"AI: Startup vs Incumbent Value - Exploring the Shifting Landscape"

"AI: Startup vs Incumbent Value - Exploring the Shifting Landscape"

The world of AI has seen a tremendous shift in value distribution between startups and incumbents. In the earlier waves of technological advancements, such as the internet and mobile, startups managed to capture a significant portion of the value, alongside incumbents. However, when it comes to AI, the majority of the value has surprisingly gone to incumbents, with startups struggling to make a dent in the market.

In the internet wave, giants like Google, Amazon, and Facebook emerged as dominant players, while incumbents like Microsoft and IBM extended their reach onto the internet. The split was roughly 60:40 or 70:30 in favor of startups. Similarly, in the mobile arena, incumbents like Apple and Google reigned supreme, but startups like Whatsapp and Uber managed to secure a 20:80 split. Crypto, on the other hand, has been a game entirely dominated by startups, with very little involvement from existing financial services or infrastructure companies.

To compete with incumbents, startups need to present a product that is dramatically better and overcome the distribution, capital, and pre-existing product moats. Alternatively, they can focus on untapped customer segments or distribution channels that incumbents are unable to serve. In general, a 10X better product is required to stand a chance against incumbents. However, the advantage that incumbents enjoyed, thanks to their data advantage, may be diminishing as companies now have access to a broader internet training set and are adopting models that work effectively with smaller data sets.

Many AI companies in the prior wave directly challenged incumbents or operated in challenging markets such as education or healthcare, where market structure and regulations often impede technological innovation. However, the current wave of AI feels different for several reasons. The speed of innovation across various areas is remarkable, and the technology itself has become dramatically stronger. This enables startups to create products that are 10X better, making it easier to overcome incumbent advantages.

While GPT-3, a highly advanced AI model, has not yet paved the way for startups to build significant businesses, a model that is 5-10X better could potentially create a new startup ecosystem while enhancing incumbent products. Additionally, there is a clear emergence of infrastructure-centric companies with widespread adoption, such as OpenAI, Stability.AI, Hugging Face, Weights and Biases, and others. These companies provide crucial access to AI technologies, opening up more opportunities for startups.

The key lies in identifying actual end-user needs and underserved markets that can benefit from the advancements in AI. Focusing on the needs of end users is crucial to avoid the trap of building solutions in search of problems. This wave of exciting technology should be harnessed to address real market demands.

Moving on to the "6 New Theories About AI," we find that AI holds the potential to push creation costs toward zero, just as the internet eliminated distribution costs. However, the economic value from AI will not be distributed evenly along the value chain. Instead, it is expected to consolidate rapidly among infrastructure players and end-point applications, resulting in power law outcomes.

With widely available data sets and mathematical frameworks, the barrier to entry for building AI models is primarily compute power. Any company with sufficient skills and resources can potentially replicate existing models. The real differentiator lies in the developer community, ease of use, and the creation of a robust ecosystem. Open-source models also exert downward pricing pressure on API-based model providers, forcing them to compete with free alternatives.

Long-term differentiation in models comes from data-generating use cases, rather than fine-tuned models. Open-source AI startups tend to function more as consulting shops than traditional SaaS companies. The battle for dominance among endpoints is often driven by go-to-market strategies and marketing prowess, rather than pure model performance. AI becomes a powerful marketing tool, but the winners are determined by software questions and the ability to leverage existing distribution channels.

For startups competing in the SaaS space, having inherent distribution or product capabilities becomes a crucial competitive advantage. Integrating AI into existing products for large companies is easier than building a full-suite AI product from scratch. Distribution ultimately triumphs over pure AI capabilities.

In a world where content creation costs are nearly zero, the real differentiator is distribution. Creators who effectively utilize AI tools to generate better content at a faster pace will be able to build a substantial fan base. However, this also means that the revenue distribution will become more skewed, with a small percentage of creators capturing the majority of the revenue. AI has the power to amplify this dynamic in the digital media landscape.

Lastly, there is the concept of "Invisible AI." Some companies harness the power of AI without explicitly mentioning it. Instead, they use AI to create something previously unimaginable, resulting in delightful experiences for users.

As we look ahead, it appears that startups will finally start to gain real value from AI. The rapid pace of innovation, coupled with the strength of the technology itself, creates a favorable environment for startups to thrive. Exciting times lie ahead as AI continues to shape various industries and revolutionize the way we live and work.

In conclusion, here are three actionable pieces of advice for startups entering the AI space:

1. Focus on building a product that is truly 10X better than what incumbents offer. Overcoming distribution, capital, and pre-existing product moats is essential to compete with incumbents.

2. Identify untapped customer segments or distribution channels that incumbents cannot serve. This can provide a unique advantage and increase the chances of success.

3. Pay close attention to actual end-user needs and unserved product/markets. Avoid the trap of building solutions in search of problems. Understanding the market demand and addressing it effectively is crucial for long-term success.

With these strategies in mind, startups can position themselves to make a significant impact in the evolving AI landscape. The potential for value creation is immense, and those who navigate the terrain wisely will reap the rewards of this technological revolution.