Maximizing Korea's AI momentum

Troy Stangarone

There is a tendency to view the world through the status quo. If the economy is growing, the set expectation becomes that it will continue to do so. That mentality is already taking hold in Korea as it relates to soaring profits at Samsung Electronics and SK hynix from artificial intelligence (AI).

However, if Korea is to make the most of the current AI boom, it needs to think beyond the status quo to use this moment to create new strengths in the economy rather than just buttress existing ones.

Current expectations are being driven by the record corporate profits from Samsung and SK hynix for high-bandwidth memory (HBM) that is critical to the continued development of AI. Many politicians now treat these profits as a new normal, debating how to spend additional tax revenue or whether to impose excess‑profit taxes on both firms.

But there is no guarantee that this future will hold. Samsung and SK hynix will continue to be highly profitable companies in the near term, but we should not expect this run to last indefinitely. The early internet era illustrates why Korea must take a broader view of Samsung and SK hynix’s good fortune.

In the early days of the internet, the biggest winners were infrastructure firms. Cisco Systems, which built routers and switches, and Oracle, which provided databases and software, were among the initial leaders. Over time, however, they were surpassed by companies that learned how to harness the internet’s potential such as Amazon, Google and Facebook.

A similar fate likely awaits Samsung and SK hynix. Their profitability will eventually decline, and their centrality to AI will diminish. Cisco and Oracle remain profitable, but they are no longer the defining companies of a new technological era.

Today’s boom is driven by demand far exceeding supply. Google, Microsoft, OpenAI and others are rapidly building data centers to advance large language models, pushing up demand for HBM. That surge has driven up prices and forced Samsung, SK hynix and their American counterpart, Micron, to shift production away from consumer-electronics memory chips to meet HBM demand.

Shortages and price increases have already pushed Apple to raise prices for some of its products and request approval from the Donald Trump administration to purchase Chinese-produced memory chips. This is a challenge, however, for all consumer electronics firms. According to Benchmark Research, prices for DRAM and NAND flash memories, key components for consumer electronics, have risen by roughly 660 percent over the last year.

These price levels cannot hold. Even if Apple avoids Chinese producers such as YMTC or CXMT, other firms will turn to them to cut costs. Both Chinese firms are also set to expand production capacity in 2027.

This is where Korea’s semiconductor megaproject becomes central to its competitiveness. Without expanding production capacity, Korean firms will lose market share to Chinese competitors. This would have negative consequences for both Samsung and SK hynix in the medium-term if they face deeper Chinese competition outside of China.

There is another reason profits at Samsung and SK hynix will decline: Continued data center expansion is less certain than it appears. Neither Anthropic nor OpenAI, nor other AI developers, have demonstrated a compelling, profitable use case for AI. Meanwhile, data center expansion is facing growing backlash in the United States, and current plans are already stalling due to equipment shortages and energy constraints.

If current profit levels at Samsung and SK hynix are unsustainable due to the uncertainty around the data center buildouts and the likelihood that other firms will seek lower cost memory producers, what should Korea do?

Expanding semiconductor production capacity under the first megaproject is essential. Without it, Samsung and SK hynix will lose market share to Chinese firms producing lower‑end memory chips. Traditionally, the argument would be that Korean firms move up the value chain while Chinese firms handle basic memory. But in an era of weaponized trade, that division of labor is a national security risk.

Just as with semiconductor production capacity expansion, a buildout of data centers is needed if Korea is going to remain competitive. Both of these projects are as much about maintaining pace as getting ahead in the AI era.

The most compelling of the three megaprojects, however, is physical AI. Clear uses of AI include coding, as well as research and development, but the world is still searching for the application that will truly unlock AI’s potential. AI deployed in the physical world through robots, autonomous systems and industrial automation has strong potential to be one of those breakthrough areas.

But to harness this moment, Korea must invest in more than physical AI. It must create an environment where startups can develop new applications that could shape new industries, just as Amazon, Google and Meta did during the internet era. As Korea considers how to invest excess tax revenue or profits from today’s HBM boom, innovation beyond existing industries must be a priority.

Troy Stangarone is a visiting senior fellow at the Korea Economic Institute of America. The views expressed here are the author’s alone.