Key Motivations and Outcomes:
The Artificial Intelligence (AI) industry is developing at a rapid pace. However, the AI market is becoming increasingly complex due to the diverse end devices and use cases that leverage AI capabilities. It can no longer be treated as a single market or as a single technology with a specific addressable market or harmonized implementation requirements.
In light of this development, ABI Research has designed a comprehensive market data (MD) model to help our clients assess Total Addressable Markets (TAMs) and related business opportunities for AI chipset vendors across different industry verticals and device types, and to evaluate deployments of the technology over the rest of the decade. It looks at paths to monetization across key segments of the market and identifies market forces driving the technology implementation across each segment studied.
Market Data Output and Segmentation:
This MD provides forecasts for the following key components of the AI market:
• Tiny Machine Learning (TinyML) chipset shipments and revenue
• Breakdowns per vertical/device
• Split between architectures (Microcontroller Unit (MCU), Ultra-Low Power Central Processing Unit (ULP-CPU), Field Programmable Gate Array (FPGA), and Neural Processing Unit (NPU))
• Regional breakdowns
For each segment, ABI Research looks at market penetration of AI technologies across different use cases and analyzes the key factors that will drive business opportunities in each segment.
Below is a schematic diagram on the methodology of the forecasting process and key components of the market data.
For more details on each end device, use case, AI process, and chipset architecture covered in this report, please refer to the Definitions tab covering use cases.
Data Validation:
ABI Research has been tracking the AI market since 2013 and publishing market forecasting data and application-specific and vertical-specific reports. Data were generated by tracking companies and trends in different vertical application segments. Before publishing any output from ABI Research’s forecasting model, data have been shared with key industry players and stakeholders for validation. Feedback received from different parts of the value chain are then harmonized and the results enable ABI Research to update the model’s parameters and assumptions, and to align them with the overall industry expectations. Key assumptions have been made to adjust the forecast to reflect recent developments caused by industry trends like the deployment of NPUs in embedded devices.
Forecasting Models:
This MD covers all types of embedded, or TinyML, chipsets servicing major use cases and implementations at the far edge, from always-on sensors in personal and work devices, to predictive maintenance in industrial applications. Segmentation via device type for each vertical area was established by speaking to and studying relevant chipset suppliers, device vendors, infrastructure suppliers, and technology implementers. The same approach was used to forecast each AI use case and the architecture of the dedicated hardware for performing inference at the edge. It was also important to detail the use cases enabled by AI at the device level, and then to detail the hardware that would be used to support inference workloads.
The granular segmentation provided by this model, together with the range of our analysts' expertise, mean that ABI Research has been able to provide reliable datasets based on top-to-bottom and bottom-up approaches, reflecting on the balance of technology supply and market demand. This approach enables ABI Research to be unique in tracking both market and technology transformations across various industries. These have been based on dynamic changes in the typical characteristics of each market vertical, their relative pain points, and how the technology supply chain is aligned to solve these pain points.
Key Input Data in This Iteration:
Internal modeling, as well as primary and secondary research involving key chipset vendors and Original Equipment Manufacturers (OEMs).
