A Chinese artificial intelligence company has cracked a critical safety problem plaguing data centers worldwide: detecting electrical fires before they ignite. Shenzhen HQVT Technology Co., Ltd., which went public on Hong Kong's Main Board in June 2026, reported that its multispectral vision platform successfully secured large-scale deployment contracts at major data centers operated by a Shanghai state-owned enterprise and a Shenzhen-listed company during the first half of 2026. The technology works by extending detection capabilities beyond what human eyes can see, monitoring the invisible thermal and ultraviolet signatures that precede catastrophic electrical failures.

Why Data Centers Need a New Kind of Fire Detection?

Traditional smoke detectors fail in modern data centers because the forced-air cooling systems designed to keep servers running actually dilute smoke before it can trigger alarms. This creates a dangerous blind spot: electrical arcing, insulation degradation, and localized thermal spikes can develop undetected until they become fires. HQVT's approach bypasses this problem entirely by monitoring the physical signatures of electrical stress long before combustion occurs. The company's proprietary "Super Eye" perceptual platform detects faults across three distinct light spectrums: visible light (380 to 780 nanometers), ultraviolet (10 to 380 nanometers), and thermal infrared (780 nanometers to 1 millimeter). This multi-layered sensing catches micro-arcing and dielectric degradation that traditional detectors simply cannot see.

How Does HQVT's Three-Tier Defense System Work?

• Tier-1 Rack-Level Monitoring: Internal sensors embedded within server racks detect electrical stress at the component level, catching faults in power distribution units and cooling systems before they spread.
• Tier-2 Room-Level Spatial Sensing: Multispectral cameras mounted throughout the data center floor monitor the broader environment, detecting thermal anomalies and ultraviolet signatures across multiple racks simultaneously.
• Tier-3 Autonomous Robotic Patrol: Quadruped robots equipped with multispectral AI modules conduct contact-free, online inspections of high-voltage substations, cable interlayers, and battery storage systems, eliminating human exposure to hazardous environments.

The system processes all this sensory data through HQVT's proprietary "Zhiyuan Origin Large Model," an artificial intelligence foundation model that has completed dual regulatory filings with China's Cyberspace Administration and secured official model service provider qualification from Shenzhen. This AI layer transforms raw sensor data into actionable safety alerts, enabling instantaneous response to emerging faults.

What Changed: Moving AI Detection to the Edge?

A major technical breakthrough during the first half of 2026 was HQVT's successful migration of its foundation model from centralized server clusters to compact edge AI terminals powered by NVIDIA's Jetson AGX Orin platform. This shift matters because it eliminates two major obstacles to widespread adoption: the need for expensive on-premise compute servers and the latency introduced by sending data to cloud servers for analysis. By processing multispectral data locally on compact edge devices, HQVT's system responds to electrical faults in real time, without waiting for cloud processing. This architectural change also dramatically lowers the initial capital expenditure required to deploy the system, transforming HQVT's commercial model from serving only large state-owned enterprises to reaching broader commercial markets and small and medium-sized businesses.

The financial results underscore the commercial momentum behind this technology. HQVT reported consolidated interim operating revenue of 410.5 million Chinese yuan (RMB) for the first half of 2026, representing an 84.5% year-over-year increase. The primary growth engine was multispectral AI foundation model services, which surged 382.5% year-over-year to 320.0 million RMB, now accounting for the majority of the company's revenue. Adjusted net profit reached 27.6 million RMB, up 21.9% compared to the prior-year period, while cash used in operating activities narrowed by 50.2% year-over-year.

Beyond data centers, HQVT has expanded into electrical power systems and new energy infrastructure. The company entered into a strategic collaboration agreement with Deep Robotics to integrate multispectral AI modules into quadruped robotic inspection platforms for high-voltage substations, cable interlayers, and distributed battery energy storage installations. This diversification reduces dependence on any single market while addressing a broader infrastructure safety challenge: the electrical grid itself faces similar detection blind spots as data centers.

What Does This Mean for Global Data Center Expansion?

Data center operators worldwide face mounting pressure to expand capacity to support artificial intelligence workloads, yet communities increasingly oppose new facilities due to power consumption and environmental concerns. A proven, scalable fire detection system addresses one critical safety objection that regulators and local governments raise during facility approval processes. By demonstrating that advanced multispectral AI can prevent catastrophic electrical fires more reliably than traditional methods, HQVT's technology could remove a significant regulatory barrier to data center expansion in regions where fire safety compliance is stringent.

Looking ahead, HQVT is directing its IPO capital toward expanding production capacity, advancing edge-model iterations, and executing an international go-to-market strategy. The company has partnered with Singapore-based LINKWISE to accelerate deployments across Southeast Asian data centers and is actively preparing channel rollouts in the Middle East and Europe. By pairing standardized edge perception hardware with cloud-based Software-as-a-Service (SaaS) subscription tiers for analytics and diagnostic reporting, HQVT is positioning itself to build recurring, high-margin international revenue streams within the global physical AI safety sector.

Soochow Securities, in its inaugural coverage report dated July 31, 2026, assigned HQVT a "Buy" rating with a target price of 58.57 Hong Kong dollars, implying approximately 113% upside from the September 4 closing price of 27.42 Hong Kong dollars. The broker's valuation is anchored on 2027 projected revenue of 1.31 billion RMB at a 32x price-to-sales multiple, citing the stock's scarcity value as the only Hong Kong Main Board-listed multispectral physical AI pure-play and the industry's 40% or higher compound annual growth rate. The report notes that the edge AI breakthrough serves as the single largest near-term catalyst, with commercialization orders representing the critical verification point for investors.

HQVT's interim results approval by the board on August 28, 2026, and the September 7 Stock Connect effective date represent the two immediate catalysts for investor attention. The Stock Connect inclusion will enable mainland Chinese investors to trade HQVT shares through their Stock Connect accounts, significantly broadening the company's investor base and enhancing trading liquidity.