Deepgram acquires Daily, will keep Pipecat open source as voice AI consolidates

Deepgram has acquired Daily, the company behind Pipecat, a widely used open-source framework for building voice agents, TechCrunch reported on Thursday. The companies did not disclose financial terms, but confirmed that Pipecat will remain open source under its new owner.

What each side brings

Deepgram, a speech AI company that reached unicorn status earlier this year after raising a $130 million Series C round, is buying a firm with two assets: hosted infrastructure for real-time voice and video calls, and stewardship of Pipecat, one of the most popular open-source projects in the voice AI space.

Pipecat handles the plumbing of a voice agent. It coordinates audio and video streams and connects the components of a typical pipeline — speech-to-text, large language models and text-to-speech — while keeping latency in check. For Deepgram, which has largely operated at the model layer, owning that orchestration layer gives it a presence across more of the stack developers actually assemble.

Daily had raised more than $62 million from investors including Tiger Global, Highfield Capital, Nat Friedman, Lachy Groom and Kevin Weil, according to PitchBook data cited by TechCrunch. The two companies also share a history: both went through the same Y Combinator batch in 2016.

The open-source commitment

According to TechCrunch, Daily co-founder Kwindla Kramer said there had been plenty of acquisition interest, but Deepgram CEO Scott Stephenson's pledge to keep Pipecat open source and vendor-neutral indefinitely was a deciding factor. Kramer argued that openness is also commercially sensible: a neutral orchestration layer at the center of a fast-growing ecosystem benefits when as many participants as possible can plug into it.

Stephenson told TechCrunch that Deepgram had wanted to work in open source for some time and identified agent orchestration as the natural entry point. He framed the deal as a way to let developers mix and match models regardless of vendor — using Deepgram's where they fit and competitors' where they are stronger — with Deepgram working to close gaps in its own products.

Kramer also described Pipecat's origins: after GPT-4 arrived, Daily began experimenting with AI-to-human communication over voice and text, and that work grew into the framework. He said the team never viewed Pipecat as narrowly a voice toolkit, but as tooling for real-time, AI-native software — which suggests Deepgram could apply it beyond audio.

A new protocol ships alongside the deal

Together with the acquisition, Deepgram announced the Real-Time Context Standard (RCS), an open protocol for voice agents co-developed with the Pipecat team. As described by TechCrunch, RCS defines metadata drawn from audio — a whisper, for instance — along with instructions for how a language model may use those signals and guidance on how the response should sound. The goal is a shared way for agents to interpret and act on the expressive qualities of speech, not just its words.

A sector in consolidation

The deal is the latest in a string of moves across voice AI. Model developers such as ElevenLabs, Gradium and Fish Audio have raised substantial sums, while infrastructure providers including LiveKit and Vapi have landed multi-million-dollar rounds this year, TechCrunch notes. As business models in the space mature, established players are increasingly locking in their positions through partnerships and acquisitions rather than building everything alone.

Why it matters

The acquisition concentrates a significant piece of voice AI infrastructure — an orchestration framework many developers already build on — under a major model provider. Deepgram's promise to keep Pipecat vendor-neutral will be tested over time, but if honored it could reassure a developer community that has watched open-source projects absorbed into commercial stacks before. The launch of RCS also signals where the sector is heading: beyond transcription and generation, toward agents that understand tone, emphasis and delivery in real time. For teams building voice products, a better-funded Pipecat paired with a standardized context protocol could simplify the fragmented work of stitching speech-to-text, LLM and text-to-speech components together — whichever vendors they choose.