Introduction
In an increasingly connected world, the amount of data generated by devices and digital systems keeps growing. To cope with this explosion of data, companies and researchers are turning to edge computing and artificial intelligence (AI) to process and analyse information as close as possible to its source. This trend towards decentralisation offers considerable benefits in terms of reduced latency, data security and energy efficiency.
In 2025, edge computing and AI have become essential pillars for many sectors, ranging from telecommunications to healthcare and the automotive industry. This development is largely driven by the integration of AI capabilities directly into edge devices, an evolution made possible by advances in computer hardware and by the synergy with 5G networks. This article explores in detail the recent progress, applications, challenges and future of this revolutionary technology.
Background/History
The beginnings of edge computing date back to the 2010s, when companies began to realise the limitations of centralised cloud architectures, particularly with regard to latency and bandwidth. Over the years, technological progress has made it possible to move data processing closer to its sources, thereby reducing the need to transfer large amounts of data to remote data centres.
A major turning point was reached around 2020 with the emergence of the Internet of Things (IoT), which amplified the need for a more decentralised infrastructure. This period also saw the first integrations of AI capabilities into edge devices, made easier by progress in the miniaturisation of electronic components and in computing power. Since 2024, the focus has been on optimising AI models to run efficiently on these devices, a trend that continues to grow in 2025.
Applications/Use cases
Edge computing and AI have found practical applications in various sectors. In healthcare, for example, medical devices equipped with AI capabilities can analyse patient data in real time, enabling faster and more accurate interventions. Similarly, in the automotive industry, autonomous cars use edge computing to make instant decisions based on sensor data, thereby improving passenger safety.
Industrial sectors also benefit from this technology through the optimisation of production lines. Smart sensors collect and analyse data in real time, enabling immediate adjustments that increase efficiency and reduce costs. These applications demonstrate how edge computing and AI are transforming operations by taking advantage of the speed and efficiency of local processing.
Technologies/Methods
At the heart of these innovations are advanced technologies such as quantisation and the size reduction of AI models, which allow these systems to run on devices with limited resources. Quantisation, for example, reduces the precision of model weights to lower computing and memory requirements, while model size reduction removes redundant parts to optimise performance. These techniques give rise to compact models designed to run locally, such as Phi, released by Microsoft, or Llama, released by Meta.
The synergy with 5G networks also plays a crucial role, as it offers ultra-low latency and high bandwidth, enabling near-instantaneous interactions between edge devices and core networks. On the hardware side, platforms such as NVIDIA's Jetson modules embed the computing power required for inference directly on the device. Software platforms such as AWS Greengrass and Azure IoT Edge facilitate this integration by providing the tools needed to deploy and manage intelligence at the edge.
Challenges/Limitations
Despite its many benefits, edge computing presents notable challenges. One of the main ones is data security: the more data is processed locally, the more exposed it is to cyberattacks. Companies must therefore invest in robust cybersecurity solutions to protect sensitive information.
This requirement for data control also guides the way Koaee designs its multilingual conversational assistants for businesses: available in sixteen languages, they rely on servers located in Europe and do not retain any conversations. Processing relies on models from Mistral (Voxtral) and OpenAI (GPT). The spirit is the same as that of edge computing: keeping data as close as possible to where it is used, and under control.
Another challenge lies in the heterogeneity of edge devices, which can complicate system integration and compatibility. Manufacturers must ensure that their devices can work together seamlessly, which requires common standards and greater interoperability. Finally, power management and energy consumption remain a significant obstacle, requiring continuous innovation to improve the energy efficiency of edge devices.
Outlook
In the future, edge computing and AI will continue to develop, driven by technological innovation and the growing demand for decentralised solutions. Companies will invest more in research to overcome current challenges and fully exploit the potential of these technologies. Continued integration with 5G networks will open up new opportunities, particularly in the fields of augmented reality and smart cities.
Progress in the miniaturisation of electronic components and in computing power should also facilitate the adoption of edge computing in currently under-exploited sectors, such as agriculture and energy. With strategic investment and cross-sector collaboration, edge computing could fundamentally transform the way data is processed and used on a global scale.
Conclusion
Edge computing and AI represent a major evolution in the way we process and analyse data in the 21st century. By bringing intelligence closer to data sources, these technologies offer significant benefits in terms of speed, security and efficiency. Although challenges remain to be addressed, the outlook is promising, and the potential impact on various sectors is immense.
In conclusion, as the world continues to generate unprecedented amounts of data, edge computing and AI will provide the tools needed to exploit this information in innovative and efficient ways, paving the way for a new era of discovery and technological progress.
Source: Eurostat, survey “Artificial intelligence by size class of enterprise” (isoc_eb_ai) — companies with 10 to 249 employees, France. Retrieved via API on 30 August 2026. The survey publishes neither 2022 nor 2026: the series ends at the latest available year.
