The Rise of Generative AI and the Ownership Dilemma

The Rise of Generative AI and the Ownership Dilemma

Generative AI, with its ability to create images, copy, and even code, has experienced tremendous growth in recent years. Companies in this space have seen significant revenue, but they also face challenges in retention, differentiation, and gross margins. The question arises: who truly owns the generative AI platform?

Infrastructure vendors have emerged as the biggest winners in this market, capturing the majority of the revenue flowing through the stack. They provide the necessary backbone for generative AI applications to thrive. On the other hand, application companies struggle with retaining customers and differentiating their products. Model providers, despite being responsible for the existence of this market, have yet to achieve large-scale commercial success.

The growth of generative AI applications can be attributed to novelty and a wide range of use cases. B2B and B2C apps create long-term customer value through network effects, data retention, and complex workflows. However, it is not clear whether selling end-user apps is the best path to building a sustainable generative AI business. Vertically integrated apps may have an advantage in driving differentiation.

One key observation is that commercialization in generative AI is closely tied to hosting. Demand for proprietary APIs is rapidly increasing, and hosting services for open-source models are emerging as useful hubs for sharing and integrating models. Model providers have even incorporated the public good into their mission, without hindering their fundraising efforts.

While the generative AI market appears promising, a significant portion of the revenue flows to infrastructure companies. App companies spend a substantial amount on inference and fine-tuning, either directly to cloud providers or to third-party model providers who, in turn, spend a significant portion of their revenue on cloud infrastructure. Nvidia, a major player in this space, has reported substantial revenue from data center GPUs.

Infrastructure companies, like Nvidia, enjoy a lucrative and seemingly defensible position in the stack. They benefit from various moats, including scale, supply-chain, ecosystem, algorithmic, distribution, and data pipeline moats. However, the durability of these moats over the long term remains uncertain. It is also unclear if any layer of the stack will experience strong network effects.

The ownership dilemma extends beyond infrastructure companies to the entire generative AI ecosystem. It is not yet clear if there will be a long-term winner-take-all dynamic in this market. Both horizontal and vertical companies have the potential to succeed, depending on the end-markets and end-users they serve. Verticalization may be advantageous when the AI itself is the primary differentiating factor, while horizontalization may be more suitable when AI is part of a larger feature set.

Moving beyond the ownership question, the article shifts its focus to DAOs (Decentralized Autonomous Organizations) and the Iron Law of Oligarchy. DAOs aim to combine the benefits of democracy with ownership and governance. However, direct democracies are impractical due to the challenges of coordination and information problems. The evolution from democracy to oligarchy, known as the Iron Law of Oligarchy, suggests that all organizations, even those committed to democratic ideals, will inevitably be ruled by an elite few.

Monarchy, oligarchy, and democracy represent different forms of governance, each with its strengths and weaknesses. Monarchies are efficient but can lead to poor outcomes if the monarch is ineffective. Oligarchies are stable but tend to serve themselves rather than the entire population. Direct democracies are cumbersome and suffer from information problems.

The cyclical theory of governance, known as anacyclosis, suggests that there is an evolution between monarchy, oligarchy, and democracy over time. Elites in an oligarchy develop values that differentiate themselves from the democratic masses, leading to decisions that benefit the elite few rather than the entire group.

While DAOs aim to overcome the challenges of democratic governance, they are not immune to the tendency towards oligarchy. Financial incentives and pre-programmed rules may slow down this process, but the default trend within most democratic organizations is towards oligarchy.

In conclusion, the rise of generative AI and the ownership dilemma highlight the complexities of this rapidly evolving field. Infrastructure companies currently hold a significant position in the market, but the durability of their advantages remains uncertain. The question of who truly owns the generative AI platform is still up for debate. Meanwhile, DAOs strive to combine the benefits of democracy with ownership and governance, but they must contend with the Iron Law of Oligarchy. As the generative AI landscape continues to evolve, it is crucial to consider the implications of ownership and governance structures.

Three actionable advice before conclusion:

1. For companies in the generative AI space, consider exploring vertical integration to drive differentiation and long-term customer value.
2. Model providers should prioritize commercialization by focusing on hosting services and proprietary APIs to capture value in the market.
3. DAOs should carefully design financial incentives and pre-programmed rules to mitigate the natural tendency towards oligarchy and ensure a more democratic and inclusive governance structure.