HighTek Takes on NVIDIA
Russia's leading microelectronics developer, HighTek, has unveiled the country's first AI accelerator for embedded systems, delivering up to 30 TOPS of computing performance. The new processor is designed to bring advanced artificial intelligence directly to edge devices, allowing them to process data locally without relying on remote cloud infrastructure.
HighTek has developed the country's first AI accelerator for embedded systems based on a Russian-designed microprocessor. The LinQ HX accelerator is built entirely from Russian-made components and does not require foreign technology licenses.
First Russian Edge AI Accelerator
The architecture combines the LinQ H1M microprocessor, designed to accelerate neural network workloads and included in the Russian Ministry of Industry and Trade's official registry, with a general-purpose coprocessor. The module delivers 30 TOPS (trillion operations per second), the highest performance reported for a Russian-developed AI accelerator.
The module can run up to five different neural network models simultaneously, allowing multiple AI workloads to execute in parallel with latency as low as 2.3 milliseconds for certain neural networks. After optimization, latency drops to 1.5 milliseconds. A single module can perform numerous real-time computing tasks at once, including building terrain maps, detecting and classifying objects, performing semantic segmentation, planning routes, and transmitting telemetry. Power consumption remains below 40 watts, making the module suitable for mobile and autonomous platforms. A broad range of interfaces – including Ethernet, CAN, USB, HDMI, and others – supports cameras, including thermal imaging cameras, sensors, and control systems. The software ecosystem includes a proprietary developer toolkit with a Performance Profiler, supports TensorFlow 2.x and PyTorch through the ONNX format, and provides a complete software stack for building custom AI applications.
One Module, Dozens of Tasks
The LinQ HX accelerator is designed for a broad range of applications. In transportation and logistics, it can power autonomous vehicles operating in urban traffic. The platform can also serve as the computing core for autonomous robots capable of navigating complex environments and carrying out sophisticated tasks independently. In healthcare, it can enable AI-powered diagnostic systems and around-the-clock patient monitoring. Drones equipped with LinQ HX can inspect buildings, bridges, power transmission lines, and pipelines while identifying defects immediately. The module can also support multispectral urban monitoring, geological exploration, remote sensing, coordination of autonomous systems during search-and-rescue missions, and many other scenarios. In addition, LinQ HX is well suited for the Industrial Internet of Things (IIoT), where it can support predictive maintenance.
From Strategy to Product
HighTek has invested approximately 800 million rubles (about $10 million) in research, development, and commercialization of the project. The accelerator also aligns with Russia's broader technology strategy. The updated Strategy for the Development of the Russian Electronics Industry Through 2030 prioritizes the creation of a competitive domestic electronics base while expanding the adoption of AI technologies across the economy.
The relevance of HighTek's approach is reinforced by growing demand for embedded AI systems. According to Research Nester, the global embedded AI market is expected to reach $13.2 billion in 2026 and expand to $44.53 billion by 2035. Russia's AI accelerator market totaled approximately 63 billion rubles (about $820 million) in 2025 and, based on projected compound annual growth, could exceed 250 billion rubles (about $3.3 billion) by 2030.
The main growth drivers include the launch of new projects, modernization of existing production facilities, wider deployment of industrial AI inference, large-scale adoption of IIoT technologies, and growing demand for fully autonomous intelligent systems.
From Servers to Autonomous Systems
Until now, Russian developments addressed only part of the domestic market. NTTs Modul has introduced the NM Quad server neural accelerator, designed primarily for data centers and specialized computing platforms. The Elbrus processors developed by the Moscow Center of SPARC Technologies (MCST) can execute neural network algorithms, but they remain general-purpose processors with limited specialization for AI workloads. Meanwhile, the Baikal processor project developed by Baikal Electronics is still under development. Against that backdrop, LinQ HX fills a major gap in Russia's edge AI ecosystem while completing a full portfolio of domestic AI accelerators, strengthening the country's technological independence in this strategic field.
