AssemblyAI Raised $115M: How Solo Founder Dylan Fox Built a Speech AI API Seven Years Before the Boom

October 19, 2026

TL;DR: AssemblyAI is a speech AI company that sells speech-to-text and speech understanding models through a developer API. It was founded in 2017 by Dylan Fox, a former Cisco machine learning engineer who started it alone and got into Y Combinator a month after quitting his job. As Fox shared on the show, the company raised about $1M from angels after YC, a $5M angel round in summer 2020 when it had passed $1M in ARR with a team of under 10 people, and then an A, a B and a C in quick succession. TechCrunch reported in December 2023 that a $50M Series C led by Accel brought total funding to $115M. Fox told the early story on The Product Market Fit Show. The episode aired in 2024 (Season 3), so every figure he gave was as of then.

A note on sourcing. Anything labeled "as he shared on the show" comes from the interview and has not been independently confirmed. Funding and customer figures labeled with a source and date were checked against press coverage and AssemblyAI's own announcements in October 2026.

What does AssemblyAI do?

AssemblyAI builds AI models that turn audio into text and then into structured information, and sells them to developers through an API. The customer is usually a software company that wants to add transcription or voice features to its own product without training its own models.

That was the plan from day one. Fox described the original goal this way.

"The goal really when we started the company was let's build really advanced new models for speech tasks, speech to text, speaker diarization, speech understanding, and let's make them available through a developer platform that is super easy to use, that has great docs, that you can get started on for free, and that can scale with you if you deploy it at really large scale, even if you're a big enterprise organization, but let's build that platform that anyone can just pick up and use, whether you're a college student or a developer at a Fortune 500." — Dylan Fox, AssemblyAI

Key stat: as he shared on the show, the first speech-to-text model he ever trained used 10,000 hours of audio. The model AssemblyAI was about to release at the time of the interview was trained on 12.5 million hours, on hundreds of TPUs.

As of December 2023, TechCrunch reported about 4,000 paying customers (up 200% from the prior year), more than 200,000 developers on the platform and roughly 25 million API calls per day. As of February 2026, AssemblyAI's changelog lists the release of Universal-3-Pro, a promptable speech-to-text model, so the company has kept shipping new models since the interview. I could not find a current, independently reported customer count, so none is given here.

Who founded AssemblyAI?

AssemblyAI was founded by Dylan Fox, who is the founder and CEO. TechCrunch's December 2023 coverage describes him as a machine learning engineer, and the company dates its start to 2017.

Fox's background, as he told it on the show, was builder-first. He had started a company in college that he called "a terrible company, terrible idea," but it was where he taught himself to program, because he built everything himself. After two years as a contract developer, he moved to San Francisco for a machine learning engineer job at Cisco, working on natural language processing.

The spark was the Amazon Echo. He bought one, tested it from rooms away with the shower, the sink and the TV running, and was struck that it worked. That pushed him toward voice interfaces.

He described the Cisco job as "active rest mode" and said he felt bored because nothing he worked on had impact. When he quit, there was nothing built. "It was just me," he said on the show, and he applied to YC's summer 2017 batch 30 days after the deadline, expecting to be rejected. He got in.

"I believed at the time and still believe that to create the best product is not just a function of resources. You don't just put a budget and people in a bottle and shake and then out comes this amazing product." — Dylan Fox, AssemblyAI

His YC group partner was Daniel Gross, who had seen first-hand how weak speech technology was and how hard it was for companies to get access to it. Gross became a major investor in the company, as Fox shared on the show.

How much has AssemblyAI raised?

The figures below mix what Fox said on the show with press-verified numbers. Sources conflict on exact dates for the early rounds, so this list keeps to what at least one primary source supports.

YC, 2017: $125,000. As Fox shared on the show, that was the standard YC investment at the time.
Seed, after YC: about $1M. As he shared on the show, it came only from angels, with no institutional funds.
2020: $5M from angels. As he shared on the show, the round came in summer 2020, right after the start of COVID, from angels including Daniel Gross and Nat Friedman.
Series A: $28M, led by Accel. AssemblyAI's own July 2022 Series B announcement references the $28M Series A led by Accel. Sources disagree on the exact date, which falls between late 2021 and early 2022, so treat the date as approximate.
Series B, July 2022: $30M, led by Insight Partners with Y Combinator and Accel participating (AssemblyAI announcement and TechCrunch, July 2022). AssemblyAI put total funding at just over $65M after the round.
Series C, December 2023: $50M, led by Accel with Insight Partners, Keith Block (former Salesforce co-CEO), Nat Friedman, Daniel Gross and Y Combinator (TechCrunch, December 2023).

Key stat: TechCrunch reported total funding of $115M as of December 2023, after the Series C.

As of October 2026, a web search found no official announcement of a round after the December 2023 Series C. Some third-party data sites list later rounds and valuations, but they conflict with each other and I could not confirm them with a primary source, so no valuation is given here.

Why did VCs pass, and how did he handle the "Google will do this" question?

Nearly every skeptical investor asked the same thing: how will you compete with Google? Fox heard it from VCs at the end of YC, and he said it was the only reason they passed.

His answer had two parts. First, building the best product is not purely about resources, and more people can slow a project down. Second, speech-to-text was not core to the big tech companies. He relayed a point from a YC partner that you are not competing with Google, you are competing with one product manager at Google who is trying to ship a successful product while navigating internal politics. In his view, the real opportunity was working more closely with developers and customers in an emerging market.

He also made a fundraising mistake he is open about. His very first investor meeting, with no fundraising experience and no real progress, was with Sequoia.

"No experience fundraising, no real progress, don't really know what I'm doing, and just start with the most experienced investors possible, probably not the right strategy." — Dylan Fox, AssemblyAI

The angels backed him instead, because they had seen developer-first companies beat large incumbents before. Since then he has used one habit before every raise.

"I actually write up a big document with all of the objections I anticipate to get. I write out lengthy answers to all of them." — Dylan Fox, AssemblyAI

For more on raising money alone, see fundraising as a solo founder and how to run a fundraising process.

How did AssemblyAI find its first customers and product-market fit?

The first two years were slow, and Fox called them "wandering in the forest." As he shared on the show, each model took one to two weeks to train, so a 90-day YC batch allowed only about six training runs. That is the opposite of the weekly shipping rhythm YC pushes.

Never miss a founder's PMF story

Subscribe to The PMF Show

First customers arrived in late 2019, once the models were good enough to power real use cases. As he shared on the show, the early applications were media monitoring and call centers, with a few voice bot projects.

"That was one of the first customers we had and they actually found out about us on Hacker News." — Dylan Fox, AssemblyAI

That customer analyzed TV and radio stations around the clock for spoken brand mentions and alerted the brands. Its CTO found AssemblyAI on Hacker News, tried the API and built a product on it.

Then the curve bent. As he shared on the show, by summer 2020 AssemblyAI had passed $1M in ARR with a team of fewer than 10 people.

"We had passed a million dollars in ARR. We had decent traction at that time." — Dylan Fox, AssemblyAI

The A, B and C followed over the next few years. Fox credited a mix of better technology, more market adoption as early adopters showed what was possible, and then the generative AI wave from 2022.

"The 2020 is really when things started to compound and the trajectory really changed." — Dylan Fox, AssemblyAI

On product-market fit itself, Fox said something unusual. He knew there was demand before he built anything, because the demand for accurate speech-to-text already existed. The risk was whether the technology was good enough, not whether anyone wanted it.

"For us, I always had conviction that there was market demand and product market fit for what we were building." — Dylan Fox, AssemblyAI

He added that it never stops: AssemblyAI serves many use cases, and the work is improving fit for whatever customers are building most.

What was the mistake Fox would fix?

Fox said his main regret was the first year after YC. YC trains founders to show progress every week. For a team training models, that meant retraining and chasing customers repeatedly instead of setting a clear accuracy target and hitting it before going to market.

His reasoning: demand was known, accuracy benchmarks already existed in the industry, and a research-heavy product is different from a productivity app where you need early feedback to know if anyone cares. He also warned that an early customer can pull a startup in the wrong direction.

Fox also noted that the team stayed at three people for a long time.

Key lessons from Dylan's playbook

1. Know which risk you are de-risking. Fox said demand was not the unknown, technical quality was. When the risk is technology, set explicit accuracy goals and clear them before you chase customers, rather than iterating for weekly traction. 2. Build the developer platform, not the app. Fox picked an API with great docs and a free tier so anyone from a student to a Fortune 500 developer could use it. Platforms let you serve use cases you did not predict, like brand monitoring, call analytics and voice bots. 3. Bet on a category the giants treat as a side project. Fox's response to "what about Google" was that the incumbents' focus was elsewhere. Competing with a team inside a large company is different from competing with its flagship product. 4. Write down the objections. Before each raise, Fox writes a long private document of investor objections and detailed answers. It is a cheap way to be sharp on the competitive and market questions that always come up. 5. Take the long view after an accelerator. Asked what he would tell his 2017 self, he said to slow down after YC, decide where the company should be in two years, and work backward from that.

Listen to the full interview: This solo founder bet on AI 7 years ago, now he has 5,000 customers and $115M raised.

FAQ: AssemblyAI

Q: What is AssemblyAI? A: AssemblyAI is a speech AI company that sells speech-to-text and speech understanding models through a developer API. It was founded in 2017 and counted Spotify, CallRail, Algolia and Fathom among its named customers in its July 2022 announcement.

Q: Who is the CEO of AssemblyAI? A: Dylan Fox is the founder and CEO. As he shared on the show, he started AssemblyAI alone in 2017 after leaving a machine learning engineer job at Cisco, and got into Y Combinator's summer 2017 batch.

Q: How much funding has AssemblyAI raised? A: TechCrunch reported total funding of $115M as of December 2023, after a $50M Series C led by Accel. Earlier rounds include a $28M Series A led by Accel and a $30M Series B led by Insight Partners in July 2022. As of October 2026, no later round could be confirmed with a primary source.

Q: Who are AssemblyAI's investors? A: Publicly named investors include Accel, Insight Partners, Y Combinator, Daniel Gross, Nat Friedman and Keith Block. Fox said on the show that Gross, his YC group partner, became a major investor.

Q: How many customers does AssemblyAI have? A: TechCrunch reported about 4,000 paying customers and more than 200,000 developers in December 2023. I could not find a newer independently reported figure.

Sources: Listen to the Full Founder Story

Dylan Fox, AssemblyAI — a solo founder who left Cisco in 2017, got into Y Combinator, and built a developer-first speech AI API that grew from a three-person team to a company that had raised $115M by late 2023.