Rime raises $24M Series A led by M13 for enterprise voice AI

A modest $24 million round for a voice AI startup would normally be a footnote in a week of nine-figure raises, but Rime's Series A is worth reading because of where in the stack it lands. TechCrunch reports the San Francisco company raised $24 million led by M13, with Twilio Ventures, Corazon Capital, Unusual Ventures and existing backers joining, on top of a $5.5 million seed round from last May.

Rime is a model shop. It records conversational training data in its own studio rather than scraping the web, and it uses what the company describes as a phoneme-based architecture so customers do not have to retrain models on industry-specific terminology. Co-founder Lily Clifford's pitch is bracingly candid. 'The voice technology is still not there to automate the vast majority of enterprise phone calls,' she told TechCrunch, adding that talking to a voice AI agent today 'is kinda like a new IVR, but with a better voice.' Rime's answer is to push the underlying model, and it just hired Rafael Valle from Meta Superintelligence Labs and NVIDIA's audio research team as chief scientist to lead a shift toward lower-latency speech-to-speech models.

Why this matters for anyone running a contact center: the voice AI market is stratifying. ElevenLabs and others are moving up into orchestration and applications, going head to head with Decagon and Sierra. Rime's bet, echoed by M13's Morgan Blumberg who is joining the board, is that there is still a lot of technical work to do at the model layer, especially in regulated environments where latency and reliability actually decide the deal. The company says it already handles over 100 million calls a month for customers including Mayo Clinic, Dialpad, Upstart and Asurion, across food service, healthcare, airlines and fintech.

The honest caveats are the obvious ones. The 100 million calls figure is Rime's own, and the reporting does not include revenue, ARR, or the round's valuation. A shift from stitched speech-to-text plus LLM plus text-to-speech pipelines to a single speech-to-speech model also tends to buy latency at the cost of controllability, which cuts awkwardly against regulated buyers who need auditability. What the article does not give you is how those 100 million calls break down across customers, or how much of that volume actually runs on Rime's own end-to-end model rather than a hybrid deployment.

Still, the interesting signal is that enterprise voice is now big enough and specialized enough to sustain a pure model company alongside the orchestrators, and that Twilio Ventures wrote a check. If you buy voice AI for a regulated vertical, the layer you are shopping in, whether model, orchestration, or full-stack agent, is now the first decision you have to make.