Hugging Face moving from consulting to platform

This shift shows Hugging Face is trying to turn open source popularity into software like revenue, not services like revenue. Consulting pays well but resets every time a project ends. Usage based APIs, cloud marketplace referrals, and hardware sales keep earning as customers keep running models, downloading weights, or buying devices. That makes revenue more predictable and ties growth to activity on the Hub, which is the asset competitors have the hardest time copying.1 2 3 4 5 6

• The old model was heavily services led. Earlier research estimated most of Hugging Face’s roughly $70M annualized revenue in 2023 came from managed enterprise work and consulting for large AI buyers, while more recent company data tracks recurring revenue rising from $35M in 2023 to $150M by September 2026.2 1
• The new model monetizes distribution instead of labor. Microsoft says Azure AI Foundry now offers more than 10,000 Hugging Face models for one click deployment, and Hugging Face says Google Cloud usage grew 10x in three years to tens of petabytes of downloads per month. That means each model hosted on the Hub can throw off usage and partner revenue without a new consulting engagement.3 4 5
• Security and tooling make the self serve product easier to trust inside big companies. Hugging Face says all 2.2M plus public model and dataset repositories are continuously scanned through VirusTotal, and its Hub security program includes SOC 2 Type 2 controls. That matters because enterprises want to click deploy open models without manually auditing every file.6 7

From here, the likely path is a business that looks less like an AI consultancy and more like GitHub plus a cloud marketplace for models, data, and devices. If Hugging Face keeps pushing deployment, security, and partner distribution into the product, more enterprise spend should arrive as recurring infrastructure and platform revenue instead of custom project revenue.2 3 4 5 6 7