At Transforma Insights, we spent much of 2020 analyzing some of the best, most granular IoT predictions. In 2021, it was AI's turn to undergo a similar level of analysis. The results? First, it's important to clarify that our AI predictions differ from many other AI predictions. Our approach is to segment the AI market into a range of use cases, from complex image processing and natural language processing to risk analysis and threat detection. This list expands to approximately 42 use cases that we believe collectively represent opportunities to deploy AI to support enterprise efforts (we exclude AI deployed on PCs, tablets, and phones, though these applications extend enterprise processes). Furthermore, we don't predict the market size in US dollars because it's often unclear what those dollar figures actually mean: are they driven by end-user spending, licensing fees, service spending including AI features, service investment, or research investment? Or all of these? Instead, we predict AI instances based on the use cases and the location of instance deployments. For example, we might record a complex image recognition instance on an AI-enabled surveillance camera, and so on. This analysis tells us that the vast majority of AI instances will be addressed through deployments on IoT devices. In fact, over 95% of AI instances will continue to be deployed on IoT devices during our forecast period of 2020-2030. AI edge instances are growing faster than IoT, with the proportion of AI instances deployed on edge infrastructure increasing from around 0.4% to around 1.3% during this period. Unsurprisingly, given that these markets are at a more developed stage, the share of AI instances deployed in the cloud, as well as on PCs, tablets, and mobile phones (we do include AI instances as extensions of enterprise operational processes) declined during the forecast period. In summary, we predict that there are approximately 2 billion AI use case instances today, growing to over 20 billion by the end of this century. This growth is illustrated in the following figure.
It's fascinating to see how the breakdown of these use case instances changes during the forecast period (as shown in the figure below). Currently, natural language processing, chatbots and digital assistance, image processing and analysis, and activity recognition dominate, accounting for approximately 80% of all AI instances globally. These use cases will remain among the most important in 2030, along with customer behavior analytics. More importantly, although this newly added top-tier application set will only account for 66% of AI deployments by 2030, the long tail of AI will become longer and more significant over the next decade.
Clearly, the artificial intelligence market is already very important and is growing rapidly. Through these new forecasts, we aim to support discussions about how this market will evolve over time: identifying where and why AI will be deployed. Analysis of AI-related business opportunities can be derived from these forecasts, and the overall profile and size of these opportunities will, of course, depend on who you are: opportunities for different vendors will vary depending on how they interact with the AI market.
