Stress-Testing a 50% Model Price Drop on a $42k-MRR SaaS

Model price drops are usually reported as unambiguous good news. They are not. A drop you can keep is margin expansion. A drop your competitor passes through faster than you do is a forced price cut on your own product. This article runs a $42k MRR scenario through the Model Price Drop Stress Test calculator and shows the realistic outcomes at 10%, 30%, and 50% drops, then maps the pass-through dynamics solo founders should plan around.

1. The $42k MRR setup at 58% margin

The scenario is a solo AI SaaS at $42,000 MRR with $7,800 of monthly AI cost (Claude Sonnet on user prompts, embeddings, plus a small amount of GPT-4o mini for classification). Gross margin today is 58%, slightly below the SaaS median Bessemer's 2024 cloud index reports at 70-75%^[4]. Engine math: gross profit today is $24,360 ($42,000 × 58%). Non-AI cost is $9,840 ($42,000 − $24,360 − $7,800). That non-AI line covers hosting, database, payment processing, support, and the founder's own time priced at a contractor rate. The engine stress-tests three drop magnitudes from one run:

# model-price-drop-stress-test (computed live from /engines/model-price-drop-stress-test.js)
Engine input
  monthly_revenue       = 42000
  monthly_ai_cost       = 7800
  gross_margin_percent_today= 58

Engine output
  todayGrossProfit      = 24360
  todayNonAiCost        = 9840
  scenarios[0].dropPercent= 10
  scenarios[0].newAiCost= 7020
  scenarios[0].newGrossMarginKeepSavings= 59.86
  scenarios[0].newGrossMarginPassThrough= 58
  scenarios[0].newRevenueIfPassThrough= 40142.86
  scenarios[1].dropPercent= 30
  scenarios[1].newAiCost= 5460
  scenarios[1].newGrossMarginKeepSavings= 63.57
  scenarios[1].newGrossMarginPassThrough= 58
  scenarios[1].newRevenueIfPassThrough= 36428.57
  scenarios[2].dropPercent= 50
  scenarios[2].newAiCost= 3900
  scenarios[2].newGrossMarginKeepSavings= 67.29
  scenarios[2].newGrossMarginPassThrough= 58
  scenarios[2].newRevenueIfPassThrough= 32714.29
  mostLikelyMargin      = 67.29

2. The 10% drop: 1.86 points of margin

A 10% model price drop is the realistic near-term scenario — Anthropic, OpenAI, and Google have all cut individual model prices by similar margins over 12-month windows^[1][2]. In the engine block above, the 10% scenario shows new AI cost $7,020 (was $7,800), gross margin 59.86% if kept (was 58%), 58% if fully passed through, and revenue of $40,142.86 if passed through (was $42,000).

Two things to read off this. First, 1.86 points of gross margin from a 10% input-cost cut is small in percentage terms but real in dollars: $780/month, or $9,360/year of additional gross profit. Second, the pass-through case shows the trade — if competitors force a 4.4% price cut to defend share, revenue drops by $1,857/month and the founder ends with the same gross profit ($24,360) at lower scale. Pass-through is rarely all-or-nothing; 30-50% pass-through is the typical mid-case, which lands the founder somewhere between these two extremes.

3. The 30% drop: where competitor pass-through bites

A 30% model price drop is rarer but has happened (OpenAI cut GPT-4 Turbo from $30/$60 to $10/$30 over 18 months, a 67% drop at the input rate^[2]). The 30% scenario in the engine block above returns new AI cost $5,460 (was $7,800), gross margin 63.57% if kept (was 58%), 58% if passed through, and revenue of $36,428.57 if passed through (was $42,000).

The kept-savings case is $2,340/month, or $28,080/year of additional gross profit. The passed-through case shows the founder defending margin at the cost of $5,571 of monthly revenue, or $66,852 of annual ARR. The interesting case is the middle: 50% pass-through means the founder cuts price by 7%, lands at $39,060 MRR and roughly 60.5% gross margin. Gross profit holds up at $23,631, slightly below the no-drop baseline. The founder absorbed half the input savings into price defence.

4. The 50% drop: 67.29% margin and pricing pressure

The 50% drop case is what happens when a vendor ships a major efficiency improvement (Anthropic's Haiku tier launch, OpenAI's GPT-4o mini at $0.15/$0.60). The 50% scenario in the engine block above returns new AI cost $3,900 (was $7,800), gross margin 67.29% if kept (was 58%), 58% if passed through, and revenue of $32,714.29 if passed through (was $42,000). The engine's central estimate (mostLikelyMargin) is 67.29%.

The engine flags 67.29% as the "most likely margin" outcome — the assumption being that competitors will not perfectly pass through a 50% savings. That assumption is contestable. In an undifferentiated market (the AI tool wraps a single model with no proprietary data), competitors with the same input cost reduction will pass through aggressively to capture share. In a differentiated market (the AI tool has proprietary fine-tuning, a unique workflow, or owns a customer relationship), pass-through is slower and the margin gain is more durable.

The honest planning number is the midpoint: roughly 62.5% margin after 90 days, against $36,500 of MRR if half the savings get passed through. That produces $22,800 of monthly gross profit — slightly below the $24,360 baseline. The founder gained nothing from a 50% input cost cut because the market forced the savings into the customer's pocket.

5. The pass-through problem

Pass-through speed depends on three factors:

• Differentiation. The more substitutable the product, the faster pass-through. Pure model-wrapper products see 80%+ pass-through within 60 days. Products with proprietary data, workflows, or integrations see 20-40% pass-through over 6-12 months.
• Customer awareness. If customers track model prices (developer tools, AI infrastructure products), pass-through is fast. If they do not (vertical SaaS where AI is a feature rather than the product), pass-through is slow or nonexistent.
• Contract length. Annual contracts delay pass-through by 12 months on the existing book and immediately on new sales. Monthly subscriptions force the question every renewal cycle.

For most solo AI SaaS, the realistic pass-through assumption is 30-50% within 90 days. The founder keeps 50-70% of any input-cost saving. The Stanford HAI AI Index Report documents the broader trend: token prices fell ~60% across major frontier models from 2023 to 2024^[3], but SaaS pricing has held remarkably steady — meaning indie founders have generally captured the gains.

6. Second-order effects: demand and competitor moves

Two effects the engine does not model directly but matter for the real outcome:

• Elasticity of demand. A 22% price cut (the pass-through case at 50% drop) does not just hold share — it expands the addressable market. Some users priced out at $42 will buy at $33. Realistic elasticity for AI SaaS is around 0.3 to 0.7 — a 22% price cut lifts demand by 7-15%. The founder's gross-profit calculation should include the demand-expansion offset.
• Competitor cost basis. If competitors are larger and have negotiated discount terms with the vendor, their cost basis may already be lower than the founder's. A vendor price cut hits the founder's bill 1:1 but barely moves competitors' bills. In that scenario, pass-through pressure is intense and unavoidable; the founder's only defence is product differentiation.

The runway calculator handles the inverse scenario (cost spike instead of drop). The two together cover the realistic range of vendor-side surprises.

7. Solo founder playbook for a price drop

Five moves to make in the first 30 days after a vendor cuts prices:

• Recompute the unit economics. Run the engine within 48 hours. Know the kept-savings and passed-through margin numbers before deciding what to do.
• Hold list price for 60 days. Competitors will telegraph their move within that window. Cutting first is a mistake; cutting second to match is rational.
• Increase ad spend or content output. Margin expansion is a chance to fund growth. The ad-spend calculator can repoint the saved cost back into acquisition.
• Communicate the value, not the cost. Customers do not care about the founder's gross margin. They care about whether the product justifies its price. Use the saved budget to ship features that defend pricing, not to cut prices preemptively.
• Build a vendor-cost-shock buffer. A 50% cut today does not mean prices keep falling. Set aside the saved gross profit for 3-6 months as runway against the inverse scenario. The burn-rate calculator sizes the buffer.

One more pattern worth pricing explicitly: the vendor-tier swap. When a frontier vendor cuts the cheap-tier price (Haiku at $0.80/$4 vs Sonnet at $3/$15), the right response is rarely "celebrate." It is "audit which routes can move from the expensive tier to the cheap tier now that the gap is bigger." Migrating 40% of Sonnet calls to Haiku saves more than a 30% Sonnet price cut would on the same calls. This is invisible to a stress test that only models input-price changes, and it is where most of the real margin lives in a multi-tier vendor world. Re-run routing audits within 30 days of any major price announcement.

The corresponding worst-case scenario is also worth keeping in view. A vendor that lowers prices for new customers but holds old contracts at the old rate creates a margin gap between similar competitors that depends entirely on signup date. The honest defence is to renegotiate annually with each vendor, treating model contracts the same as enterprise SaaS contracts. Anthropic and OpenAI both rebill at announced rates on monthly billing; on volume contracts, ask for the new rate at renewal even if the contract has another quarter to run. Most vendors honor it.

Run the stress test quarterly and after every announced vendor price change. The single most useful number it returns is the "most-likely margin" — the central estimate of where margin ends up after partial pass-through. That number is what to plan against, not the kept-everything ceiling or the passed-everything floor. See the methodology for the full derivation^[5].