Bessemer Venture Partners led fal.ai’s Series F to fuel the inference layer and generative media stack, according to the article. It says generative media is becoming one of AI’s largest markets, with production costs collapsing and volume exploding as model capabilities improve, and that fal has evolved in two years from an inference engine to the full stack: workflow tools, an API reaching hundreds of models, custom infrastructure for enterprises’ own weights, and post-trained model development alongside labs. The article states media is a different infrastructure problem than text because image and video models refine a whole frame over many passes, requiring memory, compute, and latency to be handled differently, and that fal has spent years on this problem down to the kernel, producing infrastructure for real-time generation at scale with low latency, low cost, and high throughput. It describes fal as a two-sided ecosystem where developers access thousands of models through a single API and labs leverage fal for distribution and serverless infrastructure, and says companies including Adobe, Canva, Shopify, World Labs, and Amazon have moved production volume to fal. It adds that in August fal Research released H3 Max, a post-trained version of MiniMax H3, which immediately topped all major arenas and leaderboards, and that last week fal Research partnered with Black Forest Labs on Flux 3 Max, with enterprises able to commission post-trained versions of models exclusively on fal.