Research Publication – Artificial Intelligence – AI Ecosystem Part I

3rd September 2024: RFM expands its coverage of Artificial Intelligence with an analysis of the emerging AI Ecosystem and how it will evolve over the coming years. RFM subscribers will receive their copy by email.  

The battle for the next ecosystem has already begun and it will look very like the smartphone ecosystem. The AI Ecosystem is at an early stage with most players being very vertically integrated and doing most of the development in-house. This has meant that the platform of choice is the silicon development platform which RFM refers to as AI Ecosystem 1.0. As generative AI matures, developers are likely to become less vertically integrated and will rely on platforms that are based on foundation models (AI Ecosystem 2.0) or services themselves (AI Ecosystem 3.0). With evolution comes opportunity and leadership in the AI ecosystem could easily change hands several times.

• The Digital Ecosystem: is the virtual environment where users live their digital lives, interact with one another and consume digital goods and services. RFM sees the emerging AI ecosystem evolving in a very similar manner to the smartphone ecosystem which evolved between 2007 and 2016.
• Platforms: The optimum location of a platform is the point where creators do things the same way changes to where creators all do things differently. The point of a platform is not to have to recreate the entire product from the ground up and, as such, they tend to become points of control once they become very popular with creators. It is Nvidia’s ownership of CUDA, a key control point in the early AI ecosystem, that is mostly responsible for its recent success.
• The AI Ecosystem 1.0: is where the industry is today and is dominated by Nvidia with its CUDA platform and its ability to release cutting-edge products many months ahead of its competitors. CUDA is the current control point but there are signs that this may be changing.
• The AI Ecosystem 2.0: is where the control point migrates to the foundation model. Many LLM owners are encouraging developers to develop services on their LLMs. This would mean that the silicon development platform would become less relevant. RFM thinks that this is the opportunity for others to take share from Nvidia, but it is going to take some years.
• The AI Ecosystem 3.0: is when development shifts from creating services to combining and refining what has already been created to produce new and more complicated functionalities. Nvidia is targeting its AI Foundry and Nvidia Inference Microservices (NIMs) offering here as a strategy to mitigate the risk presented by AI Ecosystem 2.0
• The Players: largely fall into 3 camps. These are the established digital ecosystems, generative AI start-ups and tool providers. These are the players that are vying to become the go-to place to create and purchase generative AI services because it is through owning one of the control points that high returns on invested capital will be made.
• Part II: will focus on the players, their strategy and their market positioning to estimate how well they are likely to fare as the AI ecosystem emerges over the coming several years.