Deepgram announced Thursday that it's acquiring Daily, the company behind Pipecat, one of the most widely used open source frameworks for voice AI infrastructure. The acquisition comes as voice AI companies race to strengthen their positions through deals and partnerships, with both firms committing to maintain Pipecat as open source indefinitely. Neither company disclosed the purchase price.
Pipecat is a framework that handles voice and video streams while coordinating speech-to-text engines, large language models, and text-to-speech systems with reduced latency, according to the announcement. Daily, which provides hosted infrastructure for voice and video calls, has pulled in more than $62 million in funding from backers including Tiger Global, Highfield Capital, Nat Friedman, Lachy Groom, and Kevin Weil. Deepgram achieved unicorn status earlier this year with a $130 million Series C round. Both companies went through Y Combinator's 2016 batch together.
Daily co-founder Kwindla Kramer told TechCrunch that despite receiving multiple acquisition offers, he chose Deepgram because CEO Scott Stephenson committed to keeping Pipecat open source permanently and the engineering cultures aligned. "Scott was super committed to Pipecat being open source and vendor-neutral forever," Kramer said. Stephenson explained that Deepgram wants to open up the ecosystem so developers can build voice agents using whatever models work best for their needs, not just Deepgram's products. Kramer noted that Daily began experimenting with AI-to-human communication over voice and text around the time GPT-4 launched, which led to Pipecat's creation. He emphasized that the team "never thought about Pipecat as a voice toolkit" but rather as infrastructure for building AI-native software with real-time capabilities.
The deal positions Deepgram to gain deeper expertise in real-time voice and video infrastructure while expanding beyond voice applications, since Pipecat handles real-time communication across multiple formats. Stephenson said the acquisition enables developers to construct agents with expressivity and complete speech systems by connecting different models through a context bus that links all components. Deepgram is also rolling out the Real-Time Context Standard, an open protocol for voice agents developed jointly with the Pipecat team that specifies metadata from audio—like whispers—instructions for language models on interpreting those signals, and guidance on response tone. The move reflects broader investment momentum in voice AI, with model creators like ElevenLabs, Gradium, and Fish Audio raising millions while infrastructure providers such as LiveKit and Vapi secured multimillion-dollar funding rounds this year.
The open source commitment may prove strategically essential as voice AI infrastructure fragments across competing proprietary platforms. Maintaining vendor neutrality could determine whether Deepgram becomes a standard orchestration layer or just another walled garden in an increasingly crowded field.