Peter H. Diamandis
August 20, 2026
Open-Sourcing Gemini: The Strategy Google Should Have Adopted
TL;DR
A former Google advisor argues that open-sourcing Gemini three years ago would have been Google's winning move to dominate AI, similar to Netscape's transition to Firefox, and could still be transformative if tied to TPU and GCP infrastructure.
Key Takeaways
- 1The speaker wrote a letter to Google management three years ago recommending they open source Gemini to win the AI market
- 2Demis Hassabis publicly claimed Google lacked sufficient compute resources to open-source Gemini, which is why only Gemma was made available
- 3Open-sourcing Gemini tied to Google's TPU and GCP compute would create an empire on compute infrastructure, similar to how Netscape evolved into Firefox
- 4An open-source Gemini could provide America with a top-tier American-developed AI model to compete in the era of Microsoft dominance
- 5Google's core mission to organize knowledge and make it accessible aligns philosophically with open-source model distribution
Notable Quotes
“If you open source it you will win.”
— Speaker
“We need like a top tier American openweight model. Google could deliver that.”
— Speaker
“That would be the move of the century if they open sourced it and they tied it to the TPUs.”
— Speaker
“That would actually save the world potentially.”
— Speaker
Chapters
1. The Three-Year-Old Prediction
The speaker wrote a letter to Google management three years ago arguing that open-sourcing Gemini would lead to market victory.
2. Demis Hassabis's Compute Excuse
Demis Hassabis publicly explained that Google would open-source Gemini if they had sufficient compute, but instead only released Gemma due to compute limitations.
3. The Strategic Win: TPU and GCP Integration
Open-sourcing Gemini tied to TPU and GCP compute would enable Google to build an AI empire on infrastructure, similar to the Netscape-to-Firefox transition.
4. Geopolitical and Philosophical Implications
An open-source Gemini would give America a top-tier AI model and align with Google's mission to organize knowledge and make it accessible.
Key People & Entities
- Demis Hassabis
- DeepMind CEO; made public statements about why Gemini was not open-sourced
- Technology company that developed Gemini and Gemma AI models
- Tindar
- Person addressed in the transcript; invited to appear on the podcast
Glossary
- Open source
- Making software code publicly available for anyone to use, modify, and distribute
- Distillation
- A machine learning technique where knowledge from a large model is transferred to a smaller, more efficient model
- TPU
- Tensor Processing Unit; Google's custom-built hardware accelerator optimized for machine learning tasks
- GCP
- Google Cloud Platform; Google's suite of cloud computing services and infrastructure