When @beckylitv tweeted "can you tell this video is AI?” and sent Romain Torres’ company from $5,000 to $64,000 monthly recurring revenue, I’d like to imagine that Torres’ watched the spike without celebrating. He’s spent the past two years building the kinds of apps that needed something like this, and he knew that the product would eventually catch the eyes of the mainstream, because all it takes is one poster and the favor of the algorithm.
The Timing No One Saw
Arcads.ai does just one thing: you make a script and pick an actor from a 1,000+ AI library, and get a video ad in just two minutes. The format in question is user-generated content, which has taken over short-form video platforms like TikTok and Instagram Reels due to their authenticity and low barrier to entry. It’s become the most popular choice for social advertising, so marketers need a constant supply of them without hitting writer’s block. Testing around 20-30 variants of an ad has become an industry norm as a result.
HeyGen and its London-based rival, Synthesia, were both much larger startups that did similar things. However, they both optimized their platform for studio quality and enterprise clients in hopes of securing larger contracts. Neither of them built something that looked and felt as authentic as a user who’s filming a product review on the couch because they genuinely want others to try it.
Why They Got Funded
Eurazeo, a French private equity and venture capital company, led the $16 Million Seed round in December of 2025. It wasn’t a hard decision since the company was already profitable, but its investor, Thomas Turelier, gave a public thesis for his decision. He believes that the next key battle in marketin gis how it's distributed, and at what volume companies could distribute at.
What's Disruptive. What Isn't.
Arcads uses an impressive, fully autonomous acquisition loop through an AI agent that scrapes ads from competitors. The ad data gets sent to OpenAI’s LLM to analyze what’s working, and new Arcad videos are created using similar hooks. With a small team of around 8 employees and an annual recurring revenue of $10 Million, Arcads can provide compounding value in ways that a larger, and naturally slower, sales team cannot.
However, since Arcads doesn’t own its generation model, it’s heavily reliant on third-party infrastructure and has less control over quality and pricing as a result.
The Go-To-Market
Arcads employs a subscription plan as well as a generous 25% referral program. By way of illustration: if you’re a marketing agency using Arcads and you refer a friend’s brand, and he pays for the $220 per month subscription plan, you’re getting $55 (which is 25% of $220) for the next 12 months. On top of this, there’s Romain’s LinkedIn following and an AI sales agent that looks for potential clients. Let not the irony be lost with you: yes, the company that makes millions from marketers doesn’t use humans for their own marketing.
The Blind Spot
Two risks.
Infrastructure poses the most obvious one: if HeyGen, or whoever Arcads is sourcing its API from, decides to reprice or restrict third-party access, Arcads loses its core technology overnight.
Perhaps a more subtle one is their Ideal Customer Profile. Arcads’ testing mechanism includes the generation of 10-30 hooks that are all run as paid ads at a small test budget. Meta and TikTok’s algorithm picks the winner, and the client scales that one. However, by attracting smaller companies and startups through their pricing instead of going all in on enterprises like their larger competitors are doing, Arcads is trying to serve a demographic that often doesn’t have the media budget to test these ads out. Additionally, its model of copying hooks from competitors doesn’t serve the startup model well if the startup makes a disruptive product that goes outside of the existing category.
The Question
Romain, are you building toward model ownership or betting the distribution advantage becomes durable before the infrastructure risk materializes?
The Takeaway
Arcads is proof that you can build a $10M ARR business targeting the wrong customer. Direct-to-consumer brands that have at least $1,000/month to test out ads would get the most out of Arcads’ services, and startup founders making blue oceans and chasing new categories are probably somewhere in the streets of San Francisco doing something that doesn’t scale instead of using Arcad. The gap hasn’t cost Arcad anything yet, but it does set a ceiling for what they’re able to achieve, especially against their larger competitors.
