The Future of Generative AI: Exploring Ownership, Challenges, and Opportunities

The Future of Generative AI: Exploring Ownership, Challenges, and Opportunities

"Be brave. Take risks. Nothing can substitute experience." ― Paulo Coelho

Generative AI, with its ability to create new and unique content, has captured the imagination of businesses and consumers alike. From image generation to copywriting and code writing, the growth of generative AI applications has been staggering. However, as the market continues to evolve, questions arise about 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. While application companies are experiencing rapid topline growth, they often struggle with retention, product differentiation, and gross margins. On the other hand, model providers, responsible for the existence of this market, have yet to achieve large-scale commercial success.

In the absence of strong technical differentiation, businesses drive long-term customer value through network effects, data retention, or by building complex workflows. However, it remains uncertain whether selling end-user apps alone is the most effective path to building a sustainable generative AI business. There is potential for margins to improve as competition and efficiency in language models increase. Additionally, as AI tourists leave the market, customer retention is expected to increase. Vertically integrated apps, which tightly couple the user-facing app to the home-grown model, may have an advantage in driving differentiation.

For model providers, it is becoming evident that commercialization is closely tied to hosting. The demand for proprietary APIs, such as those offered by OpenAI, is rapidly growing. Hosting services for open-source models, like Hugging Face and Replicate, are emerging as useful hubs for sharing and integrating models, creating indirect network effects between model producers and consumers. Many model providers have also incorporated the public good explicitly into their mission, organizing as public benefit corporations (B corps) or issuing capped profit shares, without hindering their fundraising efforts.

However, a significant portion of the revenue in the generative AI market ultimately flows through to infrastructure companies. On average, app companies spend a substantial percentage of their revenue on inference and per-customer fine-tuning. This expenditure is often paid directly to cloud providers for compute instances or to third-party model providers who, in turn, allocate a significant portion of their revenue to cloud infrastructure. As a result, it can be estimated that a considerable portion of total revenue in generative AI today goes to cloud providers.

Behind the scenes, driving the majority of AI workloads, Nvidia emerges as the biggest winner in generative AI. With $3.8 billion of data center GPU revenue in the third quarter of its fiscal year 2023, including a significant portion for generative AI applications, Nvidia is a dominant player in the infrastructure layer of the stack. While there are standard moats, such as scale, supply-chain, ecosystem, algorithmic, distribution, and data pipeline, it remains uncertain whether any of these moats will be durable over the long term. The presence of strong, direct network effects in any layer of the stack is yet to be determined.

In the future, it is expected that both horizontal and vertical companies will succeed in generative AI, with the best approach dictated by specific end-markets and end-users. If the primary differentiation lies in the AI itself, verticalization, which tightly couples the user-facing app to the home-grown model, is likely to prevail. On the other hand, if the AI is part of a larger, long-tail feature set, horizontalization is expected to occur.

As the generative AI landscape continues to evolve, businesses and individuals must adapt to the changing dynamics. Here are three actionable pieces of advice to navigate this exciting yet complex field:

1. Embrace collaboration: As the market matures, collaboration between model providers, infrastructure vendors, and application companies will become increasingly important. By leveraging each other's strengths, these stakeholders can create more robust and differentiated offerings.

2. Prioritize data ethics: With the potential for generative AI to harm as well as benefit, it is crucial for businesses to prioritize ethical considerations. Incorporating the public good into the mission, as many model providers have done, can be a responsible approach to ensure that the technology is used for positive impact.

3. Focus on long-term value: While short-term gains may be tempting, businesses should prioritize long-term customer value. Building sustainable generative AI businesses requires a focus on retention, product differentiation, and gross margins.

In conclusion, the future of generative AI holds immense potential and challenges. Infrastructure vendors currently dominate the market, but the landscape is still evolving. As businesses navigate this dynamic field, collaboration, data ethics, and a focus on long-term value will be essential for success. By embracing these principles, stakeholders can shape the future of generative AI in a responsible and impactful manner.