002236 · SZSE · China
dahuatech.comFinancials as of FY2024 · latest on file
Builds surveillance cameras with AI built directly into the hardware, designed in Hangzhou so the software and sensor work as one unit.
Zhejiang Dahua Technology builds surveillance cameras in Hangzhou where the computer vision algorithm and the CMOS sensor inside each unit are designed together, so the AI processing runs directly on the camera's embedded chip rather than on a separate server. Because each algorithm is compiled against one specific sensor's noise floor and pixel geometry, swapping in a different sensor means rewriting and revalidating the algorithm from scratch — the camera, the firmware, and the analytics are a single unit, not interchangeable parts. Every new camera deployed makes switching harder, since the video management software learns from footage produced by that exact sensor-algorithm pair, and replacing cameras mid-deployment means months of model retraining per site before the system can understand the new hardware's output. The whole architecture depends on a steady supply of CMOS sensors from foundries in Taiwan and South Korea, because the algorithm IP was compiled for those specific sensor geometries — if that supply were cut by export controls or geopolitical disruption, no substitute sensor exists for which the co-design has already been done, and the integrated architecture could not simply be transferred to a different foundry.
How does this company make money?
The company earns money when it sells cameras, NVRs, and complete surveillance systems to distributors and system integrators, who then supply end customers. It also charges licensing fees for the AI analytics software embedded in the hardware. On top of that, it collects recurring fees from support contracts that cover firmware updates and ongoing technical help.
What makes this company hard to replace?
Switching means replacing the entire video management software system, not just the cameras — there is no upgrade path that keeps the old software running on new hardware. Every installed site has also built up months of AI training on footage from the existing cameras; a new vendor's hardware produces different-looking footage that the system does not yet understand, requiring months of retraining. Government customers face an additional barrier: security certifications are tied to a specific vendor, so switching means going through a full re-qualification process from the beginning.
What limits this company?
The company can build more software and assemble more units, but it can only ship as many cameras as Sony and Samsung choose to allocate CMOS sensors to it. That allocation is negotiated deal by deal — no amount of factory investment buys a guaranteed supply.
What does this company depend on?
The company cannot operate without CMOS image sensors from Sony and Samsung, embedded ARM processors, clean-room manufacturing facilities in Hangzhou, vendor-specific ISO 27001 security certifications required for government sales, and its own proprietary computer vision algorithm IP developed in-house.
Who depends on this company?
Chinese public security bureaus rely on it for facial recognition and traffic monitoring — if the company stopped, those AI analytics capabilities would go dark. International system integrators managing multi-site surveillance installations would face full hardware replacement to keep their software working. Smart city projects in Belt and Road countries would need to completely re-engineer their integrated camera-and-analytics setups.
How does this company scale?
Once a computer vision algorithm is written and validated, it can be copied into millions of camera units at almost no extra cost. What does not scale automatically is sensor supply — every increase in production volume requires renegotiating foundry allocations with Sony and Samsung individually, and no amount of extra money removes that constraint.
What external forces can significantly affect this company?
The company is on the U.S. Entity List, which blocks it from buying American semiconductor components and software tools. The Chinese government also pushes for domestic surveillance infrastructure to rely on Chinese technology, which shapes what the company can use internally. In Europe, laws like GDPR restrict how biometric data — including the facial recognition these cameras perform — can be collected and stored, limiting what the cameras are legally allowed to do in those markets.
Where is this company structurally vulnerable?
If export controls expanded or a geopolitical disruption cut off CMOS sensor supply from foundries in Taiwan and South Korea, the company's algorithm library would be stranded — it was written for those specific sensors, and no substitute sensor exists for which that work has already been done. Rebuilding the co-design from scratch for a different sensor would take years.
Sign in to view price data.
Sign inScreen for dividend patterns
Find other stocks with similar dividend characteristics in the screener.
As of FY2024 (year ended December 31, 2024). Newer annual figures aren't yet on file.
Structural observations derived from financial data, industry benchmarks, and supply chain position.
Companies that share the same coordination system — how they create, deliver, or capture value.
Companies that share active interpretations — structural patterns currently present in both stocks.
CompanyGraph