SKU-ladder velocity: vendors restructure the menu faster than they change the prices
AI vendors now rework their plan *menus* faster than their prices. In the June 2026 capture batch, tier restructures — adding, retiring, renaming, or re-metering plans — outnumbered simple price changes nearly 2-to-1, and the churn has a direction: ladders fan out for self-serve buyers and collapse to a single 'call us' tier for enterprise.
What's happening — and why
What's happening: when an AI company changes its pricing, the part that moves is increasingly the structure of the plan ladder itself — a tier added, retired, renamed, split, or switched to a different meter — rather than just the dollar figure on an existing plan. In one week of June 2026, Tavus widened its ladder from four tiers to six while Synthflow collapsed its to a single Enterprise plan; Dust, WellSaid, Runway and Frase all rebuilt their menus the same week.
Why: AI unit economics are still in motion. Model costs move underneath every plan, the right value metric is unsettled (seats vs credits vs tokens vs outcomes), and vendors are still discovering what each buyer segment will pay. Restructuring the ladder — not just the price — is how a vendor searches that space: fan tiers out so sophisticated buyers self-select, or collapse them to force a sales conversation.
How it works
Evidence over time
27 supporting · 2 counter — hover or tap a point for detail, click to jump to the row.
Evidence
| Company | Date | What happened |
|---|---|---|
| Tavus | Jun 2026 | Restructured developer CVI plans from 4 to 6 tiers — added a $22 Starter and a $975 Business, renamed the old $59 plan to Builder, and cut per-minute overage rates. |
| WellSaid | Jun 2026 | Replaced its single $50 Creative seat with two tiers (Starter $10, Pro $33) and switched the capacity meter from downloads/year to minutes/year — a full ladder rebuild plus a meter swap in one change. |
| Dust | Jun 2026 | Scrapped a flat €29/seat unlimited plan held ~2 years for a USD, credit-metered Free/Pro($30)/Max($150) seat structure, adding a first free tier. |
| Runway | Jun 2026 | Retired the $95 Unlimited tier and replaced it with Max at the same price — unmetered Explore-mode generation swapped for a 9,500-credit monthly allowance with 1-month roll-over. |
| Synthflow | Jun 2026 | Removed its self-serve per-minute pay-as-you-go path and collapsed the ladder to a single Enterprise plan (contracts from $30K/year) — a multi-tier ladder retired to one quote-only SKU. |
| Frase | Jun 2026 | Held its $49/$129/$299 prices but rebuilt packaging as a 'content operating system' — added per-plan meters (AI generations, Content Guard pages, API caps) and switched overage to opt-in PAYG, off by default. |
| AiSDR | Jun 2026 | Moved from credit-and-message tiers to AI-researched-contacts metering (Solo $250 / Explore $900 / Scale $2,500) and added a new $250 month-to-month entry plan — re-based the meter and re-cut the ladder together. |
| Sourcegraph Cody | Jun 2026 | Retired its standalone self-serve Cody tiers into a single Enterprise plan (from $6K) with seat-scaled pooled credits; the agentic successor Amp was spun out with separate billing. |
| Windsurf | Apr 2026 | Killed the Cascade prompt-credit system for daily/weekly usage quotas, raised Pro $15→$20, and added a $200 Max tier — a meter swap plus a tier addition that drew a 33% price-hike backlash. |
| Cursor | Feb 2026 | Settled into a formalized six-plan lineup following its 2025 credit-pool transition — the ladder widened as the pricing model changed underneath it. |
| Perplexity | Jan 2026 | Restructured the developer API into a Search API plus a new Agentic Research tier (Jan 1), then split the enterprise offering into Enterprise Pro and Enterprise Max (Jan 15) — two ladder restructures inside two weeks. |
| Ideogram | May 2026 | Published a fresh four-tier grid (Free / Plus $15 / Pro $42 / Team $20-per-user) alongside a per-image PAYG API from $0.025. |
| HeyGen | May 2026 | Cut every plan over to a single unified credit currency spanning avatars, dubbing, and video — a whole-ladder metering change. |
| Galileo | Oct 2025 | Repackaged from two tiers (Developer $0 up to 3 users / Enterprise) to three (Free $0 unlimited users / Pro $100 / Enterprise) — a new mid tier inserted and the free cap removed. |
| Qodo | Jun 2026 | Collapsed a three-product split (Gen / Merge / Command) into a single pooled-credit Pro Team plan, dropped per-seat pricing entirely, and retired the permanent free Developer tier for a 14-day trial — a whole-ladder re-architecture plus a metering model swap (per-seat → $.012/credit pool) in one change. |
| Vercel | Jun 2026 | Retired the legacy Ultra/Premium plan, leaving Free/Team/Business/Enterprise, while cutting the v0 Max Fast model ~3× — a tier retirement bundled with a price move. |
| RunPod | Jun 2026 | Surfaced four distinct compute modes on the pricing page (Pods, Serverless, Clusters, per-request Public Endpoints) and re-sorted its GPU rate ladder (added B300 at $7.39/hr, re-priced L40S/L40) — menu expansion at the infra layer. |
| Weaviate | Jul 2026 | Dropped the mid Plus tier ($280/mo, SSO, 99.9% SLA), consolidating Weaviate Cloud from four plans to three (Free / Flex $45 / Premium $400) with Plus's SSO/SAML and uptime SLA rolled up into Premium — a tier deletion at the vector-DB layer. |
| Oxylabs | Jul 2026 | Flattened the Web Scraper API from a two-tab Regular/Enterprise layout into a single four-column grid (Micro $49 / Starter $99 / Business $999 / Custom+), removing the Advanced, Venture and Corporate named tiers from the page while leaving per-1K rates unchanged — a presentation/menu restructure, not a price move. |
| AssemblyAI | Jul 2026 | Reorganized the pricing page into six product tabs (Pre-recorded STT, Realtime STT, Voice Agent, Speech Understanding, Guardrails, LLM Gateway), broke Guardrails out as a distinct priced family, and swapped Whisper-Streaming for Universal-Streaming Multilingual — a whole-page re-architecture alongside a new product launch. |
| Flexprice | Jul 2026 | Reshaped both ends of the self-serve ladder without moving list prices: converted Free from a 1-month trial into an always-available $100K-cumulative-billings value cap, and moved the $1000/mo Scale tier's CTA from self-serve 'Get Started' to sales-assisted 'Contact Us' — a re-metering of the free tier plus a self-serve→sales boundary shift. |
| Fathom | Jul 2026 | A tier added by subtraction, and a direct hit on the collapse-for-enterprise half of the hypothesis: Fathom went from four plans to five by inserting a quote-only Enterprise tier — and populated it by REMOVING entitlements from the tiers below. SSO left Team ($19/$15) and custom data-retention policies left Business ($34/$25); both now sit in Enterprise alongside Launch Assist onboarding, org-wide security controls, Okta SCIM and a HIPAA BAA. Every seat price is unchanged. The ladder grew and no existing customer gained anything. Source: changes/fathom-2026-07-22-packaging.md. |
| Lago | Jul 2026 | The consistent case, at the other end of the same day: Lago replaced side-by-side Business and Enterprise columns with a single 'Available in Lago Premium' column, moved self-hosting into the base offer so deployment stopped being a tier gate, and unbundled 24x7 incident management and an Assigned Solution Engineer into on-demand add-ons. Both tiers were always quote-only, so no dollar amount moved — a pure menu collapse. The matrix simultaneously GAINED ten new feature rows (AI Billing Assistant, AI Finance Assistant, MCP server, wallet alerts, prepaid credits, grace-period management, receipts, security logs, RBAC with custom roles, Okta SSO), which is why tier count and capability count have to be read separately. Source: changes/lago-2026-07-22-packaging.md. |
| Fireworks AI | Jul 2026 | Menu expansion inside a single product rather than across tiers: serverless inference went from two quality-of-service tiers (Turbo, Priority) to three named serving paths — Standard (default), Priority (set via service_tier) and Fast (separate model ID, 100+ tokens/sec) — each with its own published input / cached-input / output rate, plus a new Fire Pass promo-code SKU carrying zero per-token charges for personal agentic coding. The ladder that grew was the SERVING-PATH ladder, not the plan ladder. Source: changes/fireworks-ai-2026-07-22-launch.md. |
| Observe.AI | Jul 2026 | Platform-level menu expansion with no prices attached: a new top-level Platform menu now lists five named modules — Agent Platform, Agent Harness, CoBuilder, Interaction Intelligence, Interaction Fabric — each with its own product page, replacing a navigation that had exposed only AI Agents plus Integrations and Trust. Modularization ahead of monetization: the same vendor publishes no prices on its own site at all (observe.ai/pricing 404s). Source: changes/observe-ai-2026-07-22-launch.md. |
| Hyperbolic | Jul 2026 | Expansion that moves capability INTO self-serve, which is the direction the hypothesis predicts: two self-serve surfaces became four documented ones — On-Demand (hourly, 99.5% SLA), Reserved (prepaid, self-serve from 1 week to 1 month), Private Cloud (custom contract) and Serverless Inference (99.9% SLA), plus separately metered storage volumes — where reserved clusters and dedicated hosting had previously been sales-quoted only. Funded by never-expiring prepaid credits with a $5 minimum and Auto Top-Up. Source: changes/hyperbolic-2026-07-21-packaging.md. |
| Mercor | Jul 2026 | Pure renaming as a ladder event — the form of restructure that no price diff detects: Mercor's enterprise product lines were renamed site-wide from Business / Enterprise evals / Data partnerships to Enterprise agents / Human data / Data monetization, with a new APEX research line added (APEX Benchmarks, APEX-Agents, APEX-SWE, Off-the-shelf data) and the data-partnership surface grown from 34+ to 50+ integrations. Nothing was priced, added or removed for a buyer; the menu was relabelled. Source: changes/mercor-2026-07-21-packaging.md. |
Counterexamples
- E2B · May 2026 — Headline prices identical across the entire Wayback record (2024-12 through 2026-05): $150/mo Pro and unchanged per-second vCPU rates — a durable, un-restructured ladder.
- Intercom · Jun 2026 — Ran a new-customer promo on the Essential plan ($29→$19) but left the Advanced ($85), Expert ($132), and $0.99/Fin-resolution structure entirely intact — a price tweak with the ladder held stable.
Trivia
-
Of the 16 customer-facing changes captured in the 2026-06-17 → 06-24 batch, 11 were plan-ladder restructures (tiers added, retired, renamed, or re-metered) rather than simple price moves — restructuring the menu outpaced changing the numbers nearly 2:1 in a single week.
-
Tavus expanded its CVI ladder 4→6 tiers and Synthflow collapsed its ladder to a single Enterprise SKU on the very same day (2026-06-24) — same category (voice/conversational AI), opposite directions, same date.
-
Dust's flat €29/seat Pro had held since at least May 2024 — through two funding rounds and a model-generation transition — before being scrapped on 2026-06-24 for credit-metered Free/Pro/Max seats. One of the corpus's most durable self-serve prices fell this cycle.
-
Qodo's 2026-06-30 change is the corpus's most total ladder re-architecture in a single move: it simultaneously collapsed a three-product split into one plan, swapped the metering model (per-seat → pooled $.012/credit), and deleted its permanent free tier — three structural changes most vendors stagger across multiple cycles, done at once. The same batch tilted toward simple price moves (7 of 12), so the restructure-beats-reprice ratio that defined the 2026-06-24 batch did not repeat — the tempo persisted but its composition shifted toward repricing.
-
Across three consecutive capture cycles (2026-06-24, 06-30, 07-06) the restructure direction converged on simplification: Synthflow → 1 SKU, Sourcegraph → 1, Qodo three-products → 1, Weaviate 4 → 3 plans, Oxylabs dropped three named tiers. The 07-06 cycle didn't add a single new customer tier to any corpus vendor's ladder — every restructure that week either merged or deleted a tier.
-
Free-tier retirement became its own restructure signature in the 06-30 → 07-06 window: Lokalise and Qodo both deleted permanent $0 plans, and Flexprice converted Free from a 1-month trial to a revenue-capped ($100K cumulative billings) always-on tier — three vendors in two weeks rewriting what 'free' means as part of a ladder rebuild, not a standalone price move.
-
The ratio survived the test it was set. It was flagged in June 2026 as a possible artifact of one capture batch; measured across all 303 logged change entries it is 131 packaging events to 61 price changes (2.15:1), and in the 15 days to 2026-07-30 it ran at 71 to 26 (2.73:1). Seventy-one restructures in a fortnight is too many to be a presentation effect of a single sweep.
-
Three vendors moved the enterprise end of their ladder in three different directions on the same day, 2026-07-22: Lago collapsed Business and Enterprise into one quote-only Premium package; Fathom ADDED a brand-new quote-only Enterprise tier as a fifth plan, filling it by pulling SSO out of Team and custom data retention out of Business; and Inflection AI deleted its Enterprise offering outright, leaving /enterprise a 404. The tempo claim survived this cycle; the "collapse for enterprise" directional claim did not.
-
Fathom's restructure is the clearest example of a ladder edit that adds a tier by SUBTRACTING entitlements rather than adding capability: seat prices did not move at all (Free $0, Premium $20, Team $19, Business $34), but single sign-on left Team and custom data-retention policies left Business, and both reappeared inside the new quote-only Enterprise tier. The menu grew; nobody got anything new.
For buyers
The plan you sign up on may not exist at renewal — especially on self-serve. Before committing annually, ask what happens to your tier if it's restructured (grandfathering, price-lock), and watch the meter, not just the price: a 'held price' can hide a new usage cap or an overage meter switched on. Frequent restructuring isn't automatically bad — it often adds cheaper entry tiers — but your line item is a moving target.
For vendors
Restructuring is a legitimate price-discovery tool, but each move spends trust — Windsurf's credit→quota switch drew a 33% price-hike backlash even though the headline barely moved. Make ladder changes legible: grandfather existing customers, name exactly what changed (tier, meter, allowance), and decide deliberately whether you're fanning out (self-select) or collapsing (sales-led). Build pricing infrastructure that lets you restructure without breaking live contracts.
Outlook — what to watch
Logged emerging in June 2026. The honest caveat is that absolute monthly counts are inflated by capture cadence — the weekly sweep logs in bursts — so the durable claims are the restructure-beats-reprice ratio and the expand-vs-collapse split, not a precise acceleration curve. It strengthens into a confirmed tempo if restructures keep outpacing price moves across the next capture cycles once the June burst washes out; it weakens if ladders stabilise as AI unit economics settle, marking this as a one-time round of pricing discovery.
Bottom line
Vendors are rebuilding their plan menus faster than they change prices: in June 2026, tier restructures outnumbered price moves nearly 2:1, fanning out for self-serve buyers and collapsing to one tier for enterprise.
FAQ
What is SKU-ladder velocity?
The rate at which a vendor changes the structure of its plan ladder — adding, retiring, renaming, splitting, or re-metering tiers — as distinct from simply changing the price of an existing plan. In the June 2026 corpus batch, 11 of 16 customer-facing changes were structural restructures rather than price moves.
Why are AI companies changing their pricing tiers so often?
AI unit economics are unsettled — model costs move under every plan and the right value metric (seats, credits, tokens, outcomes) is still being worked out. Restructuring the ladder is how vendors search for what each buyer segment will pay, fanning tiers out for self-serve buyers or collapsing them to force an enterprise sales motion.
Does frequent repricing mean a tool is a bad buy?
Not necessarily — restructuring often adds cheaper entry tiers or cuts prices. But it does mean the plan you buy may be reworked at renewal, so look for grandfathering and price-lock terms, and watch whether a 'held price' hides a new usage cap or overage meter.