China’s AI-Powered Radar Takes On Drone Swarms as US-Iran War Drives New Detection Race

The Iran conflict is doing more than destroying infrastructure and tankers. It’s stress-testing air defense systems in real time, and Beijing is watching every radar return very carefully.

China has publicly confirmed it is integrating AI algorithms into its air defense radars to better detect low-altitude drone swarms. Xu Jin, chief engineer for early warning and detection at the 38th Research Institute of China Electronics Technology Group Corporation, made the announcement as a CPPCC member attending Beijing’s annual “two sessions” political meetings — speaking to the South China Morning Post on the sidelines. That platform matters. The US-Iran conflict pushing drone swarm tactics into prime time at the same moment was, at a minimum, convenient timing.

• The Development: China’s 38th Research Institute has tested an AI algorithm that delivered an “unexpected boost” to radar target detection, specifically against large numbers of low-altitude drones operating in swarm formations.
• The Driver: The ongoing US military campaign against Iran, which both sides have prosecuted heavily with low-cost suicide drones, is the stated real-world reference point for this research push.
• The Problem It Solves: Traditional radar systems struggle to distinguish, track, and fully detect large numbers of coordinated, inexpensive drones simultaneously — a gap that swarm tactics deliberately exploit.
• The Source: South China Morning Post interview with Xu Jin, published during the two sessions meetings in Beijing.

Drone Swarms Are Breaking Traditional Radar’s Processing Limits

Traditional air defense radar was designed around a different threat model: a limited number of fast, high-value targets. Drone swarms invert that entirely, flooding radar processing pipelines with dozens or hundreds of slow, cheap, low-observable signatures at once. Xu Jin’s team at the 38th Research Institute — the organization that built China’s first low-altitude early warning and detection radar — has concluded that conventional signal processing can no longer keep up.

“When inexpensive suicide drones are deployed, their numbers are usually quite large, making them difficult to distinguish and completely detect,” Xu said. “Some of them may also have the ability to coordinate and divide tasks. This puts enormous processing pressure on traditional radar detection.”

That’s a precise description of what the US-Iran conflict has demonstrated since late February. Iran’s Shahed-136 salvos against Gulf targets and US retaliatory strikes have both leaned on sheer numbers to overwhelm defense layers. The friendly fire incident over Kuwait, where three US F-15E Strike Eagles were shot down during an Iranian drone defense engagement, is the sharpest evidence yet of how chaotic mass drone engagements get for any detection and intercept system.

The 38th Research Institute’s AI Fix

China Electronics Technology Group Corporation’s 38th Research Institute is one of China’s primary air defense radar development bodies. Its role in building the country’s first low-altitude early warning and detection radar gives it a specific institutional focus on exactly the detection gap drone swarms expose: the terrain-hugging, slow-moving, radar-cross-section-minimal threat profile that commercial and military drones share.

The AI algorithm Xu describes works on the detection processing layer — essentially improving the radar’s ability to separate real drone returns from noise and clutter when dozens of targets appear simultaneously. The “unexpected boost” language is notable. It suggests the performance gain exceeded what the team modeled before deployment in tests, which could mean the algorithm is finding signal patterns that classical processing discards as ambiguous.

This isn’t China’s first move to embed AI into its drone-related military systems. The PLA’s integration of DeepSeek AI across drone swarms and autonomous combat vehicles showed Beijing moving quickly to apply large model inference to battlefield systems. The radar work is the defensive side of the same coin: if your adversaries are using AI to coordinate swarm attacks, your detection systems need AI to parse the resulting radar picture.

The Broader Counter-Drone Detection Race

China is not alone in chasing this problem. The counter-drone radar market has seen significant private and government investment across multiple countries. Chaos Industries raised $510 million at a $4.5 billion valuation specifically to scale its counter-drone radar technology. NATO’s Bold Machina exercise tested a counter-drone prototype developed by the Naval Postgraduate School aboard a Dutch Navy vessel, drawing on sensors and systems from the US, Norway, Sweden, and Australia. Taiwan built the T-Dome system with Chinese drone incursions explicitly in mind.

What’s different about China’s announcement is the specific acknowledgment that current radar architectures are structurally inadequate for swarm scenarios — and that the gap was confirmed by live combat data from the US-Iran theater. That’s a rare admission from a state-controlled research body, and it signals that Beijing considers this deficiency urgent enough to say publicly at a politically sensitive moment like the two sessions.

The New York Times editorial board’s warning last December — that AI drone swarms could hunt and kill autonomously, and that the Pentagon was losing war games simulating such scenarios — frames the same problem from the American side. Every major military is now in a race to figure out how to detect, track, and engage drone swarms before they find a way to saturate defenses entirely.

What the Iran War Is Teaching Everyone About Drone Defense

The US-Iran conflict has compressed years of theoretical drone defense research into weeks of live data. Operation Epic Fury’s debut of the LUCAS drone — a US weapon derived from a captured Shahed airframe and rebuilt with American electronics and AI flight systems — showed that offensive swarm capability is now accessible to any well-resourced military. The economics are impossible to ignore: a $2,500 interceptor drone versus a $3 million Patriot PAC-3 missile is a cost ratio that breaks any conventional defense budget at scale.

China’s radar AI initiative is a direct response to that arithmetic. If you can’t afford to shoot down every drone with expensive interceptors, you at least need radar that can accurately sort and prioritize the swarm. That’s what Xu Jin’s team is building.

DroneXL’s Take

The most telling detail in Xu Jin’s interview isn’t the AI boost itself. It’s that he’s talking about it publicly, as a CPPCC member with a platform, and specifically citing the US-Iran conflict as the reference case. Senior Chinese research officials don’t use two sessions media access to announce incremental laboratory results. This reads as a deliberate signal that Beijing sees drone swarm saturation as an imminent challenge to its own air defense architecture, not just something happening to other people in the Gulf.

I’ve covered the swarm detection problem since the NYT editorial board piece last December, and the consistent thread is that radar systems designed for Cold War threat profiles — fast jets, ballistic missiles — are being asked to do something they were never built for. AI is the obvious fix because the problem is fundamentally one of pattern recognition at scale, which is exactly what large model inference handles well.

My prediction: by the end of 2026, at least two NATO members will publicly announce AI-augmented radar procurement specifically framed around swarm detection, citing the Iran conflict as justification. The war in the Gulf isn’t just reshaping Middle East geopolitics. It’s rewriting air defense procurement specs globally, and China just confirmed it’s rewriting its own.

Editorial Note: AI tools were used to assist with research and archive retrieval for this article. All reporting, analysis, and editorial perspectives are by Haye Kesteloo.