Investing in Flower

We’re delighted to announce our $15M investment in Flower’s Series A. Flower offers a unified approach to federated learning, analytics, and evaluation. This is important for three reasons:

1. There’s a new regulatory imperative to train ML models securely, ensuring PII and sensitive information are handled judiciously. New laws like the EU AI Act will create GDPR-like constraints for enterprises to navigate. For this reason, training and serving models in a federated manner will become more critical. But the value of Flower goes beyond compliance. By training models on, say, cross-border data—which otherwise could violate local data privacy laws—models get better, driving revenue for customers. 
2. There’s more data in the private domain than in public, yet most foundation models aren’t trained on siloed, private data. Flower will help change this.
3. Cloud-hosted inference is increasingly cost-prohibitive. By 2027, there will be 11B installed devices¹ with ML capabilities, and by 2025, 75% of enterprise data² will be generated at the edge, up from 10% in 2018. This push to the edge is exactly the kind of market shift we look for at Felicis. Flower is set up for success because their framework is purpose-built for locally training models on distributed devices.

When we met Nic, Daniel, and Taner, it was clear they’d tapped into something big. Today, the Flower community consists of over 3K developers and hundreds of contributors on the OS framework, which has garnered 3.6K+ GitHub stars and 1.1K+ dependent projects. They’re the leading framework for federated learning, showcasing 10%+ weekly growth on models trained and widespread adoption across Fortune 500s and major universities like Oxford and Harvard. 

With ML, data is everything. Flower breaks down data silos, encoding the distributional information of private data into ML models without exposing any specific data to the model builder. This is a powerful, albeit well-established, idea that will gain popularity in the coming years. We think Flower’s unified developer experience, from cloud to edge, is a winning approach to becoming the standard approach to federated learning.

Federated learning, federated evaluation, and federated analytics require infrastructure to move machine learning models back and forth, train and evaluate them on local data, and then aggregate the updated models. Flower provides the infrastructure to do exactly that in an easy, scalable, and secure way. In short, Flower presents a unified approach to federated learning, analytics, and evaluation, bridging the gap between data privacy and ML. We think this bridge couldn’t come at a better time.

If you want to learn more about Flower and federated learning, join us at the Flower AI Summit in London in March!