Russia has unveiled LinQ HX, a domestically developed neural network accelerator designed specifically for embedded systems. Created by Hi-Tech, a resident of the Technopolis Moscow special economic zone, the module enables local AI model execution without relying on cloud resources-ideal for applications demanding low latency, energy efficiency, and independence from foreign hardware components.
LinQ HX made its debut at Innoprom 2026. According to its developers, the accelerator is built around a Russian microprocessor and employs locally sourced components, operating free of any foreign licensing restrictions. This is significant for edge AI applications, where neural networks often run directly inside cameras, robots, or medical devices, and supply chain autonomy is increasingly critical in practice, not just theory.
Unlike supercharged GPUs designed for data centers, LinQ HX is engineered for the ”smart edge.” It targets compact embedded systems where heat dissipation, noise, and power consumption must be tightly controlled, while still delivering millisecond-level response times. This niche is growing worldwide, with local AI inference becoming increasingly common in surveillance, industrial automation, and transportation.
LinQ HX neural accelerator specifications
- Performance up to 30 trillion operations per second
- Latency between 1.5 and 2.3 milliseconds
- Supports multiple neural networks running simultaneously
- Processes data locally without cloud connectivity
- Power consumption capped at 40 watts
- Compact form factor tailored for embedded systems
LinQ HX addresses three main use cases. First, video surveillance where event detection happens on-site without streaming footage to distant data centers. Second, robotics and industrial machinery requiring onboard AI inference. Third, healthcare devices for diagnostics and patient monitoring where low latency and autonomous operation are vital.
The edge AI hardware sector is expanding rapidly worldwide. Market research firm MarketsandMarkets projects the global edge AI hardware market will approach $6 billion by 2030. Growth drivers include surveillance cameras, autonomous vehicles, industrial automation, and medical technology. Established players in this segment include Hailo with their edge-focused accelerators, Qualcomm’s embedded AI platforms, and NVIDIA’s Jetson lineup, a go-to benchmark for robotics and smart camera developers.
LinQ HX takes a different approach: it prioritizes full localization and technological sovereignty over outright peak performance. While Russia has introduced home-grown processors and AI solutions before, a ready-made accelerator emphasizing complete domestic assembly and zero foreign licensing remains rare. Despite LinQ HX’s pricing and commercial availability timelines being undisclosed, it signals an intent to carve out a space where import substitution is proven through hardware integration, not just announcements.
Hi-Tech joined Technopolis Moscow in 2025 and has invested around 800 million rubles (approximately $11 million) into R&D so far. For a relatively young hardware developer, this level of investment covers chip design, embedded module refinement, thermal management, software stack development, and sector-specific certification-critical challenges beyond the processor architecture itself.
Looking ahead, LinQ HX is set to move beyond exhibitions into pilot deployments. Its true test will be integration ease-how simply AI models can be ported onto the module-and the total cost of ownership compared to imported alternatives. These factors usually distinguish a showpiece prototype from serial production hardware capable of expanding edge AI adoption.