Why People Contribute to Something: Exploring Generative AI and the Value Chain

Why People Contribute to Something: Exploring Generative AI and the Value Chain

Introduction:

The concept of contributing to something, whether it's a community, a cause, or a technological advancement, has always been deeply ingrained in human nature. There's a sense of fulfillment and purpose that comes from being kind and helpful to others. Kindness not only makes us happier but also gives us a sense of meaning and achievement. This desire to give back is especially prevalent in the field of generative AI, where individuals and companies strive to create something that will leave a utilitarian legacy for future generations.

The Value Chain of Generative AI:

When delving into the world of generative AI, it's important to understand the different players in the value chain and their respective roles. Infrastructure vendors, such as cloud platforms and hardware manufacturers, currently dominate the market, capturing the majority of revenue. Their role is to run the training and inference workloads for generative AI models. While application companies experience rapid growth in revenue, they often struggle with retention, product differentiation, and gross margins. On the other hand, model providers, responsible for training generative AI models, have yet to achieve large commercial scale. Despite their significant contributions, they haven't captured the majority of the value created in the market.

The Search for Differentiation:

In the quest to understand the market structure and drivers of long-term value in generative AI, it is crucial to identify the truly differentiated and defensible parts of the value chain. Traditional moats for incumbents, such as scale moats, supply-chain moats, ecosystem moats, algorithmic moats, distribution moats, and data pipeline moats, have limited durability in this rapidly evolving landscape. Thus, it becomes important to explore other avenues for differentiation.

The Role of End-User Applications:

In previous technology cycles, owning the end-customer was considered essential for building a large, independent company. However, in generative AI, the assumption that end-user applications will be the biggest players may not hold true. While several product categories, such as image generation, copywriting, and code writing, have already generated significant revenue, many apps in this space suffer from relatively low gross margins and a lack of differentiation. The reliance on similar underlying AI models and the absence of obvious network effects or unique data/workflows make it unclear whether selling end-user apps is the best path to building a sustainable generative AI business.

Vertical Integration and Differentiation:

The debate between consuming AI models as a service or training models from scratch is a crucial one. Vertical integration, where app developers continuously retrain models on proprietary product data, offers a way to create defensibility. However, this approach comes with higher capital requirements and a less agile product team. The decision between building features or standalone apps also shapes the future landscape of generative AI companies. Desktop apps, mobile apps, plugins, extensions, and bots all hold potential, but their fate in the market remains uncertain.

Managing Hype and Churn:

As generative AI continues to gain traction and buzz, it is essential to manage through the hype cycle. While revenue associated with generative AI companies may still be relatively small compared to usage, churn remains a concern. The early market dynamics and the presence of numerous competing products make it unclear whether churn is inherent or a temporary artifact. However, the potential for explosive growth in specific product categories, such as image generation, suggests that there is room for sustained success if the right conditions are met.

The Promise of Open Source and Hosting:

Open-source models and hosting services have emerged as key players in the generative AI landscape. Models released as open source can be hosted by anyone, including companies that don't bear the costs of large-scale model training. The success of Stable Diffusion, which offers major checkpoints for free, highlights the potential of open-source models to compete with proprietary alternatives. Additionally, demand for proprietary APIs and hosting services for open-source models is rapidly growing, creating new opportunities for commercialization and integration.

The Role of Money and Public Benefit:

The financial aspects of generative AI cannot be overlooked. Cloud providers and hardware manufacturers benefit significantly from the market, with a substantial portion of revenue flowing through to infrastructure companies. Startups training their own models also raise significant venture capital, much of which is spent on cloud providers. However, the potential harm associated with generative AI has led many model providers to incorporate the public good explicitly into their mission. Public benefit corporations and capped profit shares are examples of how money is being approached in this field.

The Role of GPU Manufacturers:

In the generative AI market, GPU manufacturers, particularly Nvidia, have emerged as major winners. Nvidia's strong moats, built through decades of investment in GPU architecture and a robust software ecosystem, have positioned them as leaders in the field. Nvidia GPUs are cited in research papers significantly more than the top AI chip startups combined. However, challengers such as Google and TSMC, with their TPUs and chip manufacturing capabilities, respectively, pose potential threats.

Surviving and Thriving in a Competitive Market:

The generative AI market holds immense potential, but competition is fierce at every level of the value chain. Durability and differentiation remain ongoing challenges. As AI models converge in performance over time, it becomes crucial to find sustainable advantages. Vertical clouds with specialized offerings may take market share from the dominant players. However, it's too early to determine if strong network effects will shape the market structure. The potential size of the market ensures healthy competition, with both horizontal and vertical companies likely to succeed based on end-market and end-user demands.

Conclusion:

In the world of generative AI, the drive to contribute and create something meaningful remains a powerful force. Kindness and a sense of purpose motivate individuals and companies to give back to their communities. However, understanding the value chain and the different players involved is essential to navigate this complex landscape. Vertical integration, differentiation, and managing through hype and churn are critical factors in building sustainable generative AI businesses. As the market continues to evolve, it's important to stay adaptable, embrace new technologies, and explore innovative approaches. Three actionable pieces of advice for anyone venturing into the generative AI space would be to:

1. Focus on vertical integration and differentiate your product or service from competitors to create a defensible position in the market.
2. Embrace the potential of open source and hosting services, as they offer opportunities for commercialization and integration.
3. Stay informed about the evolving landscape and be prepared to adapt to changing market dynamics.

By understanding the motivations behind contribution, the value chain of generative AI, and the challenges and opportunities in the market, individuals and companies can navigate this exciting field with purpose and success.