Chinese army adopts DeepSeek to integrate AI into battlefield operations.

China is standardizing on DeepSeek across parts of the PLA, with Reuters documenting procurement, tests, and early field uses that include autonomous ground support and decision support at the edge. The shift compresses staff work, links AI to mobile carriers like the CS/VP16B, and could quicken the detect, decide, act loop across Indo-Pacific flashpoints.

China’s military is not waiting for a leap of faith; it is building a repeatable software stack around DeepSeek and then bolting it to real platforms. Reuters reports PLA entities buying and testing DeepSeek for autonomous reconnaissance and real-time decision support, while state giant Norinco has shown an autonomous support vehicle operating at maneuver speed and presented as DeepSeek powered. Tie that to a light 6x6 carrier like the Lynx-family CS/VP16B shown at IDEX 2025, and the picture is a methodical move from lab demos to deployable kits that ride with motorized sub-units and feed the common operational picture.
Follow Army Recognition on Google News at this link

The Chinese CS/VP16B 6x6 light unmanned all-terrain vehicle for battlefield support and logistics has a 200 km cruising range, including 20 km in pure electric mode. (Picture source: Army Recognition)

The PLA concentrates its effort in three structuring directions. First, integration on land platforms dedicated to close support. Norinco presents a support vehicle with autonomous operation at 50 km/h, keeping an uncrewed carrier within the maneuver speed of motorized sub-units. Second, standardization around a single DeepSeek model family in public tenders and technical filings, creating a coherent stack that reduces training costs and eases software maintenance. Third, compression of staff-work timelines. Academic work linked to the Chinese ecosystem claims the evaluation of 10,000 scenarios in 48 seconds, where a human team would require hours, even dozens of hours. Taken together, these elements sketch a pragmatic scale-up of perception and decision-support functions.

The CS/VP16B presented by Norinco at IDEX 2025 serves here as an example of a carrier. As part of the Lynx line, this compact 6x6 prioritizes modular payloads and mobility, with a chassis designed for light transport, reconnaissance, and auxiliary fire roles depending on configuration. This mechanical base, already fielded in multiple variants, provides the volume, interfaces, and stability required to host autonomous navigation sensors, ruggedized compute, and short-range data links. In that perspective, coupling to a decision core such as DeepSeek enables edge-side image triage, obstacle detection, micro-route planning, and prioritization of alerts forwarded to higher echelons.

Three parameters illustrate the intended employment envelope. First, autonomous speed at 50 km/h, as previewed by Norinco on a support vehicle, keeps an uncrewed ground vehicle in step with mechanized maneuver without breaking the cadence of short-range convoys. Second, the CS/VP16B’s 6x6 architecture supports negotiation of low obstacles and soft ground, opening last-mile routes to outposts. Third, short-range links and prioritized image transmission address a recurring problem, preserving useful throughput under radio congestion by sending only detected and qualified events from as close to the sensor as possible. These choices, modest in isolation, combine to save minutes in the detect, decide act loop, which is decisive for feeding a battery fire mission or extracting a casualty.

At the tactical and operational level, a CS/VP16B-type chassis offers three immediate contributions. First, contact-range logistics. It replenishes ammunition, water, and batteries with low human exposure and a reduced acoustic and thermal profile when propulsion shifts to a discreet mode. Second, mobile sensing. It navigates autonomously in narrow streets or woodland, probes technical corridors, places small charges to open a passage, and seeds a temporary sensor mesh that enriches the local picture. Third, data gateway. The vehicle collects, filters, and pushes relevant items to the common operational picture (COP) while maintaining emissions control to reduce vulnerability to interception. The performance center of gravity is resilience under degraded GNSS, algorithm robustness in poor visibility, and the ability to fall back to degraded driving rules when perception is partial.

On semiconductors, the landscape remains dual. US controls from September 2022 limit access to A100 and H100, yet open traces show Chinese teams have continued to draw on Nvidia hardware in research while the domestic line progresses and equips priority deployments. The issue is technical rather than doctrinal. Model portability, compiler maturity, and thermal management dictate the split between edge compute and offboard compute. For a CS/VP16B-class carrier, power allocation across perception, communications, and propulsion determines endurance and the ability to sustain a section’s pace.

This trend is not confined to ground platforms. Reuters places the whole in a broader competition, where the same core technology supports drone swarms, scouting robot dogs, and planning tools. Beijing’s objective is to secure a sovereign chain from algorithm to silicon, then disseminate homogeneous components across multiple vectors. Regional adversaries and partners respond with denser short-range air defense, distributed architectures, and cost-controlled counter-swarm measures. The defense industrial and technological base in Asia and Europe sees rising demand for sensors, edge processors, and software-defined radios that ensure interoperability without relying on fragile nodes. If the Chinese ecosystem stabilizes this combination of light chassis and onboard processing, the operational tempo in the Indo-Pacific tightens and crisis management margins narrow.