Week in voice AI #5
Rise in enterprise
I am Ivan Mehta, a consumer tech reporter at TechCrunch. For more than a year, I have covered different aspects of voice AI. This newsletter is an experimental attempt at covering what is happening in this industry, which is growing at a rapid pace.
Top News
Deepgram raises $130 million as demand for voice models grows
It was a good week for companies working with enterprises around voice AI. Deepgram became the first voice AI unicorn of 2026 with $130 million in funding led by AVP at a $1.3 billion valuation. Notably, Twilio, which is a partner to Deepgram, is also investing in the round.
In terms of numbers, the company is serving more than 1,400 organizations and 200,000 developers with its models. The company’s CEO, Scott Stephenson, told me that while customer experience and customer success are key focus areas, healthcare is a growing area for the company.
Notably, the company also said that India is the second biggest market by ARR and Hindi is the second biggest language it caters to. This is after ElevenLabs said last year that India is its second biggest market in terms of enterprise revenue last year.
The company also wants to expand its Saga platform, which lets developers integrate voice with their apps, to a wider audience beyond app builds.
ElevenLabs crosses $330 million in ARR, eyes an $11 billion valuation
Another voice AI model company is progressing nicely. ElevenLabs CEO Mati Staniszewski said that the company had crossed $330 million in ARR by the end of last year.
Staniszewski noted that the company took 20 months to cross $100 million in ARR, 10 months to cross $200 million, and just five months to reach the current figure of $330 millon.
What’s more, he indirectly acknowledged that voice AI solutions could have an impact on call center business and jobs. Something a16z partner Oliva Moore also noted in her outlook for voice AI for 2026.
ElevenLabs seems to be steaming ahead, though. The company has announced new partnerships with companies like Deutsche Telekom. Plus, it’s on track for a $11 billion valuation according to the Financial Times. The company had raised money at $3.3 billion valuation last January. It then raised its valuation to $6.6 billion later in the year. Essentially, in a year, the startup has more than tripled its valuation.
Restaurants want voice AI for ordering: Deepgram acquires OfOne, Presto raises $10M
Besides raising a new round of funding, Deepgram acquired Y-Combinator-backed OfOne, which allows quick service restaurants to use voice AI to capture orders. Deepgram CEO Stephenson said that the solution has over 93% accuracy in correctly identifying orders.
Food-focused startups within the voice AI segment were on a hot streak last week as Presto, another YC startup, raised $10 million led by Metropolitan Partners Group, with participation from Remus Capital, Link Ventures, and ElevenLab CEO Mati Staniszewski.
In the last few years, there have been a few instances of voice AI snafus that had restaurant chains rethink their AI strategy. Despite this, restaurants like McDonald’s have been open to trying out new providers and technology.
Quick bytes
It was a good week for enterprise voice AI providers as Flip raised $20 million from Next Coast Ventures and Ridge Ventures. The company also crossed the milestone of handling over 300 million calls.
Matthew McConaughey trademarked his voice, including his famous “alright, alright, alright” phrase, to prevent AI misuse. The actor has a stake in ElevenLabs and lets the company use his voice for certain projects.
Instagram added support for dubbing Reels in six new Indian languages, including Bengali, Tamil, Telugu, Kannada, Marathi, and Hindi.
Signals & Experiments
I’ve been using the new Plaud NotePin S quite frequently when I meet sources or attend meetups. For conferences or formal interviews, the Plaud Note Pro or the Comulytic Note Pro is more suitable.
With the NotePin, I have recorded a single conversation of more than two hours with a friend (with consent), and the notes largely came out accurate. Since we were in a restaurant, I didn’t expect the transcript to be spotless. But I was able to look at startups my friend mentioned, or numbers they spoke about, and that was pretty useful.
During that conversation, potential misuse of voice AI came up. While researching that topic, I came across this story from The Guardian that highlights that extremists are using voice tools to recreate inciting materials with the voices of problematic leaders like Adolf Hitler. While voice AI and translation make it easier for content to spread, some tools lack the ability to determine what kind of content it is, and those guardrails will have to be built within models and applications.
Thank you for tuning in. Keep listening.
Email: voiceaiweek@gmail.com or im@ivanmehta.com