REPORT: Pricing Strategy Trends for 2025

Written By Simon Moolman

Why Subscriptions, Credits, and Hybrids Are Converging (Especially for AI Products)

Summary

AI has changed how software costs work. In 2025, many companies mix subscriptions (steady access and service) with credits/usage (pay for what you use). That hybrid approach is growing because AI work—like generating images, video, or long answers—costs more some months than others.

You can see the pattern in the market: Adobe Firefly’s generative credits, Canva’s AI credits per seat, Midjourney’s Fast/Relax capacity tiers, and credits-heavy AI tools such as NanoBanana.ai.

Bottom line: If your product includes AI features, start with hybrid. Give customers a monthly credit bundle sized for typical use, show clear meters and alerts, price overage fairly, and design an ethical, easy-to-understand experience (no fake urgency; easy cancel). See the Appendix for source links.

The Landscape: Why Pricing Is Shifting Now

AI makes costs swing up and down

Every AI task has a small cost—writing text, generating images, or processing video/audio. Heavy months cost more than quiet months. Flat, “all-you-can-use” pricing can get risky unless you set fair limits.

Companies are moving to hybrid

Across the industry, more teams combine a subscription (for access and support) with credits (for the heavier, variable AI work). It’s a practical way to keep revenue predictable while staying fair to light and heavy users.

What leaders are doing

  • Adobe Firefly: Uses generative credits that renew each billing month.

  • Canva: Includes AI credits per seat on paid plans.

  • Midjourney: Bundles Fast (priority) time and a Relax queue when capacity is tight.

  • NanoBanana.ai: Maps image credits to images, with monthly or annual bundles and simple top-ups.

Models Explained—And When to Use Them

Subscription (per seat/plan)

  • Strengths: Predictable revenue, easy buying, helps users build a habit.

  • Risks: A few heavy users can drive up AI costs and squeeze margins.

  • Use when: Usage is steady and AI costs are low or capped.

Credits / Pay‑as‑you‑go

  • Strengths: You earn more when people use more; feels fair to many buyers.

  • Risks: Revenue can jump around; some customers dislike “metered” feelings.

  • Use when: Workloads are spiky, compute-intensive, or enterprise teams need tight cost control.

Hybrid (subscription + included credits + transparent overage)

  • Strengths: Balances predictability and fairness. The subscription covers access and support; credits cover heavier AI use.

  • Risks: A bit more complex to explain and design.

  • Use when: AI features can get expensive, or your users range from casual to power-user.

Psychology: Why These Models Work
(When Designed Ethically)

  • Less “payment pain.” Subscriptions and prepaid credits separate the moment of paying from the moment of using, which makes it feel smoother to use the product.

  • Payment fades. After you pay, the “sting” fades and people are more likely to use what they’ve already paid for.

  • Defaults drive behavior. Auto-renew and simple defaults keep people on a plan—so make cancellation just as easy and transparent.

  • Progress nudges usage. Starter credits and a visible progress bar help people get started and keep going.

  1. Guardrail: Use these effects to make value clearer, not to trick people (see §6).

Costs, in Simple Terms (How to design plans)

What drives COGS (cost of goods sold) for an AI product

Think of every user action (generate an image, summarize text, etc.) as a small “order” your system has to fulfill. The cost to fulfill that order comes from two big buckets:

1) AI model usage (the metered bits)

  • Text (“tokens”): Models charge for how much text you send in and get back.

    • Input tokens = words you send in.

    • Output tokens = words the model writes back.

    • Different models have different prices, and output usually costs more than input.

  • Images / audio / video: Visual or audio features are billed per image, per second/minute, or per frame—separately from text.

  • Bigger/fancier models cost more: A lightweight model is cheap and fast; a reasoning or high-quality image model costs more per request.

Plain English: the more (and “heavier”) AI you use, the more it costs you—like paying per dish in a restaurant, with premium dishes priced higher.

2) The supporting infrastructure (the “kitchen” around the model)

  • Storage: Keeping users’ files, images, and histories.

  • Search & indexing (vector DB): A special index that lets the AI find similar things quickly.

  • Retrieval: Pulling the right docs or images into the prompt before the model answers.

  • Safety/moderation: Automated checks to block unsafe or disallowed content.

  • Observability (logs/monitoring): Tracking performance, errors, and usage so you can fix issues.

  • Bandwidth (“egress”): The cost for data leaving your servers when users download results.

  • Support & payment processing: Human support time and payment fees (e.g., card fees).

Plain English: even when a single request is cheap, running the kitchen—storing things, serving downloads, keeping it safe and monitored—adds real cost.

A simple way to think about it

Per request cost = AI work (text/images/minutes) + a small share of storage/safety/bandwidth + support/payment fees.

Because a small group of power users can use a lot more, hybrid plans (with included credits and fair overage) protect margins without punishing casual users.

Practical Patterns (Borrow What Works)

How leaders package usage

  • Adobe: Generative credits across apps; renew monthly on billing date.

  • Canva: Per-seat AI credits, explained on pricing/help pages.

  • Midjourney: Fast vs Relax modes reflect capacity; plans auto-renew.

  • NanoBanana.ai: Image credits mapped to images; monthly/annual bundles; easy top-ups.

A five‑step mini‑playbook

  1. Pick your unit(s): images, minutes, text “amount,” or “actions.”

  2. Set the monthly bundle: aim for what your typical (70–80th percentile) user needs.

  3. Publish overage, clearly: simple, fair, and visible pricing for extra use.

  4. Build trust UX: live usage meters, alerts at 50/80/100%, budget caps, and (optional) rollover.

  5. Explain in plain English: Add a “How credits work” box with examples.

Compliance & UX: 2025 Guardrails That Matter

EU

  • No “dark patterns.” Don’t hide cancel or nudge people into choices they didn’t mean to make.

  • No fake urgency. If you show “only 3 left” or a countdown, it must be true and provable.

  • AI Act timeline: Some duties already apply; more arrive through 2026–27.

UK

  • Stronger consumer rules. The new law boosts enforcement and focuses on subscription fairness and harmful choice architecture.

USA

  • The federal “click-to-cancel” rule was blocked in July 2025, but states like California still require simple online cancellation if customers signed up online.

South Africa

  • Consumer contracts: Renewal notices and a right to cancel (usually with 20 business days’ notice) for consumer deals; many B2B contracts are outside this rule.


    Quick checklist:     Clear “How credits work,” no fake urgency, easy online cancel, pre-renewal reminders, transparent overage, clear receipts, and an easy-to-read usage dashboard.

Implementation Guide (Operator’s POV)

Packaging & pricing

  • Use Good/Better/Best plans with rising credit bundles. For teams, allow shared (pooled) credits.

  • Choose bundle sizes based on typical usage; consider a limited rollover bank to reduce “wasted credit” anxiety.

  • Keep overage math simple and visible: e.g., $X per extra 100 images.

  • Route work to cheaper models by default; switch to premium only when needed (tell customers what’s happening).

UX & comms

  • Show live usage and budget caps.

  • Send alerts at 50/80/100% used.

  • Add a short, plain “How credits work” explainer.

  • Make cancellation as easy as sign-up; document why your UX choices are fair.

Metrics to watch

  • Money: Gross margin by plan; how much revenue comes from overage; cost per user at different usage levels.

  • Adoption: Activation rate; time-to-first-value; % of users hitting their cap; rollover usage.

  • Trust: Customer satisfaction (NPS), how many clicks to cancel, and pricing complaints.

Risks & Mitigations

1) Bill shock
Why it happens: People exceed their included credits without realizing it, then get an unexpected bill.
How to mitigate:

  • Show a live usage meter and send alerts at 50%, 80%, and 100%.

  • Offer budget caps and a soft stop (short grace window) before overage kicks in.

  • Keep overage pricing simple (e.g., “$X per extra 100 images/tokens”) and visible on the pricing page.

2) Confusion about credits
Why it happens: Credit units feel abstract; customers can’t tell what a “credit” buys.
How to mitigate:

  • Map credits to plain units (e.g., “1 image = 2 credits”) and include 2–3 concrete examples.

  • Add a short “How credits work” panel on the pricing page and inside the product.

  • Use tooltips and on-hover math (e.g., “You have 120 credits ≈ 60 images”).

3) Cross-subsidy & abuse by heavy users
Why it happens: A small group consumes a large share of compute, squeezing margins for everyone else.
How to mitigate:

  • Include a monthly credit bundle sized to typical usage (P70–P80) and price overage fairly.

  • Add fair-use limits, rate limits, or queued “relaxed” mode at extreme usage.

  • Consider role-based tiers or team pools so power users fund their own extra use.

4) Regulatory exposure (UX & claims)
Why it happens: Dark-pattern-like flows (e.g., hard-to-find cancel), fake scarcity, or unclear terms.
How to mitigate:

  • Make cancellation as easy as sign-up; send pre-renewal reminders; show renewal dates on receipts.

  • Avoid false urgency/countdowns; if you show scarcity, ensure it’s real and verifiable.

  • Publish clear pricing, overage rates, and a plain-English “Pricing & Cancellation” policy; keep screenshots of flows for your compliance log.What to Ship This Quarter

  • Money: Gross margin by plan; how much revenue comes from overage; cost per user at different usage levels.

  • Adoption: Activation rate; time-to-first-value; % of users hitting their cap; rollover usage.

  • Trust: Customer satisfaction (NPS), how many clicks to cancel, and pricing complaints.

One‑Page Checklist

Goal: Launch or improve a hybrid pricing model that’s fair, margin-safe, and compliant.

Model & Math

❏ Choose unit(s): images / minutes / text amount / actions

❏ Size included bundle to P70–P80 monthly usage

❏ Set clear overage and list it publicly

❏ Default to cost-efficient models; auto-escalate only when needed

UX & Communication

❏ Live meter + 50/80/100% alerts

❏ Budget caps + team pools

❏ Plain-English “How credits work” box

❏ Optional limited rollover to reduce waste anxiety

Compliance

❏ No fake scarcity/urgency; easy online cancel

❏ Pre-renewal reminders (trials, intro offers, annual)

❏ Receipts show plan, renewal date, and how to cancel

❏ Regional review: EU (DSA/UCPD/AI Act), UK (DMCC/CMA), US (state ARLs), SA (CPA s14)

Metrics

❏ Activation rate; time-to-first-value

❏ % hitting caps; overage share

❏ Cost/user percentile spread; gross margin by plan

❏ Customer satisfaction (NPS); cancel-time clicks

Appendix — Source Links

Product & pricing examples

AI model costs

Market trends

EU/UK/US/SA law & guidance

Simon Moolman
Previous

REPORT: Digital engagement using Conversational UX
Next

REPORT: Mobile Web - Real Issues, Pain Points & Fixes You Can’t Ignore