The Jevons Cliff in Outcome Pricing

I’ve been looking for this framework in the public conversation about AI pricing for a year. It doesn’t exist. The strategic decision around what to do as inference costs fall under your priced outcomes is one of the most consequential CPO/CFO decisions of the next 24 months, and there’s no public playbook.

This essay is the framework. The companion model is in /toolkit/margin-recovery-curve-model.

Inference costs are dropping ~50% per year (industry-wide). Your outcome pricing was set at one moment in time. As costs fall, the gap between your price and your cost grows. You face a decision: hold the price (capture margin), pass the savings (capture share), or split the difference.

This decision isn’t gradual. It comes as a cliff: a competitor undercuts you 12–18 months in, and you have to react. The companies that planned for the cliff handle it cleanly. The companies that didn’t lose share or margin instantly.

The framework: model three scenarios over 24 months, track competitor pricing monthly, and decide which scenario you’re running deliberately.

The math.

Token costs dropped from ~$10/M tokens (2024) to ~$1/M tokens (2026). Industry trajectory suggests ~$0.50/M tokens by 2027.

If your outcome price is $0.50/resolution and inference costs are 25% of revenue today, in 24 months inference costs will be 5–8% of revenue. GM rises from 60% to 80%+ if you hold price.

That’s a 20-point margin lift. Sounds great. Until the competitor with 5% inference cost prices at $0.30/resolution and offers free trial conversion. Now you’re sitting at $0.50, your customers are looking at $0.30, and the cliff hits.

Three sections covering each scenario in detail with the math, the risks, and when each is the right call:

Scenario A: Hold prices

When it works: differentiated product, sticky customers, low competitor threat, high switching costs. Examples: enterprise products with long contracts.

The math: revenue stable, GM rises 20+ points over 24 months. Margin dollars per customer up significantly.

The risk: a 12-month-out competitor undercut that you can’t respond to without admitting your old price was inflated.

Scenario B: Pass full savings

When it works: high-volume products, price-sensitive customers, competitive market, growth-stage company.

The math: revenue per outcome drops in proportion to cost. GM stays at trough level. Volume grows because customers can afford more.

The risk: volume doesn’t compensate for the price drop. Revenue grows slower than expected.

Scenario C: Pass 50% of savings

The pragmatic middle. Most common choice. Captures half the margin, halves the competitive exposure.

Four moves:

1. Build the 24-month model now (use the toolkit linked above).
2. Track three competitor prices monthly. Set a threshold (e.g., “if any direct competitor prices >25% below us”) that triggers a price review.
3. Decide which scenario you’re running. Write it down. Get CFO sign-off.
4. Communicate the decision internally. Sales needs to know whether to defend price or follow.

Three things:

• Build the cliff model on day one of pricing strategy, not month 12.
• Set explicit price-review triggers. Don’t wait for a customer to churn to wake up to the issue.
• Communicate the strategy to sales early. Sales reps making one-off discounts in the field is the number one way the cliff strategy gets undermined operationally.
• The Pricing Migration Sequence, the strategic context for pricing in transition.
• Margin Watch Agent, the agent that operationalizes Jevons-cliff forecasting.
• Pricing Migration: 18-Month Playbook, the pillar chapter.

Originally published at https://falkster.com on May 8, 2026.