Nordic, Telit, and MeiG Highlight the Industry-Wide Applicability of AI for the IoT at CES 2026

Nordic Semiconductor acquired Atlazo’s Axon technology in 2023. The Axon Neural Processing Unit (NPU) is an Artificial Intelligence (AI) hardware accelerator for edge AI workloads. The NPU is a physical element integrated into a Nordic chipset to avoid the need to run AI inferencing on a Central Processing Unit (CPU) or to add a discrete NPU alongside the CPU. This makes for a simplified final product, with fewer components, more efficient operation, better power consumption, and a better battery life. The NPU runs proprietary “Neuton” edge AI models, which, according to Nordic, will usually be less than 5 Kilobytes (KB) in size. Neuton.AI’s Machine Learning (ML) models were acquired by Nordic in June 2025. What stands out is that Axcon and Neuton are specifically designed for small form factor, battery-powered AI-Internet of Things (IoT) devices, with example tasks cited by Nordic as being, but not limited to, sound classification, keyword spotting, and image-based detection.

The purpose of the Axon NPU is to make millisecond decisions on an edge device, without any latency from communications to and from the cloud, ensuring any applicable compliance requirements for local processing only—and to do so with the best possible battery life for always-on wireless connected IoT devices. Neuton models are built using the Nordic Edge AI Lab. It is available on Nordic’s Short-Range Wireless (SRW) nRF54 System-on-Chip (SoC) range, and its long-range cellular IoT System-in-Package (SiP) modules. The nRF54 is the only product from Nordic with an integrated Axon NPU at present, with Nordic’s cellular SiPs running Neuton models on their CPU. While this is less efficient, it makes sense that future cellular products will be available with NPU-enabled variants. Axon products were announced as samples at CES, for general availability in 2H 2026.

Telit Cinterion has partnered with Nokia to integrate its newest models of cellular, Wi-Fi, and Non-Terrestrial Networking (NTN)-enabled modules with Nokia’s Cognitive Digital Mine (CDM) platform. This will target wireless IoT applications in the oil & gas, logistics, and mining industries, plus others. The stated outcomes are to minimize operational downtime, minimize the production of emissions, and maximize workforce safety and enterprise productivity. This is an AI strategy that leverages Telit’s existing core competency, rather than acquiring or developing new abilities in-house, to then seek to sell directly. To be trusted by network infrastructure heavyweight Nokia is to be allowed to tap into its sales channels and network of enterprise customers to potentially develop lucrative new opportunities, especially with regard to 5G.

Nokia's CDM platform makes use of AI and digital twins in far-edge computing use cases that require making real-time decisions on-site, and without recourse to remote resources. This is particularly relevant for specialized industrial vehicles and equipment that may be deployed in remote locations. The principal physical component of Nokia’s CDM is its “Black Box”, which is a ruggedized integrated compute and connectivity hardware system for deployment in physically demanding industrial environments. One of the Black Box’s tasks is intelligent redundancy, dynamically switching between and combining communication channels to guarantee Service Level Agreements (SLAs) for uninterrupted operation and Quality of Service (QoS).

This is where the integration of Telit’s cellular, Wi-Fi, and NTN-enabled comes in, to provide the wireless carrier-certified communications hardware subsystem needed to bridge the transmission of mission-critical IoT data between local-area SRW nodes, with wide-area 5G network backhaul and emergency satellite fallback. Communications is critical to Nokia’s industrial platform customers, as is autonomous operation. It is not a case of one or the other, but of each being a different side of the same coin of resiliency. Nokia and Telit state they seek to enable, through collaboration, “a new class of intelligent, self-optimizing machines and systems capable of sensing, deciding, and acting locally, even when disconnected from the cloud.”

MeiG Smart is a notable survivor of the long tail of Chinese module vendors, seeking to follow in the wake of Quectel by not only building a high-growth business domestically, but also actively seeking expansion throughout the Rest of the World (RoW). MeiG is looking to have experienced the most impressive growth in both percentage terms and actual shipments after Quectel in 2025, with an expected module product line turnover of close to US$500 million, and with a third of its business expected outside of China. At CES, MeiG launched three products: the Qualcomm Dragonwing IQ-9075-based SNM983 AI-only module with up to 100 Trillion Operations per Second (TOPS) of computing power; MeiG’s MT200 series of all-in-one AI compute boxes with a tunable 24 TOPS to 100 TOPS of performance depending on the choice of module inside; and its SRT8710 AI-integrated 5G mobile (MiFi) hotspot.

The SNM983 module is aimed at the high-performance edge compute markets for automotive, robotics, and unmanned systems, while the MT200 series looks to smart manufacturing and industrial monitoring, although both are effectively general-purpose devices for high-compute plug-and-play AI requirements. Most interesting is the SRT8710 hotspot with an AI assistant. It is not intended as a consumer product but as a smart business hub to provide voice interaction, image recognition, text translation, QR code-scanning, mobile payment QR code display, and intelligent network traffic optimization for Local Area Network (LAN)-attached devices. It may be the next-level successor to the Cat-1bis-based “cloud speakers” that proliferated in manned kiosks and single-employee retail operations in China and India in recent years.

Nordic Semiconductor is investing in AI for low-power optimization to do as much as possible with as little as possible, to place intelligent compute and connectivity into as many endpoints as possible. MeiG Smart has a simpler approach, taking Qualcomm’s most powerful new chipsets and positioning them in fewer but higher-value use cases. While Telit is neither the investor in nor the buyer and packager of AI, it is a strategic partner to those with a mature AI competency and a trusted relationship with valuable regulated industries. AI has a complementary appeal across the board in the IoT, for the exact same reason that simple connectivity always has: to enable enterprises and municipalities to do exactly what they already do, but better than ever before, taking efficiency and effectiveness to the next level.