For decades, operating a steel converter was an art form. Veteran workers, standing near furnaces hot enough to melt iron at 1,600°C, relied on intuition, experience, and the color of the flame to make split-second decisions about temperature, carbon content, and timing. The process was a "black box"—too extreme for sensors, too complex for simple automation, and entirely dependent on human judgment honed over years. Today, inside the smart operations center of Nanjing Steel Group, giant screens display real-time data on converter temperatures and steel grades. An AI model developed with Huawei predicts furnace conditions with over 90% accuracy, boosting the first-grade rate of hot metal by more than 10 percentage points and slashing energy consumption per unit by 12%. The black box has been opened—and the results are transforming the world's largest steel industry.
The Scale of the Challenge
Understanding why AI matters for steel requires understanding the scale of China's steel industry. China produces over 1 billion tons of crude steel annually—more than the next 10 steel-producing nations combined. The industry employs millions, consumes enormous amounts of energy, and accounts for roughly 15% of China's total carbon emissions, making it the second-largest emitting sector after power generation. Any efficiency improvement in Chinese steelmaking has global implications for energy consumption, carbon emissions, and industrial competitiveness.
By the end of 2025, 227 Chinese steel enterprises had achieved full-process ultra-low emissions, covering 844 million tons of crude steel capacity. Another 40 enterprises had completed partial upgrades covering 88 million tons. The industry's "Extreme Energy Efficiency" program has already delivered energy savings equivalent to over 24 million tons of standard coal, reducing carbon dioxide emissions by approximately 60 million tons. These are substantial achievements, but the industry's leadership recognizes that further gains require a fundamentally different approach—one that artificial intelligence is uniquely positioned to provide.
Breaking Open the Black Box
The core challenge of steelmaking AI is that the most critical processes happen inside vessels where conditions are too extreme for direct measurement. A basic oxygen furnace operates at temperatures exceeding 1,600°C, with chemical reactions occurring in seconds. Traditional automation could handle the simple parameters—blow oxygen for X minutes, add Y kilograms of alloy—but the subtle interactions between temperature, carbon content, phosphorus levels, slag formation, and dozens of other variables were beyond the reach of conventional control systems.
AI changes this by treating the furnace as a prediction problem rather than a control problem. Machine learning models trained on thousands of historical heats, combined with real-time sensor data from thousands of internal monitoring points, can predict where the process is heading before it gets there. Nanjing Steel's system, powered by a specialized large language model, dynamically predicts furnace conditions, enabling operators to make adjustments proactively rather than reactively. The result is not just better quality—it's a fundamentally different relationship between human operators and industrial processes.
The impact extends across the entire steelmaking chain. In Inner Mongolia, Bao Gang United Steel has deployed a fully closed-loop AI smelting system that transitioned from manual judgment to autonomous algorithmic control. The system increased endpoint carbon-temperature hit rates by over 15% and shortened the average smelting cycle by 1.5 minutes per heat. Multiply 1.5 minutes across thousands of heats per year, and the cumulative productivity gain is substantial—not to mention the reduced oxygen and raw material consumption.
Lighthouse Factories and AI Agents
The most dramatic illustration of AI's impact on steelmaking comes from Beijing Shougang Co., Ltd., whose cold-rolling subsidiary earned a coveted "Lighthouse Factory" designation from the World Economic Forum in early 2025. The Lighthouse program recognizes factories that have successfully integrated Fourth Industrial Revolution technologies—AI, IoT, big data, cloud computing—at scale. Shougang's achievement demonstrated that steelmaking, long considered a "dirty" heavy industry, could achieve the same level of digital sophistication as electronics manufacturing or automotive assembly.
Since the start of 2026, Shougang has deployed 45 AI agents across its operations and implemented 929 intelligent application scenarios. These aren't theoretical concepts or pilot projects—they are production systems generating measurable economic value. The AI agents handle everything from quality inspection (computer vision systems that detect surface defects invisible to the human eye) to production scheduling (optimizing the sequence of orders to minimize changeover time and energy consumption) to predictive maintenance (anticipating equipment failures before they cause downtime).
China Baowu Steel Group, the world's largest steelmaker with annual crude steel output of approximately 130 million tons, has taken the AI transformation even further. In 2023, Baowu achieved a "lights-out factory"—a fully automated production facility operating around the clock with minimal on-site human intervention. In 2026, the company launched its AI 2.0 strategy, aiming to integrate artificial intelligence across all four pillars of its business: research and development, manufacturing, management, and customer service. The company has deployed AI-powered "converter furnace foremen" and "cold rolling operators" trained on decades of production management expertise, combining big data with large language models to optimize scheduling, improve resource efficiency, and enhance safety.
The Carbon Connection
AI's role in steel is not just about efficiency—it's about survival. China has committed to peak carbon emissions before 2030 and achieve carbon neutrality by 2060. The steel industry, as the second-largest emitter, faces existential pressure to decarbonize. AI is proving to be one of the most effective tools available, not by replacing existing processes but by optimizing them to levels that were previously unattainable.
The China Iron and Steel Association (CISA) has launched a three-year digital transformation initiative that explicitly ties AI adoption to carbon reduction goals. The logic is straightforward: AI-optimized processes use less energy per ton of steel produced, which directly reduces carbon emissions. But AI also enables more sophisticated decarbonization strategies, such as optimizing the integration of hydrogen-based direct reduction ironmaking with traditional blast furnace routes, or managing the complex energy flows of a steel plant that combines grid electricity, self-generated power, waste heat recovery, and renewable sources.
Baowu is pursuing a dual-track strategy: digitizing existing processes for immediate efficiency gains while investing in breakthrough technologies like hydrogen-based shaft furnaces that could fundamentally transform steelmaking's carbon footprint. The hydrogen route, being piloted at Baowu's Zhanjiang facility, can dramatically reduce carbon emissions compared to traditional blast furnaces—but it requires sophisticated process control that only AI can provide at industrial scale.
Robots on the Factory Floor
The AI transformation of steel is not limited to software and algorithms. Physical robots are increasingly taking over the most dangerous jobs in steel plants. At the 2026 World AI Conference, Baowu's subsidiary Turin Robotics showcased a robot designed specifically for steel converter sampling—a task traditionally performed by workers carrying steel sampling rods near furnaces radiating extreme heat. The robot keeps humans away from high-temperature hazardous areas while improving sampling consistency and frequency.
These robots are not generic industrial arms repurposed for steelmaking. They are purpose-built for the unique challenges of steel plants: extreme temperatures, heavy dust, electromagnetic interference, and the need for millisecond-level precision in positioning. The integration of AI vision systems allows these robots to adapt to variations in furnace conditions, positioning themselves based on real-time visual data rather than pre-programmed coordinates.
The broader trend is toward what Baowu calls "scenario-platform-model-data-computing" integration—a systematic approach that treats AI not as a collection of point solutions but as a comprehensive capability spanning every aspect of steel production. The vision is a fully connected steel plant where every furnace, every rolling mill, every quality inspection station, and every logistics vehicle is part of a single AI-orchestrated system.
Industry-Wide Implications
The steel industry's AI transformation carries implications beyond China's borders. The technology and operational models being developed in Chinese steel plants are increasingly being exported. Baowu has operations in more than 20 countries and maintains technology exchanges with major steelmakers in Japan and South Korea. The AI systems that work in China's steel plants can be adapted to steel plants anywhere, creating a potential competitive advantage for Chinese steel technology providers.
In June 2026, China opened a national AI pilot base for the metallurgical sector in Nanjing, designed to accelerate the transition from testing to large-scale industrial deployment. The base serves as a shared resource for the entire industry—a place where steel companies can test AI applications without the risk and expense of deploying them directly in production environments. This collaborative approach, unusual in an industry known for fierce competition, reflects a recognition that the AI transformation is too complex and too important for any single company to tackle alone.
Li Yiren, vice president of CISA, captured the broader significance when he noted that the value of AI lies "not merely in improving the efficiency of a single position, but in redefining the entire value chain—from R&D and manufacturing to procurement, logistics, and marketing." Data is becoming a new factor of production in steelmaking, and AI is becoming a key tool in the production process itself. The transformation is not about replacing steelworkers with machines—it's about giving steelworkers capabilities that no amount of experience could provide.
China's steel industry, long stereotyped as a landscape of billowing smoke and brute-force labor, is proving that even the heaviest of heavy industries can be smart, sophisticated, and increasingly green. The algorithms running inside Chinese steel mills today are not just optimizing production—they are rewriting the rules of what steelmaking can be.