[Weekly Kyunghyang] Recently, A (42), who had been browsing websites to buy a new laptop, was startled. The new model of the company laptop he has used for five years had jumped by nearly 500,000 won compared to the previous one. A said, “Performance matters, but I thought a familiar brand would be better, so I was looking at the new model, yet the price has gone up too much,” and added, “I considered waiting a little before buying, but after checking the internet, there are many opinions urging a quick purchase with phrases like ‘Today is the cheapest,’ so I am undecided.”

Recently, as the prices of key computer components have surged one after another, consumers are growing anxious.

According to the PC price comparison site Danawa on February 4, the price of a custom-built PC (performance PC, 16 GB) is currently 1.52 million won, up about 50% from October last year (1 million won). The Danawa standard gaming PC (16 GB) is 2.60 million won, about 45% higher than October last year (1.79 million won). For laptops, the latest 2026 models from Samsung (Galaxy Book) and LG (Gram) are about 300,000~700,000 won more expensive than the previous models with similar specifications. Because AI-related features have been added, a simple comparison is difficult, but the price burden felt by consumers has increased. With price hikes driven by semiconductor shortages, the so-called ‘chipflation’in which the prices of related items such as electric vehicles and smartphones riseis expected to continue.

At the core of the sharp rise in prices of laptops and other electronic devices is AI. The global AI boom has led to a rapid increase in data centers, causing shortages not only of memory semiconductors such as DRAM and NAND flash used in them, but also of key components such as central processing units (CPUs) and printed circuit boards (PCBs).

Most notably, the price of standard DRAM has jumped 5~6 times in just half a year. According to Danawa, the lowest price for general-purpose PC DRAM (Samsung DDR5-5600 16 GB) rose from 69,246 won in September last year to 379,780 won in early February this yearan increase of 5.5 times.

AI training requires more data centers, whose key components include GPU (graphics processing unit) and CPU (central processing unit) servers for high-performance computing, storage to hold massive datasets, network switches and routers for fast data transfer, and UPS (uninterruptible power supply) units and air-conditioning and humidity-control equipment (cooling systems) for stable operations. Among these, CPUs and storage differ from the models used in consumer laptops and PCs, but as demand for them increases, production of consumer CPUs and the like is squeezed, pushing up prices. Rising memory prices can also affect smartphone prices.

On February 4 (local time) at CES 2026 in Las Vegas, Samsung Electronics President Noh Tae-moon said, “The cost of major components, especially the rise in memory prices, is the biggest burden,” adding, “Such increases in material costs will inevitably have some impact on product prices, including smartphones.”

Some caution that this situation should not simply be dismissed as ‘the direction of the market’. As the COVID-19 pandemic revealed, laptops and smartphones today are close to essential goods for sustaining work and daily life. If prices for such products rise across the board, lower-income households will be hit directly.

An industry insider explained, “Because silicon wafers are a core material in semiconductor manufacturing, they are essential to virtually all electronic products, including computers, telecommunications devices, and consumer electronics, yet most of the limited wafer supply is currently being allocated to HBM (high-bandwidth memory) production,” adding, “There are exchange rate issues and some diagnose this as a temporary bottleneck, but AI-related excess demand of this kind will persist for at least several years, and since factories and resources have limits, consumer electronics are also likely to become more expensive over the long term.”

This phenomenon is unlikely to end with a simple rise in electronics prices. AI data centers require enormous resources and energy beyond semiconductors, including electricity, water, and copper for fiber-optic cables.

Kate Crawford, a senior principal researcher at Microsoft Research, emphasized in <Atlas of AI> (2022) that “The skeleton of AI is minerals, and the blood of AI is electricity.” In other words, large-scale training demands massive hardware and resources. In particular, as AI performance improves, the amount of electricity required for training and inference increases sharply. Former Google CEO Eric Schmidt also said, “What constrains AI is not chips but power.” According to the AI Institute at the University of Rhode Island last year, OpenAI released its latest model in August last year, ‘GPT-5’, whose average power consumption per query was estimated at 18.35 Wh (watt-hours), about 8.7 times the power used by the previous generation, ‘GPT-4’ (2.12 Wh per query). OpenAI has not clearly disclosed the specific amount of electricity consumed for ChatGPT training and inference.

According to the International Energy Agency (IEA), as of 2024, data centers consume 415 TWh (terawatt-hours) of electricity, accounting for 1.5% of global power consumption. By 2030, data center power consumption is projected to double to 945 TWh, with their share of total consumption rising to 3%. A 2024 report from the U.S. Department of Energy stated that, as of 2023, electricity consumed by data centers accounted for 4.4% of total U.S. power use, and this share is projected to reach 6.7~12% by 2028.

This could inevitably lead to power shortages and higher electricity bills. In Virginia, which has the world’s largest concentration of data centers, electricity rates in August last year jumped about 13% from the same month a year earlier. Illinois, another high-density data center region, saw an increase of about 12% over the same period. As power demand grows, thecosts of infrastructure such as transmission towers rise, and ordinary residents end up shouldering the rates needed for such expansion. According to research by the Green Transition Institute, among domestic companies, only LG CNS discloses electricity use by data center, while most companies do not properly disclose power consumption or water usage.

Kim Byung-kwon, head of the Green Transition Institute and author of <AI and the Future of Climate>, predicted that phenomena like chipflation will intensify, saying, “For AI based on LLM (large language model), performance improves as energy input increases. It is essentially a battle to secure energy.”

He continued, “Right now only hardware companies like Nvidia are in the black, while AI firms such as OpenAI are still posting massive losses. AI services may appear inexpensive, but this is part of a bid for winner-takes-all; we need to recognize that AI is not cheap at all, whether economically or in terms of the resources it consumes,” adding, “Germany has passed an Energy Efficiency Act that, among other things, requires all data centers to be supplied with electricity from renewable energy starting in 2027, and Korea urgently needs national policies to define and promote sustainable AI.”