According to global tech market advisory firm, ABI Research, the global edge Artificial Intelligence (AI) Software-as-a-Service (SaaS) and turnkey service market will grow at a Cumulative Average Growth Rate (CAGR) of 46% between 2020 and 2025 to reach US$7.2 billion in 2025. This is 25% of the global edge AI market, which is estimated to be US$28 billion by 2025, comprising of edge AI chipsets, SaaS, and turnkey services, as well as professional services.
As the benefits of edge AI becomes more obvious, enterprises are searching for edge AI solutions that are low latency and fully secured to assist them with data-based decision-making. “The proliferation of edge AI chipset options means enterprises are no longer limited by hardware choices and can select the best-of-breed solution that fits their needs. They now look to invest in SaaS subscriptions, turnkey services, and managed services that can facilitate the deployment of edge AI,” explains Lian Jye Su, Principal Analyst at ABI Research.
Sensing this huge market opportunity, public cloud service providers joined the ecosystem by offering edge AI development boards, hardware systems, software toolkits, and cloud-based services. Google was the first to offer a development board with its Edge TPU designed for edge applications. Over the past six months, AWS and Azure have also strengthened their edge AI portfolio through hardware and software products, managed services, and industrial partnerships. This enables existing cloud service users to get into the edge AI ecosystem, lowering the barrier to entry for enterprises who are not familiar with edge AI development.
“As in many other industries, the entrance of cloud service providers has led to a lot of hype and excitement,” says Su. However, unlike cloud environment that has standardized servers and processors, edge AI is a very diverse market that covers a broad range of device form factor, processing power, and use cases. “What enterprises need are industry-grade edge ML models that can be deployed for various applications across multiple asset categories. Furthermore, not all enterprises are able to build their own models using tools provided by public cloud vendors. Building the right edge AI solutions using software from cloud vendors requires in-depth domain expertise and know-how.”
This has led to the emergence of startups specializing in software-as-a-service and managed services for the design, development, and deployment of edge AI, such as Edge Impulse, Ekkono Solutions, Imagimob, Mispsology, Qeexo, and SensiML.
These companies tend to provide end-to-end edge ML Operations (MLOps) software and services that enable continuous integration, deployment, and monitoring of edge ML models, often through low code or zero code methods. In addition, these startups have specialized skillsets in model compression and hardware optimization, best practices around data governance, and seamless integration with other enterprise software and platforms.
The future of edge MLOps lies in a higher level of automation through low code or zero code design. This not only lowers the barrier to entry for end-users who do not possess data science or machine learning expertise but also enabling them to perform edge MLOps in a seamless manner. “AutoML processes, such as neural architecture search, feature store, hyperparameter tuning, and lifelong learning, allows quick onboarding and development of edge ML models. This allows enterprises to overcome the lack of data science and machine learning expertise and focus on operationalizing edge AI in their assets,” concluded Su.
These findings are from ABI Research’s Edge AI Ecosystem application analysis report.
