price for consumption. Expect 50% margins.

---

Rethinking your pricing model? The shift from per-seat to credit-based billing isn't optional — it's structural. Check out ChurnShield for a retention strategy that works with consumption pricing, or read why vertical AI attacks labor budgets for how to price against the right cost line.