Key insights
- The author argues that Google's open-source approach to AI, using TPUs and Gemini, will surpass OpenAI and Nvidia's closed ecosystems. They cite Gemini's neutrality and unique training data as advantages, and suggest AMD's ROCm could challenge Nvidia's CUDA dominance. This implies a potential shift in AI market leadership, favoring Google and AMD, with a moderately positive impact on GOOGL.
The cadence of AI progress right now is set by OpenAI and their releases of ChatGPT models.
I believe this a temporary phenomena as hyperscalers, sovereign nations, and even large corporates abandon the closed-architecture of OpenAI and Nvidia GPUs.
The future is open-sourced hardware (TPUs) and software (Gemini) and $GOOGL has shown itself to be at the forefront of this evolution.
Closed LLMs like X’s Grok are notoriously biased and users are picking up on this crux. Google’s Gemini has shown itself to be a neutral model, and has unique training data (Gmail + YouTube) that peers lack.
Ironically, the CUDA moat that made $NVDA famous will likely lead to its toppling as king (AMD ROCm is getting good enough).