fal Announces Series F to Expand AI Inference Capacity

By Amara Osei

fal announced a Series F led by Bessemer Venture Partners on November 3, 2026, while disclosing an earlier Sequoia Capital-led Series E and a $140 million Series D. NVIDIA returned as a strategic investor in Series F, while the announcement did not state the size or valuation of either 2026 round.

fal said the Series F and Series E proceeds will fund work across its ecosystem in 2027. Capital is directed toward two connected needs: inference capacity to run models at scale and employees developing future products. The company said the backing also creates room for research and new forms of creative work, pairing variable inference with dedicated compute so research and production pipelines can grow together.

What will the new capital fund?

fal described customers across entertainment studios, advertising agencies, e-commerce brands, game developers and marketing teams. These customers run image, video and audio models, including fal's own models, within pipelines that produce finished work.

During 2026, fal also post-trained existing models to test whether quality and speed could improve together. The company said its central technical focus is making generative-media inference faster and more cost-effective for scalable applications, a priority that now sits alongside hiring and product development.

fal framed the backing from Bessemer and NVIDIA as a signal of sustained investor confidence in its generative-media infrastructure ambitions. The announcement did not attach a fundraising total, valuation or per-round investment amount to Series E or Series F.

How does Serverless scale customer workloads?

fal's Serverless documentation describes a deployment system for customers' own models, pipelines and applications. fal handles provisioning, scaling, networking and observability, while customers retain control of their code, model weights and container environment.

Serverless runners can scale from zero to thousands as demand rises and return to zero when traffic stops. The company's marketplace Model APIs operate on the same Serverless foundation, bringing its model marketplace onto infrastructure used for customer-controlled deployments.

That architecture addresses traffic that fluctuates as customer pipelines move from testing into production. It gives customers access to variable capacity without requiring fal to take control of their code, weights or container configuration, according to the company's documentation.

What does fal Compute add?

fal Compute supplies dedicated NVIDIA H100 SXM instances for sustained workloads. The instances remain running under customer control and are designed for training, extended fine-tuning, batch processing and research rather than on-demand autoscaling.

Eight-GPU instances can be placed in the same sector and connected over InfiniBand for distributed jobs. fal positions Compute as a different operating model from fluctuating inference traffic, giving research and long-running customer workloads continuous access to dedicated hardware.

How do the funding rounds connect?

fal's December 9, 2025 Series D announcement said the round raised $140 million. Sequoia, Kleiner Perkins and NVIDIA joined as new investors, and fal said the capital would scale its platform globally.

fal also disclosed that Sequoia Capital led a Series E that closed earlier in 2026. NVIDIA returned as a strategic investor in Series F, extending its participation from Series D through two additional financing stages disclosed in 2026.

At the time of Series D, fal reported 70 employees and was recruiting across engineering, product, design, go-to-market and operations. The Series F announcement again links team investment with infrastructure, this time explicitly adding research to the systems serving customers.

Who is behind fal?

According to fal, Burkay Gur and Gorkem Yurtseven founded the company in 2021 after seeing AI-infrastructure challenges during their time at Coinbase and Amazon. fal is headquartered in San Francisco and has a globally distributed team.

fal framed the 2027 challenge as expanding two infrastructure demands at once: immediate, variable inference traffic and sustained computing for research. The outcome will show whether the company can support current production pipelines while giving researchers and product developers enough capacity to advance.