Fal.ai, a developer-oriented platform specializing in AI-generated audio, video, and images, announced today that it has secured $23 million in funding from prominent investors, including Andreessen Horowitz (a16z), Black Forest Labs co-founder Robin Rombach, and Perplexity CEO Aravind Srinivas.
The funding is part of a two-round investment, with $14 million coming from a Series A round led by Kindred Ventures, while the remaining $9 million stems from an earlier seed round led by a16z.
Founded in 2021 by Burkay Gur and Gorkem Yurtseven, Fal (short for “Features and Labels”) was born out of the duo’s shared vision during the pandemic. Gur, a former Oracle engineer and machine learning lead at Coinbase, teamed up with Yurtseven, who previously worked as a software developer at Amazon. They recognized a burgeoning demand for robust AI cloud infrastructure, particularly for running generative AI models. “We believed that generative media would transform how all media is consumed,” said Gur. “Our timing was fortuitous, coinciding with significant breakthroughs in AI models just as Fal launched.”
Fal provides two main offerings: a privately managed compute environment for executing models and APIs that support open-source models for creating images, audio, and video. Notably, Fal was among the first platforms to host Black Forest Labs’ Flux model, which powers image generation for Grok, the controversial chatbot by X. While competitors like CoreWeave offer similar resources, Gur asserts that Fal’s scalability sets it apart. “Our platform can manage hundreds of millions of requests, with our inference engine being the most efficient,” he explained. “Fal enables enterprises to seamlessly integrate models into their applications where media plays a critical role.”
Fal's growth trajectory is impressive, despite any claims of scrutiny. It counts major clients, including Perplexity and other enterprise customers in retail and e-commerce. Popular generative AI apps such as Photoroom, Freepik, and PlayHT are also utilizing Fal's services, according to Gur.
Currently, Fal has achieved significant profitability, with its annual run rate nearing $10 million (~$800,000 per month), marking a tenfold increase since January. The Series A funding values the startup at $80 million. “Fal has reached over 500,000 developers on our platform, creating 50 million images, videos, and audio streams daily,” Gur shared.
In light of the rising concerns surrounding deepfakes and misinformation associated with generative technologies, I inquired about Fal's moderation policies for sensitive content. Gur indicated that Fal adopts a largely hands-off approach, allowing the companies deploying models on its platform to make their own decisions regarding safety features. “Most moderation is addressed during training, which we leave to the model developers,” he noted. “Implementing a comprehensive moderation program requires substantial research and resources.”
While Fal does support open-source training efforts under its research grants program, Gur hinted at a desire to enhance in-house moderation efforts in the future, stating, “We plan to expand our internal initiatives and collaborate with specialized vendors for these tasks.”
I also questioned Gur on the issue of intellectual property liability, specifically whether Fal would defend customers against lawsuits if their models inadvertently re-used copyrighted material. He declined to provide a definitive response, but the language in Fal’s terms of service suggests that customers must navigate these risks independently. This differs from generative AI products from established companies like Adobe, Canva, Google, Microsoft, and Shutterstock, which typically include indemnity clauses to protect their users. Some vendors, including Getty Images and startups like Fairly Trained, have specifically chosen to train models using only "commercially safe" content to mitigate copyright risks.
Overall, users of Fal must be aware of the inherent risks involved.
Moving forward, Fal plans to allocate a significant portion of its funding toward enhancing its inference optimization product, making it more accessible for self-service. Additionally, the company is forming a new research team focused on model optimization, expanding its current staff of 17.
Fal's other notable investors include Vercel founder Guillermo Rauch, entrepreneur and investor Balaji Srinivasan, and Hugging Face CTO Julien Chaumond.
Ranking
