AI/ML Platforms

This week, Fabrica Ventures’ Investment Committee discussed the prospect of investing in an AI/ML platform.

AI/ML platforms offer end-to-end AI life cycle development, management and deployment of AI models and applications. They typically incorporate tools for data preparation, model training, and deployment, as well as pre-built industry-specific AI applications, and also provide language, vision and AutoML software that can run in the cloud or on-premises.

Certainly, every hyperscaler provides its own AI/ML platform, but their primary focus lies in promoting cloud computing services. Additionally, there is a concern about the possibility of sensitive company developments “inadvertently” becoming part of hyperscalers’ LLMs.

The primary public players, C3.ai and Palantir, target (very) large enterprises, leaving an opening for startups to enter the AI/ML platform market. Notably, DataRobot, Dataiku, and H2O.ai emerge as the most prominent private contenders.

* DataRobot: Once the leading name in the sector, secured $1.1B from NEA, Sapphire, Altimeter, Salesforce and other renowned VCs, reaching a valuation of $6.3B. Regrettably, the company has been plagued by numerous management scandals, experiencing turnover of three CEOs within an 18-month span, that it has become radioactive.

* Dataiku: Stands out as the most user-friendly platform, eliminating the necessity for users to be data scientists, thus widening accessibility to the middle market. Although the startup generated $220M in revenue in 2023, its current valuation in the secondary market, trading at 15 times its revenue, is expensive for us.

* H2O.ai: It’s the preferred platform among data scientists, being fully open source. Notably, it has assembled the largest group (26) of Kaggle Grandmasters (the highest level of performance in data science) within a single company. Main investors are CVCs of large financial institutions — CBA, Wells Fargo, Goldman Sachs, New York Life, Barclays, and Capital One — and the startup has established a strong presence in the financial services sector. However, its growth rate has been lagging and the recent launch of a SLM (Small Language Model) made us think twice about the soundness of the company’s strategic direction.

Conclusion

By definition, a discovery process is made of discoveries.

We discovered that AI/ML platforms are not for us (at least for the moment).