Artificial intelligence startup fal today announced that it has raised $140 million through its third funding round of the year.

Sequoia led the Series D investment. It was joined by the venture capital arms of Nvidia Corp., Salesforce Inc. and Shopify Inc. along with more than a half dozen others.

Following its previous funding round in July, fal disclosed that its revenue had quadrupled during the first six months of 2025. The company, which is incorporated as Features & Labels Inc., stated today that its run rate has grown by another 300% since the July raise. It now has a customer base of several hundred enterprises.

The company’s namesake cloud platform provides access to more than 600 AI models optimized to generate images, audio and video. The lineup includes OpenAI Group PBC’s Sora 2 and the Nano Banana Pro image generator that Google LLC released last month. They’re available alongside a long list of open-source algorithms.

Developers can access fal’s models through a single application programming interface. According to the company, that arrangement removes the need to learn the technical nuances of each algorithm and makes it easier to switch models. Moving an application from one fal-hosted model to another doesn’t require extensive code changes because the API stays the same.

According to fal, its platform also provides other benefits. It skips the manual setup process usually involved in deploying cloud-based AI models and removes the need to fine-tune them.

Developers whose requirements are not met by the algorithms in fal’s catalog can run custom models on a service called fal Serverless. According to the company, it can launch a cluster with thousands of graphics processing units near-instantaneously. That enables customer workloads to scale quickly in response to sudden usage spikes.

According to fal, fab Serverless provides access to Nvidia Corp.’s H200 and and H100 graphics cards. It also offers an observability tool that enables customers to collect telemetry about their workloads. The software tracks metrics such as number of requests received by a model and the latency with which it processes those requests.

Companies with more advanced requirements can use another fal service, fal Compute, to access dedicated GPU clusters. The offering removes the need to share hardware with other customers. It also provides a longer list of GPU customization options than fal Serverless, which enables researchers to apply more extensive performance optimizations. 

In addition to providing infrastructure and third-party AI models, fal develops its own neural networks. One of the newest additions to its AI lineup is an image generator called AuraFlow v0.3. The model is based on a flow matching design, an alternative to the ubiquitous diffusion architecture that provides faster inference times.

“Developers and enterprises are building entirely new categories of applications around real-time, personalized, generative content,” said fal co-founder and Chief Executive Officer Burkay Gur. “fal is the infrastructure that makes that possible at global scale.”

The company will use the proceeds from its latest funding round to enhance its feature set. As part of the effort, fal plans to launch new products and extend its cloud infrastructure to more locations. It will also hire more go-to-market professionals.

Image: Unsplash