Vapi

Learning Objectives

• Understand what Vapi does and where voice-agent infrastructure sits in the AI stack
• Identify the customer profile and call-volume scale that justify a voice-AI platform purchase
• Evaluate Vapi against alternative voice-AI options (OpenAI Realtime API, ElevenLabs, in-house builds)

What Is Vapi?

Vapi is a voice-agent platform that lets companies build, deploy, and manage AI voice agents for customer support, lead qualification, appointment scheduling, and outbound sales calls. Founded in 2020 by Jordan Dearsley and Nikhil Gupta — University of Waterloo classmates who went through Y Combinator with an AI therapy chatbot before pivoting into voice infrastructure — Vapi has grown into one of the most heavily used voice-AI platforms by call volume, processing between 1 and 5 million calls daily and over 1 billion calls in total.

Vapi operates two product surfaces under the same platform:

• Enterprise platform — sales-led, custom integrations, used by Amazon Ring, Kavak, Instawork, New York Life, and Intuit
• Self-serve developer platform — pay-per-use API for engineers building voice-first applications

May 2026 Series B — $50 Million at $500 Million Valuation

On May 12, 2026, Vapi closed a $50 million Series B at a $500 million post-money valuation, led by Peak XV Partners (formerly Sequoia India). Participating investors include Microsoft's M12, Kleiner Perkins, and Bessemer Venture Partners. Total funding now sits at $72 million across all rounds.

The Microsoft M12 participation is structurally notable: Microsoft positions Azure Communication Services and its own AI voice stack as competitive products, so an M12 check signals that Microsoft sees Vapi as complementary voice infrastructure (or a future acquisition candidate) rather than a pure competitor. Peak XV's lead on the round reflects continued bullishness on voice-AI as a category, after similar bets on global voice-AI infrastructure across the firm's portfolio.

The Amazon Ring Anchor — 40 Competitors Beaten

The defining 2026 commercial milestone is Amazon Ring's decision to route 100 percent of inbound calls through Vapi after evaluating more than 40 competing voice-AI platforms. Ring's scale — millions of doorbell devices generating customer-support inquiries — makes it one of the largest voice-AI deployments in any consumer hardware category, and the head-to-head selection over 40 alternatives is the kind of social proof that drives enterprise procurement decisions downstream.

Other named customers show vertical breadth rather than single-industry concentration:

• Kavak — used-vehicle marketplace, sales and customer service voice agents
• Instawork — gig-staffing platform, candidate qualification and shift coordination
• New York Life — insurance, inbound policy-holder support
• Intuit — finance, tax-season inbound support volume

The mix signals that voice AI has crossed from pilot to production deployment in mainstream enterprise IT budgets.

Core Platform Capabilities

Voice Agent Building

Vapi handles the infrastructure plumbing that's hard to get right — low-latency speech-to-text, dialogue management, text-to-speech, conversation state, and call routing — so that builders can focus on agent logic and business workflow. Agents can be configured with custom system prompts, knowledge bases, business rules, escalation paths, and analytics.

Multi-Model Voice Stack

Vapi composes voice agents from multiple model providers rather than locking customers into a single voice model. Builders can pick the speech-to-text engine, the dialogue LLM, and the text-to-speech voice independently — useful for cost-versus-quality tuning and for hedging against any single provider's outage or pricing change.

Enterprise Telephony Integration

For enterprise customers, Vapi handles the carrier-grade plumbing: SIP trunk integration, PSTN connectivity, call recording compliance, regional number provisioning, and the routing logic that's required for production-grade phone systems rather than just developer demos.

Developer Self-Serve

A self-serve API tier lets independent developers build voice agents without an enterprise sales cycle. The Y Combinator alumni network and dev-tools audience have driven significant top-of-funnel usage that ultimately graduates to enterprise contracts.

Pricing

Public per-minute pricing for the developer tier is published on the Vapi website; enterprise pricing is custom-negotiated with volume discounts. Total bill for an enterprise call program depends on call volume, voice model selection (premium voices cost more per minute), and feature mix (transcription archive, compliance, recording).

Competitive Landscape

Vapi's most direct alternatives fall into three groups:

• Other voice-agent platforms — Companies like Bland AI, Synthflow, and Air AI target similar customer profiles. Vapi's call-volume lead and Amazon Ring win create category-defining commercial proof points that smaller competitors will struggle to match in 2026.
• Foundation-model voice APIs — OpenAI Realtime API (with GPT-Realtime-2, GPT-Realtime-Translate, and GPT-Realtime-Whisper, launched May 7, 2026) and Google Gemini Live offer raw voice model access without the orchestration layer. Vapi sits on top of these models, not against them — customers who want the orchestration + telephony + analytics package use Vapi, customers building their own voice stack call the model APIs directly.
• Voice-specialist TTS/STT — ElevenLabs (premium voice generation), Murf AI, and Suno specialize in narrow voice tasks but don't provide the full voice-agent orchestration that Vapi does.

Company Details

Strengths

• Production-scale call volume — 1 billion-plus lifetime calls is the strongest empirical proof point in the voice-agent category
• Anchor enterprise win — Amazon Ring beating 40 competitors is the kind of head-to-head selection that drives procurement decisions across enterprise IT
• Multi-model voice stack — Avoids lock-in to a single voice model provider; lets customers tune cost and quality independently
• Vertical breadth across enterprise customers — Mobility, staffing, insurance, finance — voice AI has clearly moved past pilots in mainstream sectors
• Dual surface — Self-serve developers + enterprise customers share the same platform, building strong top-of-funnel that graduates into enterprise contracts
• Strong investor mix — Peak XV lead with M12 + Kleiner + Bessemer participation creates a board with deep voice-AI and enterprise-distribution experience

Limitations & Considerations

• Voice-only focus — Customers building omnichannel agents (voice + chat + email) need to compose Vapi with other platforms; Vapi does not currently span all channels
• Per-minute economics — Voice agents are billed by call duration, so high-volume use cases need careful unit-economics work to ensure the AI replacement is cheaper than the human alternative
• Enterprise sales cycle — Production deployment for large customers requires a sales-led process; developers should expect a longer evaluation cycle for the enterprise tier
• Competitive intensity — Bland AI, Synthflow, OpenAI Realtime, and Google Gemini Live are all competing aggressively for voice-AI mindshare and customer wins

Best Use Cases

When to choose alternatives:

• Raw model access without orchestration → OpenAI Realtime API or Google Gemini Live
• Premium voice generation only (no agent orchestration) → ElevenLabs
• Omnichannel agent platform across voice + chat + email → composite of voice platform + chat agent platform

Getting Started

1. Visit vapi.ai and create a developer account for self-serve API access
2. Walk through the platform's voice-agent quickstart — typically a hosted demo number you can call within minutes
3. Build a test agent for your specific workflow (support, qualification, scheduling) with custom system prompt and tool-use logic
4. For production deployment at enterprise volumes, contact sales — expect compliance review, custom integration scoping, and a volume-discount negotiation
5. Plan unit economics carefully: voice agents are billed per minute, so verify the per-call cost beats the human-alternative cost on your specific call volume and average handle time

Key Takeaways

• Vapi is a voice-agent platform processing over 1 billion calls, with Amazon Ring as anchor customer (100 percent of inbound calls, chosen over 40 competitors)
• May 2026 Series B raised $50 million at a $500 million valuation, led by Peak XV with Microsoft M12, Kleiner Perkins, and Bessemer participating
• The multi-model voice stack lets customers tune cost and quality across speech-to-text, dialogue LLM, and text-to-speech providers independently — avoiding single-vendor lock-in
• Dual surface (enterprise + self-serve developer) builds strong top-of-funnel that graduates into enterprise contracts, similar to OpenAI's consumer/enterprise split or Twilio's developer/enterprise approach
• Voice-agent infrastructure is no longer a pilot category — Amazon Ring routing 100 percent of inbound through Vapi, alongside Kavak, Instawork, New York Life, and Intuit, signals voice AI has crossed into mainstream enterprise IT budgets