Network Effects in AI Photo Enhancement Platforms

Self-Sustaining Ecosystem versus Disparate Toolkits

The more control users have over the creation of a tool, the more benefits they receive from that tool. This has created a significant market for certain types of products in social media, route-finding, and payment processing, as well as an emerging market for products using artificial intelligence technology to enhance digital photo editing processes. Based on data from Grand View Research, this market is projected to increase from approximately US$865 million in 2023 to more than US$3.5 billion by 2030. Although the software subscription model appears to be straightforward, in reality, the value of these types of products is based on the interaction between the data gathered from usage and the quality of the product and user experience.

Data Flywheel Induced Compounding

The Data Flywheel uses a simple economic concept to create compounded growth, even though it may be difficult to implement. A user provides their image to an AI-enhanced platform, which generates a learning signal for the AI model. The AI learns which modifications will yield images that meet the specific needs of the users. The more users use the platform, the more data the model receives to train on, hence increasing the amount and variety of data. The addition of new types of data allows for the creation of a better quality model. As the model improves, it attracts additional users to participate in the process and generate additional data to enhance the AI’s learning capabilities. According to McKinsey & Company, this is referred to as "the data flywheel." As the process of enhancing digital photos relies on the marginal cost of enhancing one additional photo being nearly zero, while yields from one additional photo produced often exceed zero, the data flywheel operates at great speed.

Thus, there is currently a growing gap in performance between the platforms that already exist and those entering the marketplace. The difference is much more than merely a factor of two (for example, 50 million user platform versus 25 million user platform), but a non-linear relationship, because larger and more diverse datasets allow access to more unique cases and rare failure modes than are available on smaller datasets. Well-capitalized entrants cannot quickly capture market share away from existing platforms due to their data structure connectivity barriers.

Real World Consolidation

Adobe's development of AI capabilities and AI enhancements to Lightroom and Photoshop show how incumbents use network effects to reinforce their dominance in the market. Every one of those customers using AI-driven tools creates user behaviour data, which helps improve the AI model. On the other hand, smaller platforms that focus on niche uses, such as Wink (which provides an image enhancer for mobile-first consumers), focus on a narrower use case rather than the entire creative workflow. For smaller players, being a niche player is a logical response to the existence of strong network effects, since smaller players are often better served by focusing on a niche market rather than trying to compete with the dominant player on the same level.

Impact on Competition and Consumer Welfare

Network effects on the consumer side have substantial economic ramifications. Due to the network effects of smartphone apps, consumers receive both better and lower-cost tools. For example, enhanced photo quality that would have taken a professional photo retoucher many hours to create now can be provided to a user free of charge or at a very low monthly subscription fee for the use of the smartphone app.

The same dynamics exist with respect to the competitive structure of markets that exhibit network effects. Specifically, the European Commission's Joint Research Centre published its research findings that network effects create winner-takes-all market structures with respect to artificial intelligence-enabled platform markets. The research findings included high concentration ratios of the AI-enabled platform market in the first three to five years after market emergence, with Herfindahl-Hirschman Index (HHI) scores indicating oligopoly/near-monopoly market structures.

There are several competing interests created for regulators to manage. Blocking acquisitions to maintain competitive plurality, for example, may slow down data aggregation that will enhance the underlying network technology. Allowing consolidation could potentially create a situation where consumers are locked into ecosystems in which the switching costs (for example, having to upload image libraries again, losing platform-specific editing presets, retraining workflows) create informal barriers reinforcing the market power of the platforms long after the initial network effect has been realized.