I’ll be pulling out a few points from the post https://blog.serenacapital.com/the-open-source-payoff-5e835c54c0f1 and the report http://cossreport.com/
Disclaimer:
In this report, “Open Source” is defined broadly. It covers companies whose products rely on publicly accessible source code — whether under an OSI-approved license, open weights in AI, or products with a meaningful open-source component alongside proprietary features. This reflects the evolving reality of COSS models in today’s software landscape.
This is practically open-washing. The report, for example, categorizes xAI and databricks as COSS. While some aspects of their work are available under a FOSS license, categorizing them under COSS is a stretch.
TL; DR of the report
This report delivers a clear, data-driven conclusion: Open Source is not only financially viable — it is a superior financial strategy, particularly for infrastructure software.
Our research shows that COSS companies raise capital faster at higher valuations and deliver superior liquidity outcomes, with better returns for shareholders compared to closed-source peers.
Regarding study methology
The analysis integrates quantitative financing and exit data (from PitchBook) with open-source activity data (from GitHub), giving us a combined view of financial performance and community traction
Every time I see “open-source activity data (from GitHub)”, I get slightly excited, because I presume that they look at community code contributions, issues raised by the community, etc. Sadly, it looks like the only GitHub metric that they used in the study is “GitHub stars” and whether or not the COSS company was registered before or after the repository was created. But on the bright side, they also point out that
- No Premium from Community traction: GitHub stars show near-zero correlation with round size (R² = 0.001) and valuation (R² = 0.012)
Looking at what the companies do and where funds are being invested
- Strong Infrastructure Focus: Around 90% of COSS companies operate in infrastructure software rather than business applications. Within this, funding concentrates heavily in Core Infrastructure (20%) and Data (20%), with the rest distributed across AI, DevTools, Security, and Blockchain.
- Certain categories show a strong natural bias toward Open Source, particularly DevTools (5.9x more likely to be COSS) and Core Infrastructure / DevOps (5.2x). Investors and founders alike see Open Source as the default go-to-market model in these segments.
Regarding community traction of COSS projects
- Numerous Successful Raises with Minimal Community: Many infrastructure COSS companies secured large rounds with low or even no significant GitHub traction at the time of fundraising (up to 35% at Seed)
This seems to be in line with what the broader FOSS community perceives - and IMO “securing large rounds with low or no significant GitHub traction” sounds like a bad thing and not a good thing.
AI
One of the disappointing graphs of the report is on Slide 14 - The share of unrestricted “open weights” AI models has decreased since its peak in 2020, with a majority of AI models now being “open weights” with restricted use or “open weights” with non-commercial use. Data from Open vs. Closed AI: How behind are open models? | Epoch AI
Epilogue - I’m not sure how to feel right now. Given the sustainability problems of FOSS and the value of FOSS for enterprises, it feels like COSS is here to stay, and so is the VC funding in the COSS space. But the mischaracterization of companies that release non-commercial open-weights models as COSS is open-washing and feels like a betrayal of the actual COSS companies e.g. “products with a meaningful open-source component alongside proprietary features”