What fal does
fal is a generative-media inference platform purpose-built for image, video, audio, and 3D models. Where general-purpose inference clouds focus on text LLMs, fal's stack is optimized end-to-end for the demanding compute and latency profile of generative-media models — running open and proprietary models like FLUX, Stable Diffusion, Veo, Kling, Runway, and many others on tuned GPU infrastructure with sub-second cold starts and aggressive performance optimization.
Developers integrate fal via a single REST and SDK surface, and the company runs the fastest hosted versions of major generative-media models alongside fine-tuning, LoRA hosting, and serverless workflows. fal hit $100M annualized revenue in September 2025 with just 92 employees, making it one of the fastest-scaling AI infrastructure companies in the cycle.
Who it's for
fal is for developers, AI-native product teams, and creator platforms building generative-media applications — image generators, video tools, AI marketing platforms, gaming companies, and consumer apps that need production-grade media inference without managing GPUs.
Pricing
fal is usage-based, priced per second of compute or per inference call depending on the model. There is a free tier for evaluation; production usage is metered with volume discounts at scale.
Team & funding
fal was founded in 2021 by Burkay Gur and Gorkem Yurtseven (CEO), both Turkish-born software engineers based in San Francisco. The company has raised approximately $220M total across a $9M seed (Andreessen Horowitz), $14M Series A (Kindred Ventures), $49M Series B (Notable Capital, a16z, February 2025), and a $140M round in December 2025 led by Sequoia at a $4.5B valuation with Nvidia's Nventures, Kleiner Perkins, and Alkeon. Reports in early 2026 indicate a follow-on round in talks at $300M+ at an $8B valuation.
Position vs competitors
fal competes with Replicate on hosted-model inference, Together AI and Fireworks on infrastructure, and hyperscaler GPU offerings on raw compute. Its differentiator is the deep specialization in generative media plus extreme performance tuning that closes the gap between research models and production apps.