Google invests 20 trillion won in Malaysia! Amazon, Meta and MS expand abroad… Can Korea capture major AI infrastructure investment? #AIDataCenters #Google #NVIDIA #SamsungElectronics #SKHynix

An independent exercise, not a statement by the video creator or verification of every claim in the video.

Does a 100MW plan directly determine annual electricity sales?

Assume a hypothetical data center has a total facility grid-import limit of 100MW and average annual actual electrical load of 80% of that limit. Over a non-leap year of 8,760 hours, facility consumption is 100 × 0.8 × 8,760 = 700,800MWh. If PUE, total facility electricity divided by IT equipment electricity for the same period, is 1.25, IT consumption is 700,800 ÷ 1.25 = 560,640MWh and the remainder is 140,160MWh. Treating 100MW as IT load and multiplying by PUE instead changes the scope from the original grid-import limit. Here, 80% means average electrical load relative to the limit, not GPU compute utilization or service availability. Actual electricity revenue depends on metering, tariffs and contracts, and sales volume alone does not establish net profit.

Check for yourself

1. Does the MW figure concern generation, connection limits, total facility consumption or IT equipment?
2. Are average load, period and PUE aligned, with grid upgrades, generation fuel and tariff conditions checked separately?

Limits of application

Capacity, average load and PUE are hypothetical, not estimates of a country’s or project’s actual utilization, electricity revenue or profitability. Monthly load variation, metering boundaries for losses, outages, onsite generation, storage and tariff structures are simplified. Generation nameplate capacity alone cannot establish continuous year-round supply or GPU counts.