Skip to content

The Fee Retreat: Q1 2026's Margin Squeeze

Q1 2026 brought a quiet pattern: companies pulling back on give-aways. Not price increases — capacity decreases. The common thread is AI cost pressure forcing vendors to reclaim margin they gave away during growth phases.

The Evidence

Sources: PricingSaaS Q2 2026 Report [PricingSaaS Q2 2026 Report — no public URL] and Data Labs dataset.

Company What Retreated Before After Change Source
Cledara Cashback rate 2%, uncapped 1%, capped at sub cost, time-limited, restricted eligibility -50% rate + 9 restriction events PricingSaaS Q2 2026, p.16
Relay AI credits 5,000/mo (Pro/Team) 2,000/mo -60% credits PricingSaaS Q2 2026, p.12-13
Relay Add-on bundle cap 1M credits 200K credits -80% cap PricingSaaS Q2 2026, p.12-13
Relay Add-on price (10K) $19 $23 +21% price PricingSaaS Q2 2026, p.12-13
Replit Core credit allocation $25/mo $20/mo -20% value PricingSaaS Q2 2026, p.12
ElevenLabs Included minutes Higher Lower, overage price raised Double squeeze PricingSaaS newsletter
Alchemy Enterprise webhooks Unlimited 500 Hard cap added Dataset
Taskade AI credits Renewable monthly One-time lifetime allocation Recurring → finite Dataset

Why It's Happening

Three forces converging:

  1. AI inference costs are real. Companies that bundled AI generously in 2024-2025 are discovering that heavy users drive disproportionate compute costs. Credit cuts are surgical margin recovery.
  2. Growth-phase generosity expires. Cledara's uncapped 2% cashback was a growth tactic. At scale, it becomes a margin drain. The retreat pattern is: reduce rate, cap exposure, restrict eligibility, time-limit for new customers.
  3. Usage-based companies discover floor risk. When your pricing tracks consumption, a downturn in customer usage hits your revenue immediately. Credit cuts create a synthetic floor.

The Counter-Trend

Not everyone retreated. Renderforest doubled AI credits across plans in Q1 2026. Synthesia tripled video credits. Both are creative tools competing for AI-forward positioning — they're buying share with credits while others are cutting.

What to Watch

The fee retreat is a leading indicator. Companies that cut capacity in Q1 will likely raise prices in Q2-Q3. The playbook is: reduce value per tier first (less backlash), then reprice. Relay already did both in the same quarter.

Update: May 2026

The pattern continues, now reaching the largest AI companies.

Company What Retreated Before After Mechanism
Anthropic Enterprise tokens Bundled in seat price All usage billed separately at API rates Unbundled — lower seat price, potentially 3x TCO for heavy users
Figma AI usage Loosely metered Hard caps: Free 500/mo, Ent 4,200/seat/mo Soft limits → hard credits (Mar 2026)
Notion AI agents All AI unlimited Standard unlimited, agents $10/1K credits Surgical meter on highest-cost feature only

The Q1 pattern holds: companies aren't raising prices, they're metering what was previously unmetered. The fee retreat isn't a one-quarter phenomenon — it's the new normal for AI-bundled SaaS.

May 2026 continued

Company What Retreated Before After Mechanism
Zendesk Overage billing Manual/warned overage billing Auto-bills overage resolutions, no warning, no discount (since Jan 2026) Silent policy change
Anthropic Token costs Flat per-token rates Opus 4.7 tokenizer produces 32-45% more tokens = 12-27% stealth increase Tokenizer change
Gong Feature bundling All modules bundled Forecast, Engage, Enable, Data Cloud unbundled into paid add-ons (+25-56% effective cost) Feature unbundling

New pattern: "Shrinkflation." Vertice SaaS Inflation Index ([Vertice SaaS Inflation Index 2026](https://www.vertice.one/l/saas-inflation-index-report)) reports 28% of Q4 2025 renewals reflected reduced value without price decrease. SaaS inflation rate hit 12.2% overall, 13.2% in March 2026. Companies are getting creative: tokenizer changes (Anthropic), auto-billing without notice (Zendesk), module unbundling (Gong). The fee retreat isn't just about cutting credits — it's about reducing value per dollar in ways that don't show up on a pricing page.

Sources — Companies Referenced

Every claim above traces back to structured data from these company profiles.

Data Sources

dataset Data Labs dataset query

Linked From

Contact us