AI bust? Bet goes on, albeit in moderation

The KOSPI index is displayed on a screen at Hana Bank in Seoul on July 10 2026 AJP Yoo Na-hyun
The KOSPI index is displayed on a screen at Hana Bank in Seoul on July 10, 2026. AJP Yoo Na-hyun

SEOUL, November 13 (AJP) -  Nvidia is a darling of South Koreans invested in U.S. stocks and a customer helping memory chipmakers SK hynix and Samsung Electronics mint profits. Jensen Huang, its departing founder, enjoyed a celebrity status here.

Yet Seoul took his abrupt exit with relative calm. For now, Koreans are not ready to part with faith in AI. 

The main KOSPI ended Friday around 1 percent lower, reversing morning gains of more than 2 percent after foreign investors turned net sellers. The KOSDAQ finished 0.45 percent higher while the Korean won closed mostly flat at 1,337.70 per dollar.

For now, investors expect the $4 trillion company to remain in steady hands under interim leadership.

Nvidia finance chief Colette Kress, whom the board appointed interim CEO on Thursday, said in August that revenue for the fiscal year ending in early 2028 was expected to grow 70 percent. Although slower than the previous year, that would add around $287 billion in revenue at gross profit margins expected to exceed 70 percent.

The company does not expect supply of its AI chips to catch up with demand anytime soon.

“Although we will work to close the supply-demand gap, we expect supply to remain a bottleneck at least through the end of fiscal year ‘28,” Kress said.

Economists agree the AI phenomenon has yet to fizzle out.

“It's too early to call this a period of plateau,” said Yang Jun-sok, an economics professor at the Catholic University of Korea.

“We still don't have reliable figures for future revenue and profits. The industry is driven more by expectations than by predictable returns.”

Kim Dae-jong, a professor of business administration at Sejong University, sees the industry moving into a different phase.

“The AI boom hasn't peaked,” Kim said. “But we're moving from rapid expansion to a phase where efficiency and profitability are being put to the test.”

For Asian chipmakers that have benefited from the AI spending surge, that distinction could determine whether their shares recover from a steep selloff.
 
Compiled and edited by AJP Graphics AI-assisted
Compiled and edited by AJP. Graphics AI-assisted.

The four companies — Samsung Electronics, SK hynix, Taiwan’s TSMC and Japan’s Kioxia — have ridden the AI investment wave through different businesses, from advanced chipmaking and high-bandwidth memory to data storage. 

Their shares have fallen sharply from this year’s highs, even as AI demand remains strong upon fears over over-enthusiasm and spending in AI buildout. 

The reckoning sharpened after Huang's exit and the struggle to replace the outsized figure in the AI orbit. 

Nvidia shares plunged more than 13 percent at the opening bell before recovering much of the loss to close Thursday 4.34 percent lower at $166.19. The stock has fallen roughly 31 percent from its October high, wiping out a substantial portion of the gains that had accompanied the AI investment boom.

Samsung Electronics ended Friday at 207,000 won, down 0.48 percent, and SK hynix at 1,222,000 won, down 2.55 percent. At Friday’s close, Samsung Electronics was about 43 percent below its June 18 peak of 362,500 won, while SK hynix was down roughly 58 percent from its June 22 high of 2,919,000 won.

The selloff also hit other Asian chipmakers, though the losses varied.

TSMC’s U.S.-listed shares closed Thursday at $405.49, about 17 percent below their 52-week high of $487.47, although they remained up roughly 34 percent this year. Japan’s Kioxia closed at 9,300 yen, down about 75 percent from its June peak.

Yang sees the retreat in Korean chip stocks as a correction of the excessive optimism that drove their rally earlier this year.

“The market was too optimistic about AI in the first half,” he said. “There was a bubble, and I think we're seeing something of a soft landing as it deflates.”

Kim also sees room for a reassessment of valuations, though he remains more optimistic about the industry’s long-term prospects.

“AI will remain a major growth industry shaping the global economy over the next decade,” Kim said. “But excessive expectations in the stock market can still be corrected.”
 
Compiled and edited by AJP Graphics AI-assisted
Compiled and edited by AJP. Graphics AI-assisted.

The stocks have stumbled. The earnings have not.

TSMC, the world’s largest contract chipmaker and a key manufacturer of Nvidia’s advanced processors, reported October revenue of NT$586.89 billion ($17.86 billion), up 59.7 percent from a year earlier and 14.7 percent from September. Revenue for the first 10 months rose 43.3 percent from a year earlier.

South Korea’s two chip giants have also delivered record earnings. Samsung Electronics reported third-quarter revenue of 195.3 trillion won and operating profit of 107.4 trillion won, while SK hynix posted revenue of 104.6 trillion won and operating profit of 83.3 trillion won, fueled by strong demand for advanced memory chips, including fourth-generation high-bandwidth memory (HBM4).

Still, strong AI demand alone does not guarantee sustained profit growth. Samsung Electronics and SK hynix compete in the HBM market, which supplies the fast memory used alongside AI processors. Their earnings will depend on chip prices, production capacity and their ability to secure customer orders.

Kioxia faces different challenges. The Japanese company specializes in NAND flash memory, used to store data in AI data centers rather than feed AI processors at high speed. Growing demand for storage could lift its earnings, but an oversupply of NAND chips could drive down prices and squeeze profits.

“All four companies stand to benefit from AI's long-term growth, but their share prices will increasingly diverge based on their technology, profitability and how much future growth is already priced in,” Kim said.

“Samsung has room to grow if it strengthens its HBM technology and recovers market share,” Kim said. “SK hynix is a major beneficiary of AI investment, but its stock may already reflect much of the expected earnings growth.”

Kim said Samsung also has an advantage in operating both memory and contract chipmaking businesses, giving it more ways to benefit from AI demand. SK hynix, despite its leading position in HBM, could face greater short-term share-price volatility if earnings fail to meet already elevated expectations.

For TSMC, Kim sees advanced chipmaking technology and a broad customer base as key advantages, although geopolitical risks and heavy capital spending remain concerns.

“Kioxia could benefit from growing demand for enterprise solid-state drives (SSDs) as AI data centers expand,” he said. “But it faces greater volatility from NAND oversupply and price swings.”

For the two Korean chipmakers, Kim said, much will depend on whether they can secure more HBM4 orders as competition intensifies.
 
Nvidia CEO Jensen Huang visits the College of Engineering at Seoul National University in Gwanak-gu southern Seoul on June 8 2026 AJP Yoo Na-hyun
Nvidia CEO Jensen Huang visits the College of Engineering at Seoul National University in Gwanak-gu, southern Seoul, on June 8, 2026. AJP Yoo Na-hyun

The outlook also depends on Nvidia, whose processors sit at the center of much of the AI infrastructure boom. Its leadership shake-up raises questions about whether the company can sustain innovation and keep customers’ investment plans on track.

“For the next few months, I don't expect a major difference because the GPUs have already been sold and investment plans are in place,” Yang said. “But if the company loses momentum without Huang, we could begin to see the effects in two or three months.”

Yang said Nvidia would need time to reorganize a decision-making structure centered on Huang. He estimated that the transition could take around six months, depending on how quickly the company appoints a successor and the new CEO’s experience.

Kim, however, believes Nvidia’s competitive edge extends well beyond its founder. He pointed to the company’s CUDA software platform, which helps developers use its processors for computing tasks.

“Nvidia's competitiveness does not depend on one person alone,” he said. “It has accumulated GPU design expertise, the CUDA software ecosystem, a global customer network and enormous research and development capabilities.”

Both economists see opportunities for Nvidia’s rivals, though Yang expects its technological and software lead to limit any immediate shift.
 
Designed by AJP Song Ji-yoon
Designed by AJP Song Ji-yoon

Kim sees growing competition as an opportunity for Korean memory suppliers.

“If Samsung and SK hynix reduce their dependence on a single company and diversify their customers across Nvidia, Advanced Micro Devices (AMD) and Google, they can create new growth opportunities in HBM and advanced memory,” he said.

Nvidia, for its part, has sought to reassure customers. Its board said Thursday that it expects existing commitments, operations and product development plans to continue without disruption.

Yang identified a separate, conditional risk: any evidence of leaked company secrets or management involvement in illegal chip shipments to China could expose Nvidia to criminal proceedings and regulatory penalties.

Huang’s departure adds to pressures already weighing on the AI boom. Rising borrowing costs, opposition to energy-intensive data centers and uncertainty over how technology companies will fund their expansion plans have raised doubts about whether AI spending can continue at its current pace.

Kim sees the industry shifting toward a greater focus on returns.

“The industry is moving from quantitative expansion to qualitative growth,” Kim said. “Companies that use AI to raise productivity and generate real profits will be valued more highly than those that simply build more data centers.”

He pointed to financing costs, electricity shortages, delays in expanding power grids and local opposition as constraints on further construction.

“AI companies are investing enormous sums, but the question is how quickly those investments can produce revenue and profits,” Kim said. “Growth will continue, but the days of expanding investment without regard to returns will become harder to sustain.”

For investors, the question is whether AI stocks can keep rising as growth slows. Strong earnings may not be enough if expectations are too high. Shares that have already fallen sharply, however, could recover if profits hold up.

AJP also asked ChatGPT and Anthropic’s Claude whether the AI chip boom was nearing a plateau.

Echoing Kim’s view that the boom has not peaked, ChatGPT suggested that demand for advanced processors and memory could keep growing even if data-center construction slows. Claude’s assessment was closer to Yang’s, suggesting that the selloff in chip stocks may reflect a correction in excessive expectations rather than weakening demand.

Attention now turns to Nvidia’s fiscal third-quarter results on November 17, due after U.S. markets close. Investors will be watching its earnings, sales outlook and plans for maintaining its technology roadmap under new leadership.

The AI boom may outlast Jensen Huang. But for chip stocks, the question is whether reality can finally catch up with the hype.