ARLINGTON, Va.--()--DeepSig, a pioneer in AI-native wireless communications, will showcase its latest advancements in AI-driven radio access networks (RAN) at Mobile World Congress (MWC) 2025 in Barcelona alongside the AI-RAN Alliance demonstrations. The over-the-air demonstration, as part of the AI-for-RAN Working Group of the AI-RAN Alliance, highlights how an AI-native air interface design, powered by the NVIDIA AI Aerial platform, enhances spectral efficiency, expands network capacity and delivers real-time adaption alongside a site-specific RAN digital twin—key pillars of AI-native 6G wireless systems.

Today’s RAN air interface relies on fixed, static designs that limit network adaptability and efficiency. Traditional methods—such as rectangular QAM modulation schemes and dedicated demodulation reference signals—restrict real-time wireless optimization. DeepSig’s AI-native air interface solution eliminates these constraints with a jointly learned modulation and receiver approach, optimizing for site-specific conditions in real-time using pilot-free transmissions. This reduces overhead, enhances interference resilience and further increases spectral efficiency and capacity, leading to measurable gains in network performance that lay the groundwork for AI-native 6G networks. Furthermore, this significantly reduces the Radio Unit (RU) power consumption.

“By embedding AI directly into the physical layer (PHY), DeepSig’s OmniPHY unlocks smarter, faster and more efficient networks,” says DeepSig CEO Jim Shea. “This innovation isn’t just about optimizing 5G; it’s about rethinking how wireless networks operate and accelerating the path towards AI-native 6G, where deep learning-driven air interfaces become the standard.”

AI in Action at MWC 2025

At MWC2025, DeepSig will demonstrate:

  • AI-Native Air Interface: Demo of DeepSig’s AI-native air interface that illustrates how a learned air interface technique outperforms the 5G-NR PHY. This over-the-air demo delivers measurable spectral efficiency and throughput improvements over traditional 5G approaches. Preliminary results reveal throughput increase due to reduced pilot overhead and an improvement in link margin, showing the approach can simultaneously deliver capacity and power reduction. It is the first time a pilotless air interface has been implemented over the air.
  • Digital twin simulations with NVIDIA Aerial Omniverse Digital Twin: AI-driven ray-tracing simulations show how AI-native RAN outperforms traditional wireless architectures in urban environments by optimizing spectrum use. DeepSig’s approach leverages NVIDIA’s AI-native simulation environment to validate AI-driven RAN performance in virtual environments before deployment in live networks, reducing costs and accelerating adoption.
  • Real-time performance metrics: Interactive visualization of key network performance indicators, including throughput and error rate impacts.

Accelerating AI-Native RAN with NVIDIA

DeepSig’s OmniPHY AI-Native modem stack is optimized for SuperMicro ARS-111GL-NHR based on the NVIDIA GH200 Grace Hopper Superchip on the network side and NVIDIA Jetson Orin NX system-on-modules for small form factor edge devices, leveraging the NVIDIA AI Aerial platform stack components to enable GPU-optimized L1 processing. By leveraging the Epiq Solutions Matchstiq G40 radio, DeepSig enables an integrated real-time, over-the-air demonstration of the technology.

“AI-RAN is one of the most promising and transformative areas of research and development for the telecom industry, helping deliver more powerful connectivity for today’s society,” said Soma Velayutham, Vice President Telecoms, NVIDIA. “DeepSig’s learned air interface implementation based on NVIDIA AI Aerial is a breakthrough innovation in this domain that uses a novel site-specific and pilotless transmission approach to unlock new levels of capacity, spectral and power efficiency — setting a new standard for sustainable and AI-native wireless networks of the future.”

Meet DeepSig at MWC 2025:

  • DeepSig and NVIDIA Demo Location: ARM Booth #2I60
  • Private Meetings: DeepSig Meeting Room in Hall 2—2J17MR. To schedule a personalized demo, contact info@deepsig.ai.

About DeepSig, Inc.

DeepSig is a pioneer in AI native wireless communications. Its transformative technology pushes the boundaries of spectrum sensing, wireless performance and network capabilities. Drawing on a unique blend of expertise in deep learning, wireless systems and signal processing, DeepSig’s AI/ML-powered software enhances security, efficiency and capacity for tactical and commercial wireless communications in licensed or shared radio spectrum and in existing 5G Open RAN or AI native next generation networks.