Pricing trends

• Acqui-hire wave sweeps the AI tooling layer

  Sharpens

  24 companies

  The AI tooling and infrastructure layer is consolidating rapidly through acquisition — 19+ corpus companies saw M&A or acqui-hire events in roughly two years, with LLM observability/gateways the hottest segment (four exits in the five months to May 2026: Langfuse, Helicone, Galileo, Portkey). The acquirers are established platform companies (MongoDB, Elastic, Stripe, Kong, Zuora, CoreWeave, Nebius, Cisco, Palo Alto Networks, ClickHouse, Salesforce, Microsoft, Meta, Superhuman/Grammarly, Appier) absorbing specialist tools into their stacks. The fork holds: asset acquisitions (Voyage, Jina, Tavily, Metronome) tend to preserve the rate card, while acqui-hires (Clockwise, Robin, Rows, Humanloop, PlayHT) tend to sunset or gate the product.

  Voyage AI OpenPipe OpenMeter Jina AI Togai Metronome Tavily Galileo AdCreative.ai Rows Clockwise m3ter ScraperAPI Robin AI Unbabel Langfuse Helicone Portkey Humanloop PlayHT Windsurf Reclaim.ai Windsurf (Cognition) Continue

  Reviewed June 2026 See the pattern

• Voice APIs Converge on Per-Minute Billing

  Sharpens

  24 companies

  Voice and audio AI APIs are converging on media-minutes (per-minute or per-hour) as the dominant billing unit, displacing per-character and per-request alternatives. The pattern now holds across 20+ corpus voice companies spanning real-time agent APIs (Vapi $0.05/min, Retell AI per-minute PAYG), enterprise telephony (PolyAI, Synthflow bundled-minutes tiers), WebRTC/agent infra (LiveKit agent-session minutes + media minutes), emotional voice (Hume AI EVI per-minute), STT platforms (Gladia per-minute), and legacy TTS vendors shifting from per-character to per-minute for agent workloads (PlayHT, ElevenLabs Conversational AI). The character-only exception (lmnt) remains isolated. This reflects a structural shift: AI voice is increasingly deployed in agent calls where wall-clock time — not text volume — is the natural cost driver.

  assemblyai speechmatics bland-ai elevenlabs cartesia deepgram tavus murf-ai rev-ai krisp synthesia twelve-labs wellsaid hedra fal-ai retell-ai vapi synthflow livekit daily hume-ai polyai gladia playht

  Reviewed July 2026 See the pattern

• PLG Motion Locks In Public Pricing

  Holds

  21 companies

  Among the 100 corpus companies that use a PLG (product-led growth) sales motion, 96 publish public pricing — a 96% correlation. This is not coincidental: PLG requires frictionless self-service, which in turn requires visible pricing. The 4 companies that combine PLG with non-public pricing use PLG only as a discovery layer while routing material spend through sales. The inverse also holds: of the 16 companies with gated or sales-only pricing, only 4 use any PLG motion. The structural implication is that a company's pricing transparency can be predicted almost entirely from its sales motion choice — PLG mandates public pricing, sales-led permits opacity.

  openai anthropic elevenlabs replicate fal-ai firecrawl exa-ai perplexity-ai cursor deepseek groq upstash modal turbopuffer tavily composio retell-ai vapi lovable n8n arize-ai

  Reviewed June 2026 See the pattern

• Sales-led AI buys the Salesforce stack; usage-metered AI builds the meter

  Sharpens

  21 companies

  A companys monetization stack forks on who it sells to. Sales-led, enterprise-contract AI companies buy the standard enterprise revenue stack — Salesforce CRM, NetSuite (or SAP) for ASC 606 rev-rec, and a CPQ/quote-to-cash layer (Salesforce CPQ, DealHub, or Conga) — because their pricing is custom-quoted and needs approval routing, not metering. Usage-metered, developer-facing companies instead build a metering pipeline and frequently name no CRM at all, because their pricing is computed from usage events, not negotiated. Across the 252 corpus companies with a classified monetization stack: Salesforce (basis=vendor) appears in 63, NetSuite in 25, a CPQ-category vendor in 21; 45 companies pair a Salesforce vendor CRM with no in-house meter (the sales-led pattern), while 89 run an in-house meter and name no Salesforce (the usage-built pattern) — the usage-built side now roughly doubles the sales-led side. The split correlates with GTM motion rather than scale: 18 companies carry both halves (Anthropic, Perplexity, Vercel, Nebius and other frontier/infra vendors that sell self-serve and enterprise at once). HubSpot is the PLG-side CRM variant, so a handful of developer-first companies buy HubSpot+Gong instead of Salesforce while still building the meter.

  Skydio Cohere CoreWeave Clay Cresta Harvey Scale AI Cursor Langfuse Firecrawl Tavus Lovable Kustomer Parloa Snorkel AI Observe.AI PolyAI Chroma Weaviate RunPod OpenRouter

  Reviewed July 2026 See the pattern

• SKU-ladder velocity: vendors restructure the menu faster than they change the prices

  Holds

  21 companies

  AI vendors are restructuring their plan ladders — adding, retiring, renaming, splitting, and re-metering tiers — at a tempo that now outpaces simple price changes; in the 2026-06 capture batch, ladder restructures outnumbered price moves nearly 2:1. The churn splits by direction: ladders expand for technical/self-serve buyers (Tavus 4→6, WellSaid 1→2) and collapse for enterprise (Synthflow → 1, Sourcegraph → 1). The honest caveat: 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. Outcome TBD — this is either permanent higher tempo while AI unit economics stay in flux, or a one-time round of pricing discovery that settles as the market matures.

  Tavus WellSaid Dust Runway Synthflow Frase AiSDR Sourcegraph Cody Windsurf Cursor Perplexity Ideogram HeyGen Galileo Qodo Vercel RunPod Weaviate Oxylabs AssemblyAI Flexprice

  Reviewed July 2026 See the pattern

• Vendors withdraw public pricing as they mature or get acquired

  Sharpens

  21 companies

  A run of corpus vendors withdrew or hollowed out their public pricing in 2026 — Dropzone, Ada, Gladly, Clipdrop, OpenMeter, Unbabel, Robin — most of them either repositioning up-market into regulated/enterprise buyers (legal, customer-service, healthcare) or being acquired, with the gate slamming shut around the deal. Going gated is a maturation/acquisition tell, not a pricing-confidence signal. But the corpus moves both ways in the same window: Artisan, Gorgias, Abacus, and Regie all opened up or restored public prices — so 'withdrawal' is a directional pull, not a one-way ratchet. Self-serve PLG growth and enterprise/acquisition maturation pull pricing transparency in opposite directions.

  Dropzone AI Ada Gladly Clipdrop OpenMeter Unbabel Robin AI Sourcegraph Cody micro1 Lorikeet PhotoRoom Observe.AI Uniphore Phind Continue Flexprice Mercor Artisan 6sense Resemble AI Covariant

  Reviewed July 2026 See the pattern

• Infrastructure-Layer AI Vendors Standardize on Commitment Pricing

  Holds

  20 companies

  Infrastructure-layer AI companies (GPU clouds, inference APIs, serving platforms) almost universally pair public PAYG pricing with commitment tiers or reserved capacity discounts. Fifteen of nineteen infra-cloud segment companies have has_commits: true — a 79% rate, compared to 37% across the full corpus. The counterexamples (Cerebras, DeepInfra, Novita, Modal) are either early-stage or serve primarily developer-experimentation workloads with low repeat usage. Commitment pricing at the infra layer reflects a structural dynamic: GPU cost is the primary variable, and providers must hedge utilization risk; annual commits at 20–40% discounts let both sides manage capacity planning.

  anyscale baseten browserbase cerebras deepinfra e2b fireworks-ai groq lightning-ai modal replicate runpod together-ai turbopuffer vast-ai lambda-labs weaviate pinecone milvus bright-data

  Reviewed June 2026 See the pattern

• Three usage-billing vendors win the buy decision: Orb, Stripe Billing, Metronome

  Sharpens

  18 companies

  Among AI companies that buy (rather than build) their usage-billing layer, the vendor choice has concentrated on a three-name shortlist: Orb, Stripe Billing, and Metronome. Across 190 corpus companies with a monetization_signals block, Orb is named by 8 (baseten, chroma, fal-ai, fireworks-ai, pinecone, relevance-ai, replit-ai, vercel), Stripe Billing by 9 (arcads, browserbase, elevenlabs, higgsfield, intercom-fin, intercom, langfuse, linear, wisprflow), and Metronome by 4 (anthropic, claude-code, livekit, together-ai). The tail stays thin and idiosyncratic — OpenMeter (trigger-dev), Lago (mistral-ai), RevenueCat (suno, app-store) — so the buy decision is effectively a three-horse race, not a fragmented market. The corpus more than doubled (77 → 190) without diluting the three-name concentration, which sharpens rather than weakens the original read.

  Baseten Pinecone Vercel Relevance AI Replit AI Chroma fal.ai Fireworks AI Browserbase Intercom Langfuse Higgsfield Wispr Flow ElevenLabs Linear Anthropic Together AI LiveKit

  Reviewed June 2026 See the pattern

• Cloud marketplaces as the enterprise billing rail

  Holds

  17 companies

  Inference and compute vendors increasingly route enterprise billing through AWS / GCP / Azure Marketplaces — letting customers spend down existing cloud commitments instead of signing a new vendor contract. It is the default enterprise distribution rail for the infra cohort, and almost absent from app and seat-based vendors.

  Cohere Anyscale Modal Anthropic Baseten Fireworks AI Lightning AI RunPod Together AI Groq Replicate AssemblyAI Qdrant Dify Mistral AI Weaviate Pinecone

  Reviewed July 2026 See the pattern

• Token prices fall as access tiers rise

  Sharpens

  16 companies

  Per-token API prices fall relentlessly with each model generation and via caching/batch discounts, even as consumer subscription ceilings climb — compute deflates while the price of human-facing access inflates. The deflation now bends even long-held anchors (Claude Sonnet 5's intro $2/$10 undercut the four-generation $3/$15 Sonnet band), but it is not monotonic per SKU: a vendor can reprice a specific cost-sensitive model UP mid-life (Mistral Small 4, +50%/+100%, 2026-07-06) when it decides the model was under-priced.

  Anthropic Novita AI OpenAI Google DeepSeek Hume AI Together AI Vercel Speechmatics Lambda Labs You.com Character.ai Mistral AI xAI DeepInfra RunPod

  Reviewed July 2026 See the pattern

• Vertical enterprise AI keeps pricing gated

  Sharpens

  16 companies

  Pricing gating in the corpus now clusters around three distinct segments: (1) vertical enterprise-knowledge AI that sells top-down into regulated buyers (legal, enterprise search, clinical), (2) billing-infrastructure vendors that sell to other software companies, and (3) clinical/healthcare AI companies that sell into regulated clinical workflows (ambient note-taking, diagnostics). The overall gating rate at 338 companies is approximately 30% (100 of 338 gated or sales-only), still a clear minority but up from 16% at 97 companies — because the expanded corpus added more clinical AI, contact-center AI, and billing-infra vendors. Vertical-knowledge AI is the densest cluster: 27 of the 61 vertical-SaaS companies (44%) gate, well above the ~30% baseline. The structural driver is unchanged: gating correlates with top-down enterprise procurement into regulated buyers and with products whose value is hard to communicate on a public price page.

  Harvey Glean Jasper Spellbook Lago Metronome Orb Parloa PolyAI Uniphore Observe.AI Mercor Robin AI Legora Nabla Suki AI

  Reviewed June 2026 See the pattern

• Build the meter, buy the biller: the AI-infra monetization split

  Sharpens

  15 companies

  Usage-metered AI-infrastructure companies treat the metering pipeline as core IP they build in-house, but treat invoicing/payments as a commodity they buy. The recurring shape is an in-house meter (events to aggregation to rating) that ships usage out-of-band to a third-party billing system of record (Stripe, Orb, or Metronome) for invoicing, plans, and payment. Across 190 corpus companies carrying a monetization_signals block, the metering category resolves in-house 17 / vendor 8 (built ~2:1 over bought) while payments resolve vendor 40 / in-house 1 and billing vendor 29 / in-house 5 — companies build the meter and buy the biller. 15 companies disclose exactly this paired build-meter / buy-biller split; the counter-patterns are the fully-bought stack (the vendor meters too) and the fully-built stack (no vendor at all).

  Chroma Weaviate RunPod Baseten Cursor Perplexity AI Vercel Langfuse Pinecone Relevance AI Lovable Suno Synthesia Dust Anthropic

  Reviewed July 2026 See the pattern

• Credits as the universal currency

  Weakens

  15 companies

  AI vendors increasingly denominate *usage* in an internal 'credit' currency that floats across features — simplifying the invoice and decoupling the headline price from raw compute, at the cost of unit-economics transparency. Credits are the default *consumption* meter in AI-coding, outbound sales-tech, and creative apps. But the credit is not the corpus-wide modal unit: at 338 companies seats (175) clearly lead credits (123), because credits live on the usage layer while seats anchor the base of seat+credit hybrids. The two coexist — credits meter what you do, seats meter who does it.

  GitHub Copilot HeyGen Cursor (Anysphere) Recraft Descript Synthesia Codeium Cartesia Ideogram Lovable Creatify Bardeen Qodo Dust Zapier

  Reviewed July 2026 See the pattern

• Deal-desk and billing-engineering hiring marks monetization maturity

  Sharpens

  15 companies

  Whether a company is hiring deal-desk, RevOps, and billing-engineering roles is a leading indicator of where it sits on the monetization-maturity curve. The signal escalates: across the 295 corpus companies with a monetization-stack profile, customer-success (114) is the broadest retention bench and RevOps (107, ~36%) is near-universal once a revenue org exists; billing-eng (59) appears once usage- or hybrid-billing logic needs owning; a dedicated pricing/monetization role (51) follows; and a dedicated deal-desk (45) is the rarest and most diagnostic — it surfaces only when a company is negotiating non-list-price enterprise deals at enough volume to need a quote-approval function. (These are companies-with-the-role, not headcount.) Companies that open a deal-desk or an explicit pricing/monetization role are almost always layering a sales-led enterprise motion onto a self-serve base, which is why the same companies also carry the heaviest customer-success and retention benches. The counter-signal is the lean self-serve company that meters in-house but opens no RevOps/deal-desk req at all.

  OpenAI Anthropic Clay Decagon Stripe Billing Baseten Harvey Cohere Coreweave Sequence Nebius Modal Socket Waymo Parloa

  Reviewed July 2026 See the pattern

• Outcome-based pricing

  Sharpens

  14 companies

  Charging per resolved outcome or completed task — not per seat, message, or token — has matured from a curiosity into a category standard in AI customer support, where Ada, Gorgias, Intercom Fin, Kustomer, Lorikeet, Parloa, and Gladly all meter per resolution. Beyond support it is spreading as a per-task/per-action unit (Rox Agent Actions, Pixee per-fix, Digits outcome accounting, Mercor/Micro1 per-task labeling). The enabler is attributable, countable work product; the open risk is the opacity of the outcome definition, which now varies widely by vendor.

  Intercom Rox Gorgias Ada Lorikeet Kustomer Parloa Pixee Digits Recursion Mercor Micro1 Observe.AI HubSpot

  Reviewed July 2026 See the pattern

• The middle of the ladder is the unstable part: mid self-serve tiers churn, free and enterprise hold

  Holds

  13 companies

  The unstable part of an AI pricing ladder is the middle self-serve band — roughly $10–$200/mo Starter/Pro/Plus tiers. It gets re-priced, re-metered, split, renamed, or deleted far more often than the free tier (a stable top-of-funnel acquisition gate) or the enterprise tier (a stable 'contact us'). The mechanism: the mid band is exactly where vendors are still discovering willingness-to-pay and unit economics under moving model costs, so it absorbs the experimentation, while the two ends are structurally pinned. The terminal forms are visible too — fragmenting (WellSaid 1→2) or deletion (Synthflow → enterprise-only). Outcome TBD: the middle either settles as AI unit economics stabilize, or stays permanently fluid as long as model costs keep moving under it. Counterexamples (E2B's stable $150 Pro, Intercom's held Advanced/Expert) show it is a tendency, not a law.

  WellSaid Dust Runway Tavus Windsurf Synthesia Ideogram Galileo Synthflow Shortwave Creatify Qodo Lokalise

  Reviewed July 2026 See the pattern

• BYO-API keys zero out the model-cost meter

  Holds

  12 companies

  Agentic and workflow platforms increasingly split the bill into two meters — a platform/orchestration credit they keep, and a passthrough model/data cost they mark up — then offer Bring-Your-Own-Key as a lever that zeroes (or halves) the passthrough meter. BYOK turns the vendor's role explicit: you pay them for orchestration, and pay the model provider directly for inference. It is a transparency and margin signal — the platform concedes it is not the cost center on inference — but it is not universal, and some seat+credit tools are pulling the lever back to protect bundled-credit revenue.

  Clay Relevance AI Gumloop Byword Vectara Lemlist Tabnine Deepgram Synthflow Aider Continue OpenRouter

  Reviewed July 2026 See the pattern

• Spend guardrails arrive after the bill-shock backlash

  Holds

  12 companies

  After a run of public bill-shock incidents (Cursor's refund apology being the archetype), usage- and credit-metered vendors are bolting on spend-control governance — prepaid balances, auto-recharge, flexible spending credits that absorb overage, and advance-notice pricing policies. The guardrail is becoming a packaging feature, not an afterthought.

  Cursor (Anysphere) Vercel GitHub Copilot Firecrawl ElevenLabs Manus Runway Augment Code Clay Lovable Vapi Synthflow

  Reviewed June 2026 See the pattern

• Billing-infra vendors gate their own pricing

  Sharpens

  11 companies

  Companies that sell usage-based billing infrastructure to other companies disproportionately hide their own prices behind a sales conversation — the cobbler's children go barefoot. At 219 companies, the billing-infra segment gating rate (~60%+) remains well above the corpus-wide baseline (~20%), and the majority of the largest players (Chargebee, Maxio, Zenskar, Sequence, Kill Bill, Lago, Metronome, Orb, Togai, Finout) are gated. The pure public exceptions (Flexprice open-source, Hyperline, Schematic, Alguna, Stripe Billing) are real but do not yet overturn the pattern. The universal-gating framing no longer holds precisely, but the strong clustering does.

  Lago Metronome OpenMeter Orb Togai m3ter Chargebee Maxio Zenskar Sequence Finout

  Reviewed June 2026 See the pattern

• The $200 prosumer ceiling

  Sharpens

  11 companies

  A ~$200/month 'Max / Ultra / Pro' prosumer tier has emerged across consumer AI — roughly 10× the $20 'Plus' tier set in 2023 — as vendors monetise power users with priority or uncapped access rather than raising the mass-market price.

  OpenAI Perplexity AI You.com Anthropic Cursor (Anysphere) Google Windsurf Factory Manus Hume AI Lovable

  Reviewed June 2026 See the pattern

• The meter tracks the buyer: usage for developers, commitments for enterprise

  Holds

  11 companies

  Pricing structure in the corpus correlates tightly with who buys: pure-usage metering clusters on the developer-as-buyer, while annual commitments are an enterprise artifact. The billable structure is chosen to fit the buyer's procurement habits, not the underlying compute.

  DeepSeek Modal Fireworks AI Together AI Anyscale Glean Harvey Vercel Lambda Labs Vapi LiveKit

  Reviewed June 2026 See the pattern

• AI-coding tools standardize on seat + usage credits

  Holds

  10 companies

  Every AI-coding tool in the corpus has converged on the same pricing architecture: a per-seat subscription anchoring the base, layered with a usage-credit pool that meters agentic code-generation runs. The seat buys the developer their IDE access; credits meter how much AI work the IDE actually does — and the credit pool replaced earlier unlimited or request-based models in 4 of 5 cases. The 2026-06-30 edge case (Qodo) shows the credit pool can now outlast the seat anchor, decoupling the two halves.

  Cursor (Anysphere) Codeium GitHub Copilot Augment Code Qodo Windsurf Sourcegraph Cody Tabnine Bito Continue

  Reviewed June 2026 See the pattern

• Clinical AI converges on per-clinician gated contracts

  New

  10 companies

  Every clinical AI company in the corpus that deploys at health-system scale prices on a per-clinician (per-seat / PEPM) annual contract, gates pricing 100% behind sales, and has no public self-serve plan. The PEPM unit is driven by EHR integration economics (Epic, Cerner charge by enrolled clinician), regulatory framing (FDA clearance and HIPAA compliance anchor the value to the provider seat, not to note volume), and the enterprise procurement cycle of health systems (IT committees, legal review, multi-year negotiation). The result is a sub-segment where the pricing model is as uniform as in any SaaS category yet more opaque than any other in the corpus.

  Abridge Ambience Healthcare Suki AI Nabla Freed Heidi Health Paige AI Rad AI Viz.ai Hippocratic AI

  Reviewed June 2026 See the pattern

• Unit proliferation at the infra layer

  Sharpens

  10 companies

  Infrastructure and platform vendors keep adding metered dimensions as they ship features — drifting from one billable unit toward many — while consumer apps hold to a single unit. The metering surface is fragmenting at the infra layer and consolidating at the app layer.

  Vercel Modal ElevenLabs Deepgram RunPod LangChain (LangSmith) LiveKit Weaviate Anthropic AssemblyAI

  Reviewed July 2026 See the pattern

• AI companies are staffing a cost-FinOps role to defend the margin under the price

  New

  9 companies

  A distinct margin-discipline cohort is emerging: AI companies whose price is fixed (a flat seat, a per-clinician license, a per-unit credit) while their AI compute cost per unit is not are staffing a dedicated cost-FinOps / strategic-finance / infrastructure-cost role to own the gross-margin and unit-economics floor beneath that price. Of the 295 corpus companies carrying a monetization_signals block, 10 disclose this role — abridge, apptronik, canva, langchain, lightning-ai, roboflow, runpod, tavily, and zapier each with a sourced cost-finops hiring signal, plus modal with in-house GPU-utilization/cost instrumentation as a stack entry. This is the cost-side mirror of the value-metric/outcome-pricing thesis: where outcome pricing engineers the value side of the meter, cost-FinOps hiring engineers the COGS side, so the spread the company keeps becomes an owned, staffed function rather than an accident of cloud spend. It is a cohort and not yet a corpus-wide norm — assemblyai (pure per-hour audio, in-house meter and LLM gateway) and firecrawl (first-party credit ledger, PLG-shaped hiring) carry the same GPU/COGS exposure but disclose no cost-FinOps or FinOps role at all.

  Canva Zapier LangChain Lightning AI Abridge Tavily Roboflow Apptronik Modal

  Reviewed June 2026 See the pattern

• Batch and cache discounts as a standard playbook

  Sharpens

  9 companies

  Inference vendors have converged on batch processing (~50% off) and cached-input discounts as the standard levers to price-discriminate by latency tolerance — and increasingly on a paired surcharge for priority/fast scheduling — a de-facto pricing playbook for token APIs that smaller app vendors do not offer.

  OpenAI Google Fireworks AI Groq Anthropic Baseten Mistral AI OpenRouter DeepInfra

  Reviewed July 2026 See the pattern

• Per-action pricing for agent tools

  Sharpens

  9 companies

  As agents call tools — web search, page visits, code execution, deep research — vendors price those actions individually, per call, alongside (and on top of) tokens. No longer confined to search-native APIs: it now reaches a general-model API and browser-agent infra. The billable unit is shifting from the token to the action.

  Groq Perplexity AI You.com Exa Anthropic Browserbase Composio Zapier Vapi

  Reviewed July 2026 See the pattern

• Prompt Caching as a Structural Pricing Feature

  Holds

  9 companies

  Prompt and context caching has emerged as a structural pricing layer across LLM API providers, offering discounted rates (typically 75–90% off input token price) for repeated context. Nine corpus companies have published cached-input pricing. The pattern is not yet universal — most open-source-first or smaller inference providers have not implemented it — but the major frontier-model providers and the largest inference platforms have all adopted it. Caching creates a pricing wedge: heavy users with stable system prompts get lower effective rates, while one-shot or diverse-prompt workloads pay full price. This rewards sticky, predictable usage patterns and raises switching costs.

  anthropic openai google-gemini deepseek together-ai groq fireworks-ai baseten replicate

  Reviewed June 2026 See the pattern

• Daily-refresh credit pools on prosumer AI plans

  Sharpens

  8 companies

  A handful of prosumer/agentic AI apps are layering a small daily-refresh credit drip onto (or beneath) the monthly credit pool — 300/day at Manus, ~10/day at Higgsfield free, 100/day at Genspark free, daily refresh on Julius's annual pools. The mechanic does two things at once: it caps the worst-case daily blast radius of an expensive agent run, and it nudges habitual daily engagement (use-it-or-lose-it). Whether this becomes a durable consumer-AI billing convention or stays a free-tier throttle is still open — most consumer credit plans (Suno, Runway) remain single monthly buckets.

  Manus Julius AI Genspark Higgsfield v0 Lovable Suno Powerdrill

  Reviewed July 2026 See the pattern

• From subscriptions to usage — and back to hybrid

  Holds

  8 companies

  A wave of vendors abandoned flat subscriptions for pay-as-you-go credit metering — but several then re-introduced tier structure on top, converging on hybrid rather than pure usage. The endpoint of the migration is seat-plus-usage, not pure PAYG.

  Exa Tavus ElevenLabs Deepgram Lindy Synthflow Dust Qodo

  Reviewed July 2026 See the pattern

• Incumbent CRM/sales SaaS bolts an AI credit meter onto the seat

  New

  8 companies

  The AI transformation of incumbent, seat-licensed CRM and sales-engagement SaaS is arriving as a bolt-on consumption meter, not a repricing. 8 of the 13 CRM/sales incumbents in the corpus keep the per-seat license as the revenue anchor and add a separate AI credit/consumption currency on top — HubSpot Credits, Microsoft Copilot Credits, Close's Chloe credits, Freshworks' per-session Freddy meter, and the seat+credit pools of ZoomInfo, Outreach, and 6sense (plus Keap's usage add-ons). The private-currency 'credit' does double duty: it lets the vendor keep the familiar seat contract intact while capturing volatile AI-inference cost on a meter that can be repriced without touching the seat, and it obscures the per-action dollar cost of AI. The holdouts (Pipedrive, Copper, Zoho, SugarCRM, Salesloft) are the mid-market pure-seat CRMs that fold AI into the edition instead. The split tracks category — horizontal platforms and data-heavy sales-intelligence tools meter; pipeline CRMs bundle — not company size. Whether the meter becomes the dominant growth lever (Close already bundles the agent free and monetizes only the credits) or stays a modest overage line is the open question.

  HubSpot Microsoft Dynamics 365 Close CRM Freshworks ZoomInfo Outreach Keap 6sense

  Reviewed July 2026 See the pattern

• Outbound sales-tech converges on seat + enrichment credits

  Holds

  8 companies

  Outbound-sales and marketing-automation AI companies have independently converged on the same pricing structure: a per-seat subscription tier anchoring the base, layered with a rechargeable credit pool for enrichment lookups, intent signals, and AI-generated sequences. The seat covers the workflow surface; credits meter the data and AI consumption that makes the workflow valuable.

  Apollo.io Lemlist Reply.io Artisan Regie.ai Smartlead Instantly Juicebox

  Reviewed July 2026 See the pattern

• Self-serve list-price inflation

  Sharpens

  8 companies

  Application-layer AI vendors on self-serve plans are raising headline list prices in mid-2026, often by double or triple digits, even as the inference infrastructure beneath them keeps deflating. The driver is margin recovery: as the early land-grab phase ends and free GPU-price tailwinds get competed away at the infra layer, app vendors that had under-priced to acquire users are now repricing toward sustainable unit economics — frequently bundling the hike with a meter change (more included allowance, a credit pool, or a model toggle) to blunt churn. As of the 2026-07-06 batch the raise-side is no longer app-only: a frontier lab (Mistral) raised developer-API token rates while holding its consumer ladder, and vendors are also raising prices by letting acquisition promos expire (Intercom's 35%-off Essential reset $19→$29). The open risk is that coordinated self-serve hikes invite churn to cheaper rivals in categories with low switching costs (email clients, creative tools, outbound).

  Shortwave Creatify lemlist Dust Lokalise Tavus Mistral AI Intercom

  Reviewed July 2026 See the pattern

• PLG reprices in public, sales-led reprices in private — the visible velocity gap

  Sharpens

  7 companies

  Visible pricing velocity tracks go-to-market motion: 41 of the ~49 attributable SKU-change events in the corpus come from PLG vendors and only 2 from purely sales-led ones. PLG pricing is public, self-served, and A/B-iterable, so it gets restructured in the open and often; sales-led pricing changes happen privately, inside contract renewals, and leave no public trace. The velocity gap is therefore partly real (PLG genuinely iterates more) and partly a measurement artifact (sales-led repricing is invisible to a public-pricing feed) — and the artifact is the practical lesson. Buyer takeaway: if you buy PLG, expect your plan to be restructured under you within ~12 months; if you buy sales-led, expect a stable-looking page and per-renewal negotiation instead. The rate card you can see is not the cadence you will experience.

  Corpus tally Perplexity Windsurf Dust HeyGen You.com Batch tally

  Reviewed July 2026 See the pattern

• Agents priced as headcount: the salary anchor enters the rate card

  New

  6 companies

  Agentic vendors increasingly package agents as headcount — named workers (Alice, Julian, Jason) at flat monthly salaries, anchored explicitly against human payroll — because the salary anchor justifies prices 5–20x a SaaS seat and shifts the comparison set from software budgets to hiring budgets. The frame is unstable: Motion reverted to credit plans within a quarter, and credit-metered agents (Manus, Genspark) compete directly against the salary framing with usage math.

  Artisan 11x Lindy Reply.io Juicebox Dropzone AI

  Reviewed June 2026 See the pattern

• AI usage graduates onto its own dedicated meter, decoupled from the base plan

  Sharpens

  6 companies

  Platform and infrastructure vendors are beginning to pull AI usage onto its own dedicated billing meter — a separate credit currency, marketplace consumption unit, or thin per-token platform fee — that is priced independently of the base product's seat, plan, or edition. The driver is unit-economics honesty: as AI inference becomes a large, volatile, cost-of-goods line, vendors want its price to float on its own axis (routing, model, or marketplace) rather than inherit the base plan's tiering, so that a cheaper model or a cross-region route flows straight to the meter without re-rating the whole account. The countervailing force is bundling: most SaaS still folds AI into the per-seat price (Shortwave, GitHub Copilot) because an AI-inside-the-seat story is simpler to sell and lets the vendor capture AI upside inside ARPU. Outcome TBD — with three same-day adopters at the platform layer this is a directional signal, not yet a category standard; the test is whether app-layer and seat-based vendors follow the platform layer in decoupling.

  Snowflake Cortex Vercel Anthropic Frase Weights & Biases Google

  Reviewed July 2026 See the pattern

• Retrieval and data-API vendors widen free tiers to grab evaluation

  New

  5 companies

  Retrieval, scraping, extraction, and vector-DB API vendors are competing on the width of the free/evaluation entry, enlarging recurring free tiers and cutting entry floors to win developer land while planning to monetize at scale. In one 2026-07-14 window, Bright Data added a recurring no-card free tier, Exa 20x'd its free tier, Powerdrill put a daily free-credit pool on every tier, Unstructured made its free pages a monthly reset (and capped PAYG bills), and turbopuffer cut its entry minimum 75%. The unifying logic is that for usage-metered data/retrieval infrastructure the marginal cost of a small evaluation workload is near zero, so a generous recurring free tier is cheap customer acquisition — the revenue comes at production volume. The countervailing move is that the SAME vendors gate their premium/intelligence products upward: Bright Data raised Bright Insights 5-8x and Linkup's new Extract requires a paid balance. So the pattern is a barbell — commodity request meters open to a free land-grab while high-value intelligence surfaces gate up. Open question: whether the free-tier widening is a durable acquisition strategy or a temporary response to crowded search/scraping API competition.

  Bright Data Exa AI Powerdrill Unstructured turbopuffer

  Reviewed July 2026 See the pattern

• Search APIs split: commodity calls deflate, 'research effort' inflates

  New

  5 companies

  Agent-facing search APIs are bifurcating into two price layers: a commodity per-call layer converging around $5–7 per 1,000 requests (and falling — You.com cut 10x in April 2026), and an 'effort-priced' research layer at 10–400x per call where price scales with the compute the agent spends on the task rather than the data returned. Vendors that can't sell effort (SerpApi, proxy networks) stay on pure volume ladders and compete on deflation.

  You.com Exa Tavily Linkup Groq

  Reviewed June 2026 See the pattern

• SEO tools reprice around AI-search visibility

  Holds

  5 companies

  A cohort of SEO and AI-writing tools is repricing around generative-engine optimization (GEO): instead of selling words or documents, they now meter and gate on AI-search visibility — 'visibility prompts' tracked and the number of answer engines (ChatGPT, Perplexity, Google AI Overviews, Gemini) monitored. The value metric is migrating from 'content produced' to 'presence in AI answers tracked,' often arriving alongside an acquisition or a deep annual promo. The shift is real but not yet universal — classic per-seat writing tools persist.

  Writesonic Frase Surfer SEO Scalenut Apollo.io

  Reviewed June 2026 See the pattern

• The voice minute is unbundling into pass-through component meters

  New

  5 companies

  Self-serve voice-agent platforms are decomposing the conversation minute into a thin per-minute platform fee plus at-cost pass-through of the component stack (LLM, TTS, telephony), making bring-your-own components a first-class discount lever — voice AI's version of the BYOK split between orchestration and inference. The counter-camp (Cartesia, Bland, Tavus) keeps selling one bundled minute, betting buyers prefer a single predictable number over a cheaper assembled stack.

  Synthflow Retell AI Vapi Deepgram ElevenLabs

  Reviewed June 2026 See the pattern

• Billing vendors bill themselves on their own product

  Sharpens

  4 companies

  Companies whose product is a usage-billing platform tend to dogfood it — running their own subscription/usage billing on the very engine they sell — which doubles as a credibility signal and a built-in product test. Of the 6 billing/metering vendors in the corpus (at N=190) carrying a monetization_signals block, 5 disclose a billing system of record and 4 of those dogfood: Lago bills itself on Lago, Sequence runs Sequence as its Accounts-Receivable / quote-to-cash engine, and Metronome and Stripe Billing both name their own product as their in-house billing platform. The exception is Orb, whose only named monetization tool is Salesforce — it appears as the bought biller in 8 other corpus stacks (baseten, chroma, fal-ai, fireworks-ai, pinecone, relevance-ai, replit-ai, vercel) but does not itself disclose running its own product as system of record. The sixth, Alguna, sells quote-to-cash but discloses no biller in its block, so it is left as an honest unknown rather than counted either way. The dogfood norm is strong — 4 of the 5 vendors that disclose a biller run their own meter — but one notable vendor that powers others' billing publicly breaks it.

  Lago Sequence Metronome Stripe Billing

  Reviewed June 2026 See the pattern

• Expiring generosity: unlimited plans and promo rates with printed end dates

  Emerging

  3 companies

  A small set of vendors is starting to time-box generosity — 'unlimited' plans guaranteed only until a printed date (Genspark), promotional rates displayed with their own struck-through reversal (Deepgram), and discounts with built-in six-month fuses (Oxylabs). The mechanism hedges unpredictable model and infrastructure costs without committing to a list-price change. Too early to call: the expirations land late 2026, and the test is whether they convert into price increases, quietly renew, or get dropped.

  Genspark Deepgram Oxylabs

  Reviewed June 2026 See the pattern

• Homegrown billing gets retired as companies scale

  Emerging

  3 companies

  Usage-metered AI companies tend to start with a hand-rolled billing system, then retire it for a billing vendor once the product line and scale exceed what the homegrown version supports — but they migrate only the invoicing/plan/system-of-record layer, keeping the metering pipeline in-house. Pinecone (homegrown engine that still required manual invoices, replaced by Orb at serverless launch) and Langfuse (homegrown per-event billing replaced by Stripe Billing) remain the only two clean billing-layer migration arcs as of the 190-block review; LiveKit shows the rev-rec analogue (legacy QuickBooks to NetSuite). The signal does NOT generalize cleanly: at scale, in-house billing more often appears as a build-and-keep (Synthesia, Zapier, GitLab) or a from-scratch meter-plus-bought-biller split (Chroma, Vercel, OpenAI), and some companies skip the homegrown phase entirely (Browserbase, Mistral). The counter-direction also exists — Anthropic is building a homegrown ledger beyond off-the-shelf platforms — so this stays an emerging signal, not an established law. The trigger in both clean migration cases is the same: multi-product billing, custom plans, and version-controlled pricing that the homegrown system could not deliver without engineering bottlenecks.

  Pinecone Langfuse LiveKit

  Reviewed June 2026 See the pattern

• Annual prepay discount compression

  Emerging

  2 companies

  A handful of self-serve AI vendors are compressing or removing the annual-prepay discount — shrinking the months-free incentive, or dropping the annual option altogether for monthly-only billing. The logic mirrors self-serve list-price inflation: a deep annual discount locks in a low rate during a repricing season, so vendors raising prices also want to stop pre-selling the old (cheap) rate a year forward. The countervailing force is cash-flow and retention — annual prepay is the cheapest way to lock revenue and cut churn — so this may stay confined to vendors actively repricing rather than becoming a category norm. Outcome TBD.

  Creatify Shortwave

  Reviewed July 2026 See the pattern

No trends in this view.