Retell AI at $60M/year up 650% YoY
Jan-Erik Asplund
TL;DR: As the cost and latency of the AI voice stack collapsed, Retell (2024) built the developer-first control layer for AI phone agents. Now it's gunning to rebuild the entire contact center stack, from front-line support to routing, QA, and testing, against Vapi, Sierra, and Decagon. Sacra estimates Retell hit $60M in annualized revenue in April 2026, up 650% year-over-year. For more, check out our full report and dataset on Retell.
Key points via Sacra AI:
- As the cost and latency of the AI voice stack collapsed in 2023, with GPT-4 Turbo getting 10x cheaper, Deepgram’s Nova-2 speech-to-text (STT) falling under 300ms, and Cartesia’s Sonic text-to-speech (TTS) under 90ms, Retell AI (YC W24) shifted from live-streamed dubbing for international creators into a developer platform that makes it easy for developers to create low-latency AI phone agents by writing prompts, defining tools & conversational logic, and hooking up 3rd-party tools & systems of record. Retell monetizes as a usage-based voice infrastructure layer, charging a $0.055/min platform fee on top of pass-through LLM, speech-to-text, and text-to-speech costs across providers like OpenAI, Anthropic, Gemini, Deepgram, Cartesia, and ElevenLabs, giving customers 70–95% lower per-minute costs than offshore human agents ($0.30–$0.80/min).
- After finding product-market fit with BPOs (Everise) and services businesses with high-volume, high-stakes support needs in finance (Sunshine Loans) and insurance (Matic), Sacra estimates Retell hit $60M in annualized revenue in April 2026, up 650% year-over-year and up from ~$45M at the end of 2025. Compare to enterprise AI customer support platforms Sierra at $150M in annual recurring revenue, valued at $15.8B for a 105x multiple, and Decagon at $35M in annualized revenue, up 1,567% year-over-year, valued at $4.5B for a 129x multiple, and to meeting recording developer platform Recall.ai at $31M ARR in January 2026, up 211% YoY, valued at $250M valuation for a ~12.8x multiple of its ~$19M ARR.
- Text-to-speech and speech-to-test providers like ElevenLabs ($500M ARR, up 175% YoY), Cartesia ($191M raised, Index Ventures), and Deepgram ($86M raised, Wing VC) are moving up into the agent layer, building their own workflow platforms for deploying agents on top of their native speech models & APIs, with Retell, Vapi, and Bland AI’s counter-move being to build model-agnostic, developer-centric control layers that allow teams to freely swap better performing models in & out. At the same time, AI support platforms like Sierra ($150M ARR, up 400% YoY) and Decagon ($35M ARR, up 1,567% YoY) are extending from chat into voice, attacking the top end of the market by building and managing agents for large enterprises vs. Retell’s self-serve approach of enabling developers to build their own customized agents.
For more, check out this other research from our platform:
Read more from
Retell AI revenue, growth, and valuation
Unlocked Report
Continue Reading
