The artificial intelligence industry has spent three years demonstrating what large language models can do in text. The harder problem, and arguably the more commercially valuable one, is what they can do on the phone.
Vapi, a two-year-old startup that provides infrastructure for building and deploying AI voice agents, has raised $50 million in a Series B round led by Peak XV Partners at a valuation of roughly $500 million. Microsoft's M12, Kleiner Perkins and Bessemer Venture Partners also participated, bringing total funding to $72 million.
The numbers that matter are not in the fundraise but in the usage. Vapi says it has handled more than one billion calls since launch and currently processes between one million and five million calls per day.
Its largest disclosed customer, Amazon's Ring home security division, now routes 100% of its inbound customer support calls through Vapi's platform, a full replacement of the human and legacy automated systems that previously handled the volume.
Ring's adoption is instructive because of how it happened. During the 2025 holiday season, a surge in customer support calls forced Ring to evaluate its options: expand its human call centre capacity, deploy a traditional interactive voice response (IVR) system, or trial AI voice agents. The company assessed more than 40 vendors before selecting Vapi.
Jason Mitura, Ring's vice president of software development, said customer satisfaction scores improved after the switch and that non-engineering teams were able to tune the AI agent's behaviour directly, a detail that matters because it suggests the technology has reached a level of maturity where business operators, not just developers, can manage it.
The distinction between Vapi and the wave of AI chatbot companies that preceded it is architectural. Vapi does not build the AI models itself. It provides the infrastructure layer, handling the latency-sensitive plumbing of real-time voice: speech-to-text transcription, model inference, text-to-speech synthesis and telephony integration, all within the sub-second response times that telephone conversations demand.
Developers and enterprises plug in their preferred language models and customise the agent's behaviour for specific use cases including customer support, appointment scheduling, lead qualification and outbound sales.
The approach has attracted more than one million developers to Vapi's self-serve platform, alongside enterprise clients including Intuit, New York Life, Kavak and Instawork. Enterprise customers account for the majority of call volume.
Jordan Dearsley and Nikhil Gupta founded the company after they pivoted from an AI therapist project, a shift that reflects a pattern in the current AI landscape. The most commercially successful applications are often not the ones founders originally envisioned but the infrastructure they built along the way.
Dearsley described the core challenge as "taking this indeterminate beast that is a model and taming it," a reference to the difficulty of making language models behave reliably and consistently in real-time voice interactions where there is no opportunity to edit, retry or qualify a response.
The broader significance of the Ring deployment is what it implies about the trajectory of customer service. If the largest home security brand owned by the world's largest e-commerce company is comfortable routing every inbound call through an AI agent, the barrier to adoption for smaller enterprises has effectively collapsed. The question is no longer whether AI voice agents work but how quickly they displace the estimated $400 billion global call centre industry.
Related reading
- OpenAI closes $122bn funding round at $852bn valuation as IPO pressure builds
- MoonPay launches AI trading agent after acquiring Dawn Labs
- OpenAI's $4bn consulting company is an admission that selling AI models is not enough, and a direct challenge…
Vapi has roughly 100 employees and plans to use the new funding to expand its engineering, infrastructure and sales teams. At $500 million, it is valued at a level that implies investors expect it to capture a meaningful share of a market that barely existed two years ago.
Whether that expectation is justified will depend on whether the technology can maintain quality as it scales, and whether customers like Ring continue to report improved satisfaction scores rather than the frustration that has historically defined automated telephone systems.
The recap
- Vapi raised $50 million in a Series B funding round.
- Amazon Ring routes 100% of its inbound calls through Vapi.
- Vapi will expand engineering, infrastructure and go-to-market teams.
