Not every AI job lives on a screen

Models, chips and developers dominate the AI employment story. But AI services depend on data centers, and data centers stop being useful when power or cooling fails. More AI computation therefore means more physical infrastructure.

Korea’s AI-data-center plans also include power and cooling solutions as industrial priorities. AI investment is not only an IT hiring story; it reaches into power and facility labor markets.

US data centers and power companies are already looking at the same workforce

Deloitte found that postings in 39 occupations shared by US power companies and data centers rose 64% in the data-center sector from 2023 to 2025. Electrician postings at data centers rose more than 180%.

In a 2025 survey, 63% of data-center executives identified shortages of relevant skilled labor as their biggest talent barrier. Those US figures are not Korean shortage rates, but they show where labor competition can emerge as infrastructure investment accelerates.

Why power, cooling and facilities become core AI work

Large AI workloads create major and sometimes volatile power demand. High-density computing also makes thermal management harder, increasing the importance of liquid cooling and other facility upgrades.

Reliable AI infrastructure therefore depends on switchgear, UPS systems, backup generation, power quality, cooling, BMS/DCIM monitoring, maintenance and incident response operating as one system.

The transferable signal is not the certificate name but experience in systems that cannot stop

A qualification alone does not guarantee a salary premium. What matters is the scale and criticality of the environment someone has operated.

People from semiconductor fabs, hospitals, financial data centers, energy facilities, factory facilities and other high-availability environments may carry transferable experience in redundancy, preventive maintenance, incident recovery and commissioning even if their previous title did not say “data center.”

A career transition may require operational context more than “learning AI”

A power or facilities professional does not necessarily need to become a machine-learning engineer to move into AI infrastructure. The first task is translating existing capability into mission-critical data-center language.

Redundancy, UPS and emergency power, BMS/DCIM, high-density cooling, incident procedures and commissioning are often more relevant bridges than generic AI coursework.

The labor market created by AI is not only a developer market

If we look only at office automation and software roles, we miss the people who deliver electricity, remove heat, maintain equipment and recover failures. Their labor market can move with AI investment too.

Employers should therefore define transferable experience from utilities, EPC, semiconductor facilities, manufacturing and energy rather than requiring years of identical data-center tenure for every role.

BANSEOG VIEW

The AI talent market should be read as both a software market and an infrastructure-operations market

AI-data-center expansion can create new roles while repricing existing power and facilities careers into a new industry context.

The more useful signal is not headline job counts, but which workers move from adjacent sectors into data-center roles and which combinations of experience actually convert into hires.

SOURCES

Primary sources and references

Korean investment plans and US hiring and executive-survey data measure different markets. US shortage figures are not applied to Korea; they are used to show how AI infrastructure can create overlapping labor pools.