AI's Pricing Journey
Since the end of 2022, when Chat GPT-3 was released, many businesses have adopted AI, and new AI-native startups have emerged. This period has been exciting for many reasons. As a pricing enthusiast, I found it fascinating how the industry underwent multiple rapid pricing iterations to find the optimal monetization model that balances value, growth potential, and cost. While this journey is far from over, some common patterns have emerged. Let's explore what AI companies have tried and how these models have worked for them.
Evolution of AI Pricing
Over the past two years, AI pricing has evolved from Pay-As-You-Go Pricing to Credit-Based Pricing, eventually moving toward Progressive Billing, Success-Based Pricing and beyond.
Pay-As-You-Go Pricing
This is the purest form of usage-based pricing, mirroring how we pay for essentials like gas, water, and electricity. It's no surprise that Chat GPT-3 was launched with this model, and many AI companies building foundational models followed suit. Jensen Huang, Nvidia's CEO, even envisioned a future where, alongside electricity and water, AI tokens become a standard, too.
What does this pricing model solve?
- Enables organic account growth
- Covers vendor costs (if bills are paid)
- Works well with Product-Led Growth (PLG) and self-service models
Why is this pricing model challenging?
- No upper limit on cost liability for customers
- High vendor upfront costs since usage is billed in arrears
- Lack of long-term customer commitment
- Revenue is affected by seasonality
- Customers struggle to predict their bills
Pre-Purchased Commitments
This vendor-friendly alternative to Pay-As-You-Go pricing works well for product-led sales (PLS) companies. Customers commit to a specific spend, typically a year in advance, and receive discounts in return. The committed amount is usually flexible and applicable to any feature, which benefits AI companies offering multiple models. While vendors gain predictable revenue and cover their upfront costs, customers may still need help managing their budgets due to potential overages. Large PLS enterprises like Confluent and DataDog have successfully adopted this model.
What does this pricing model solve?
- Covers vendor upfront costs
- Creates a more predictable revenue stream
- Supports organic account growth for customers
Why is this pricing model challenging?
- Customers still face unlimited cost liability
- Sales teams struggle to forecast usage accurately
- Customers find it hard to manage and control spending
Credit-Based Pricing
Credit-based pricing is the self-service version of pre-purchased commitments. Instead of negotiating terms with sales teams, customers can buy credits on-demand, top-up balances as needed, or opt for automatic top-ups. These credits are often flexible and redeemable across various features. This model is perceived as customer-friendly since it empowers users to control costs. However, in practice, customers reliant on mission-critical services may have little choice but to buy more credits when balances run low.
What does this pricing model solve?
- Covers vendor upfront costs
- Facilitates organic account growth
- Aligns with PLG and self-service approaches
- Establishes hard spending limits for customers
Why is this pricing model challenging?
- Running out of credits can cause outages
- Customers often struggle to forecast usage accurately
Recurring Prices With Usage-Based Components
Usage-based and credit-based pricing are often combined with seat-based or recurring prices to ensure baseline revenue. For example, monthly credit allowances with top-ups are common, as seen with Clay. However, this model can be challenging for customers with seasonal demand, as unused credits from slower months cannot roll over.
Another variant ties the number of seats to a shared usage or credit pool. While these hybrid models provide predictable revenue for vendors and their complexity can confuse potential customers.
What does this pricing model solve?
- Guarantees a baseline revenue stream for vendors
- Works well with PLG and self-service models
Why is this pricing model challenging?
- Seasonal customers may end up with unused credits
- Running out of credits mid-cycle can cause service disruptions
Progressive Billing
Progressive Billing represents a newer, vendor-favorable take on Pay-As-You-Go pricing. Vendors set invoicing thresholds, and customers are billed as soon as these limits are reached. Access may be restricted or blocked if invoices aren't paid promptly, minimizing vendor risk and upfront costs. Billing thresholds can also be adjusted based on the customer's account size and payment history.
What does this pricing model solve?
- Encourages organic account growth
- Keeps vendor upfront costs low
- Works effectively with both PLG and PLS approaches
Why is this pricing model challenging?
- Limited customer commitment
- Seasonality and trends impact revenue
- Customers face difficulty predicting their costs
Success-Based Pricing
Most AI companies charge for work, not outcomes. Success-based pricing flips this approach, aligning costs with measurable customer value. This model is gaining traction in applications like AI-powered customer support, where fees are tied to successfully resolved tickets. For example, Intercom's AI chatbot defines a successful interaction as one where the customer's issue is resolved or they don't return within 24 hours. While this model is highly aligned with customer goals, defining and measuring success requires mutual agreement between the vendor and the customer. To learn more about success-based pricing, read this excellent article from Kyle Poyar.
What does this pricing model solve?
- Aligns price with delivered value
- Supports organic account growth
- Works well with PLG and self-service models
- Simplifies cost forecasting by tying it to business outcomes
Why is this pricing model challenging?
- Defining and measuring success can be complex and subjective
What's Next?
Current pricing models adopted by AI companies tend to favor vendors, leaving customers with limited control over spending. As customers seek stability and value-aligned pricing, we can expect more innovative monetization models to emerge. Billing and tooling will also improve, offering real-time usage forecasting and management.
It will be interesting to see whether success-based pricing will extend to foundational models and infrastructure applications or remain limited to specific AI use cases. In the meantime, engineering teams tasked with adapting billing systems to evolving pricing models will play a crucial role in maintaining their companies' competitiveness.
When engineering cannot keep up with changing requirements, businesses lose the ability to iterate on pricing quickly. Developers who solve these challenges become the unsung heroes driving business success.
Make the move today
OpenMeter provides flexible Billing and Metering for AI and DevTool companies. It also includes real-time insights and usage limit enforcement.
We help companies with Product-Led Monetization: