The GenAI Chessboard

The strategy game for AI platform dominance is just getting started

In his book, “The Grand Chessboard,” Zbigniew Brzezinski, a former National Security Advisor to President Carter, described the complex and strategic nature of the global geopolitical landscape in the post-Cold War era. Nations and major powers compete for resources, influence, and dominance on the Chessboard. Similarly, on the GenAI Chessboard, tech giants and emerging players are competing for resources, market share, and dominance.

In my previous post, I discussed why GenAI is the next transformational technology and why powerful GenAI models will become the next tech platforms like Apple and Android for mobile applications. The technological arms race is already intensifying, with major tech companies and open-source players bringing their A-game and new GenAI models being introduced almost weekly. The stakes are high and the outcome of the game will rearrange tech companies along the power curve.

3-Dimensional Chessboard

There are 3 layers in how the GenAI technology will be developed and used. This week, I explored the platform layer.

Platform layer: companies who build and train GenAI models, including foundational and large language models like ChatGPT as well as smaller models for specific use cases.

Application layer: companies who build on top of GenAI models, fine-tune them further, and make them more user-friendly and accessible for end-customers.

Customer layer: consumers and businesses who are the end-customers of GenAI technology to meet their needs, such as increasing productivity. Customers will vote with their wallet, usage, and attention to determine the winners and losers across the platform and application layers.

Platform Layer

There are a lot of GenAI models that we don’t often hear in the headlines. Clem Delangue, co-founder and CEO of Hugging Face, said ~15,000 companies uploaded ~250,000 ML models onto Hugging Face. About half of these are open-source and the other half are proprietary models. In addition, big tech and industry leaders are also building their models. Below is a simplified illustration of the platform layer (non-exhaustive):

The debate will continue whether there will be a large number of models vs a few dominant models to rule them all. It is prohibitively expensive and complex to develop general purpose large language models. However, we are likely to see a long tail of smaller models designed and trained for specific use cases and can get the job done faster and cheaper. Thus, companies building the application layer will need to figure out what is the best model for their specific use cases.

OpenAI’s Game Strategy

OpenAI has a well-thought-out game strategy followed by ‘fast and furious’ execution. In a move reminiscent of The Queen's Gambit, Sam Altman sacrificed 49% of OpenAI to Microsoft, a calculated decision that could lead to significant long-term strategic advantages. OpenAI is going after the most critical resource, poaching AI talent from its biggest competitors such as Google, Meta, and Apple.  

OpenAI is investing in the application layer through a $100Mn startup fund. To accelerate enterprise adoption, they formed strategic alliances with Bain, BCG, and Morgan Stanley. To build its defense, OpenAI has aggressively dropped its API price by 90%, and integrated ChatGPT into SnapChat, Slack, Instacart, Shopify and more. These well-orchestrated bold strategic moves are all in pursuit of becoming the dominant player on the GenAI Chessboard.

By integrating OpenAI's powerful models with Microsoft's comprehensive ecosystem of products (Bing, Windows, Office, Azure, virtual assistant, Xbox), other tech giants will face enormous pressure to devise their game strategies and play catch-up.