Shutting down a blast furnace and building new facilities fit into the same sentence, but on the accounting books they are entirely different items. One is the early retirement of assets whose depreciation is not finished, and the other is new investment with a payback period of more than ten years. The debate over steel decarbonization has gone off track because these two items have been discussed as a single lump. The stage of arguing over the principles of low-carbon steelmaking technology has passed, and the remaining issue is who puts up the money to move those principles into facilities.
Three pieces of news came in the same week. The triple burden on the steel industry, which faces weak demand, rising costs and trade pressure at the same time, a statement by an industry group calling for the sharing of transition costs, and progress in a demonstration project that uses artificial intelligence to cut emissions. The three items, which look different in character, converge on one point. The means of reduction are increasing, while the financial capacity to adopt them is shrinking.
A policy that demands only structural improvement from companies without designing financial support brings forward industrial contraction instead of bringing forward reductions. There are two ways for emissions to fall. Cutting them by changing facilities, and having them fall because production itself declines. Both are counted as reductions in the statistics, but the latter leaves behind no jobs, no tax revenue and no follow-up investment.
The logic on the other side is not light either. Subsidies are ultimately a burden on the public, and cases of keeping uncompetitive facilities alive with taxes have been repeated in many industries. If support becomes permanent, companies have an incentive to delay the timing of reductions. This point is valid.
The reason the conclusion does not change nonetheless is that steel's emissions do not come from the management choices of individual companies but from the process itself. Unlike emissions that can be cut by adjusting supply chains, as in autos or electronics, the carbon dioxide from a blast furnace does not fall unless the reduction method is changed altogether. What is needed, therefore, is conditional support. If payments are tied to reduction performance and the rate of facility conversion, with clawback when targets are missed, concerns about moral hazard can largely be managed.
This is also where the artificial intelligence-based reduction demonstration carries meaning. Optimization using process data requires less input and pays back faster than large-scale facility replacement. It can serve as a buffer that holds emissions down in the stretch before full conversion. It is hard to see this alone as reaching the target, and it also needs to be confirmed whether the demonstration results can be reproduced at commercial scale.
The trade environment does not buy time either. As carbon regulation tightens in export markets, low-carbon production capacity is closer to a right of market access than to an environmental indicator. Facilities that fail to convert now will be far more expensive to turn around after regulation takes full effect. The price of delay grows as time passes.
What should be watched, then, is the text of the support scheme rather than the reduction target figures. How the recipients and timing of payments, the method of checking compliance, and the clawback conditions for non-compliance are written will determine the actual pace of the transition. A plan that raises targets alone and leaves the funding design empty puts pressure on industry without cutting emissions. The temperature of molten iron is more honest than the sentences in a policy document.
