Stephenson told Slator that Deepgram’s customers are using its models to build live captioning, note-taking, subtitling, and consumer-facing voice assistants, and that a core goal is to accelerate the company’s expansion to more than 100 fully supported languages and dialects.
In addition to language expansion, Stephenson told Slator that the funding will be used to expand Deepgram’s core models “in both perception and generation, so better speech recognition, richer understanding, and more expressive, low-latency voice generation across more domains and languages.”
The company also plans to “build out the global real-time infrastructure that Voice AI requires, which means more regions, more edge locations, and the compute capacity to deliver sub-second experiences for users anywhere in the world.”
Commenting on Deepgram’s fine-tuned models for regulated industries, Stephenson told Slator that “regulated industries have been some of the fastest growing segments for us, precisely because their constraints are so tough. In healthcare and financial services, for example, accuracy, auditability, and deployment options really matter, so our ability to fine-tune models on domain-specific language and deploy in virtual private clouds, specific regions, or even on premises is a big differentiator.”
“Our low latency, code switching capabilities, and growing language coverage make live conversational translation a very natural extension of what we already do.” — Scott Stephenson, CEO, Deepgram
The CEO added, “we see strong uptake in use cases like clinical documentation, patient and member support, trading and research communications, and compliance monitoring, where voice data is both highly sensitive and highly valuable. These customers are not experimenting in a lab; they are putting AI into production under strict regulatory regimes, which is why we invest heavily in security, data handling, and clear controls over how models are trained and used. The friction does not go away, but our platform is designed to help them navigate it without sacrificing performance or real-time responsiveness.”
The company was founded in 2015, and told Slator that conversations with investors “were very different from what you typically see in an early-stage AI raise,” given Deepgram’s existing enterprise customer base, which includes AWS Connect and Cloudflare.
According to Stephenson, “Deepgram is the foundational infrastructure layer for the rapidly emerging ‘voice AI economy,’ in the same way Stripe became the foundational layer for digital payments.”
“We see strong uptake in use cases like clinical documentation, patient and member support, trading and research communications, and compliance monitoring, where voice data is both highly sensitive and highly valuable.” — Scott Stephenson, CEO, Deepgram
“Ultimately, investors saw that Deepgram is becoming the backbone of a trillion-dollar voice AI economy – powering everything from call centers and enterprise workflows to highly complex consumer environments like drive-thrus and retail. That combination of technical depth, infrastructure maturity, and proven market adoption is what made this round so strong, even against a backdrop of heavy AI hype,” he concluded.
The Series C round was led by AVP, an independent global investment platform, and included participation from Deepgram’s existing investors Alkeon, In-Q-Tel, Madrona, Tiger, Wing, Y Combinator, and funds and accounts managed by BlackRock.
The round also attracted new investors, Alumni Ventures, Princeville Capital, the University of Michigan, and Columbia University.
