In April 2026, I joined the APC project for LG Energy Solution battery manufacturing. It was the biggest turning point in my career.

I don’t know much about manufacturing yet, but being on-site, I realize the differences between manufacturing and the general IT industry. Legacy systems are often treated as technical debt that needs to be refactored away immediately in IT. However, the approach in manufacturing is quite different. In IT, carrying technical debt is a major risk. But in manufacturing, attempting to fix technical debt can itself be a risk. Any decision to change equipment or source code here is very dangerous. If you rush to fix technical debt and the process gets tangled, leading to a drop in factory utilization, the loss is immediate and direct. If manufacturing deadlines aren’t met, the company faces irreversible damage. Manufacturing has a heavy constraint that prevents tech debt from being simply ripped out.

All of LG Energy Solution’s production lines are for batteries. Recently, they have been actively producing ESS batteries for data centers, driven by the AI industry’s development. Major economies like the US are in an AI gold rush, with data center investment in the first half of 2025 surging 37% year-on-year, and shipments of related servers and power systems exploding by 64%. However, these giant data centers consume massive amounts of energy. This is where the virtual software world and physical production intersect powerfully. To support this ever-expanding energy-intensive AI infrastructure, paradoxically, these huge, clunky battery production lines must run incessantly.

The current global leader in battery production is CATL in China. CATL maintains its number one position by leveraging its massive domestic market and strong raw material supply chain. They are rapidly dominating the market with product lineups that prioritize price competitiveness. Under the full support of the Chinese government, they have vertically integrated their mineral supply chain to achieve overwhelming cost reduction, focusing on LFP batteries. LG Energy Solution is executing a strong localization strategy to respond to the US Inflation Reduction Act (IRA) by establishing large joint ventures in North America. Ultium Cells is a joint venture between LG Energy Solution and GM, with plants operating in Ohio (Phase 1) and Tennessee (Phase 2), plus Michigan Lansing (Phase 3).

The project I am part of is APC (Autonomous Process Control). It is a system that uses facility data and measurement/inspection data to perform automated corrections.

The electrode process, which determines battery performance, requires high precision. Currently, samples are taken periodically for physical measurement, but if a defect is found, a significant amount of defective products might have already passed the line, causing massive loss.

As mentioned, in manufacturing, loss is a blow to the enterprise. If the real-time facility data pouring out of coater or press equipment is fed into an AI model, quality can be predicted virtually without stopping the line.

This prediction data becomes the perfect input for the APC system I handle. APC receives virtual measurement results in real-time and proactively adjusts parameters if quality deviates from the target.

In manufacturing, directly stripping away legacy systems often exacerbates operational risk, creating a technical paradox. Instead of forcibly dismantling this debt, the industry is pivoting toward an AI-driven approach. By leveraging AI-learned algorithms to optimize risk and loss, the industry is bypassing technical limitations and advancing production efficiency.

The end of infrastructure competition, the real wealth is completed downstream. This is similar to containerization, which led the logistics revolution 15 years ago. The old containerization technology had a huge social impact but reinforced existing structures rather than creating new wealth. The pioneers didn’t become very rich, but the innovation laid the foundation for the East Asian export economy, offshoring, and retail giants like Walmart or Amazon, where most wealth was created downstream.

I believe there is a sign that AI is following the exact same pattern. The current market is obsessed with building super-large models and operating data centers. But this area is very capital-intensive and forced to suffer through difficult price wars. Rather than the few who own the actual infrastructure, far more opportunities will arise where structural changes cause value-added redistribution due to AI. The core implication of this huge trend is that investment and profit occur in the AI downstream sector.

As Korea’s economy is manufacturing-based, I hope the APC project and battery manufacturing site I am currently dedicated to will become the front line of AI-based downstream innovation.