Deepgram Hits Unicorn Status at $1.3bn With $130m Raise as Voice AI Goes Mainstream

The quiet but steady rise of voice-based artificial intelligence is beginning to reshape how companies sell, support customers, and interact with consumers. What was once dismissed as clunky call-center automation or unreliable voice assistants is now emerging as a serious enterprise tool, and investors are responding accordingly.

That shift is at the heart of Deepgram’s latest milestone.

The voice AI company said it has raised $130 million in a Series C funding round led by AVP, pushing its valuation to $1.3 billion and cementing its status as one of the sector’s newest unicorns. Existing backers Alkeon, In-Q-Tel, Madrona, Tiger Global, Wing, and Y Combinator also participated, while new investors, including Alumni Ventures, Columbia University, Princeville Capital, Twilio, and SAP, joined the cap table. The deal brings Deepgram’s total funding to more than $215 million.

The size of the round and the company’s valuation underline how sharply sentiment around voice AI has changed. Once viewed as a narrow use case plagued by errors and poor user experiences, voice technology is now being embedded into revenue-generating and cost-sensitive parts of enterprise operations, from sales development to customer support and internal productivity tools.

That demand is what first drew AVP to the company. Elizabeth de Saint-Aignan, a partner at the firm, said conversations with enterprises in 2024 repeatedly surfaced voice AI as a practical application of artificial intelligence rather than a speculative one.

“When we were talking to enterprises about how they were thinking about using AI inside their business, we started to hear about them using voice AI in processes like contact centers and sales development,” she said.

As AVP dug deeper, it found that many of those systems were powered by Deepgram’s models and APIs.

For enterprises, the appeal is twofold. Voice AI promises to make interactions with customers faster and less frustrating while also reducing operational costs tied to human agents. Investors see Deepgram as a company selling the infrastructure layer for that shift, rather than a single end-product that could be easily replaced.

Founded by CEO Scott Stephenson, Deepgram builds speech-to-text and text-to-speech models, along with APIs designed for real-time conversational AI, interruption handling, and low-latency responses. The company positions itself as a developer-first platform, providing building blocks that other startups and enterprises can integrate into their own products. It says more than 1,300 organizations now use its technology, including meeting notetaker Granola, voice agent startup Vapi, and communications heavyweight Twilio.

One detail that stands out in the current AI funding climate is that Deepgram did not raise out of necessity. Stephenson said the company was cash-flow positive last year, a claim that sets it apart in an industry where many AI startups are still burning large amounts of capital to train models and acquire customers.

“In the last year, voice AI has gone mainstream, and there is more potential pull,” Stephenson said. “We see that we can make larger investment sooner in order to accelerate growth. And that is why we felt it could be a good time to raise.”

He added that Deepgram was not actively shopping for funding. Instead, interest came to the company, allowing it to be selective. According to Stephenson, the focus was on bringing in strategic investors who understand both the technical complexity of voice AI and the industries building on top of it.

The timing of the raise also reflects a broader wave of capital flowing into the voice AI ecosystem. Over the past year, several companies in the space have closed large rounds, including Sesame’s $250 million Series B, ElevenLabs’ $180 million Series C, and Gradium’s $70 million seed round. Taken together, these deals point to a growing belief that voice will be one of the dominant interfaces for AI, particularly in environments where typing is slow, impractical, or unnatural.

Deepgram plans to use the new funding to expand internationally and improve support for multiple languages, a critical step as global companies look to deploy voice systems beyond English-speaking markets. The company is also making a more aggressive push into the restaurant industry, an area that has long attracted voice AI experiments but has delivered mixed results.

To strengthen that effort, Deepgram recently acquired Y Combinator-backed startup OfOne, which built a voice AI solution for quick-service restaurants. OfOne claims its system achieves more than 93% accuracy in taking orders, a metric that addresses one of the biggest pain points in restaurant automation.

Previous attempts across the industry have often fallen short. Taco Bell, for example, pulled back from its voice AI pilot last year after a highly publicized incident in which a system processed an order for 18,000 cups of water.

Stephenson argues that food ordering could become a turning point for how consumers perceive voice AI more broadly.

“I am excited about this because food ordering might be the first positive interaction more than 300 million Americans have with voice AI,” he said. “There have been a lot of sour interactions with voice AI over the last 20 years, where assistants came out and people felt they didn’t provide a magical experience. But when you can order your food using natural conversation, people would think the technology is ready.”

Investor activity suggests that optimism around restaurant-focused voice AI is spreading. OfOne’s acquisition follows recent funding for Presto, a company that provides voice and automation tools to chains such as Carl’s Jr., which raised $10 million in new capital.

Behind these individual deals is a larger market opportunity. Analysts estimate that the voice AI market is growing at more than 30% year over year and could reach between $14 billion and $20 billion by 2030. At that scale, companies supplying core models and APIs stand to become deeply embedded in enterprise technology stacks, much as cloud providers did in the previous decade.