“What Now for Samsung and SK hynix?” ... Could News from China Rattle the KOSPI? Shareholders Anxious

Nvidia Fell 17% During Last Year's DeepSeek Shock
Upgraded V4.1-Flash: "Works Faster and Gives Shorter Answers"

Attention is focused on the impact DeepSeek, a Chinese artificial intelligence (AI) company, could have on the share prices of Samsung Electronics and SK hynix after it released a new model that outperforms major U.S. AI models in some performance evaluations and costs as little as one-fortieth as much to use.

On October 17, DeepSeek announced on its official website and other channels that it had released “V4.1-Flash-1017.” An upgraded version of “V4.1-Flash,” released on September 10, it features improved coding and AI agent performance. DeepSeek said, “Compared with the previous version, it delivers better benchmark performance, works faster, and generates shorter responses.”

In DeepSeek’s own evaluations, the new model scored 77.0 on DeepSWE v1.1, a software development benchmark, beating Anthropic’s Claude Opus 5 (74.0) and OpenAI’s GPT-5.6 Sol (73.0). It also scored higher than Opus on AutomationBench, a work automation benchmark, while its TerminalBench 4.0 score, which assesses coding agents, improved over the previous model.

During off-peak hours, DeepSeek charges $0.60 per 1 million output tokens via its API. That is one-thirty-third the price of GPT-5.6 Sol ($20) and about one-forty-second the price of Opus 5 ($25). The Decoder, a German AI news outlet, cited a cost chart released by DeepSeek and reported that the new model costs an average of less than $0.20 per task, less than half the roughly $0.50 charged for the previous model.

Markets are watching closely ahead of trading on October 19. In January 2025, DeepSeek released its R1 reasoning model, sending Nvidia’s share price down about 17% in a single day. Some $589 billion (about 787.97 trillion won) in market capitalization was wiped out. The sell-off followed concerns about whether massive investment in AI infrastructure was necessary if low-cost models could deliver performance comparable to that of U.S. models.

U.S. markets have also been rattled recently by concerns over the profitability of AI investments. On October 16, the Nasdaq Composite fell about 2.1%. The impact extended to Samsung Electronics and SK hynix, suppliers of high-bandwidth memory (HBM). Alongside the debate over returns on AI investment, higher-than-expected U.S. consumer inflation and heavy foreign investor selling contributed to sharp declines on October 15, when the two companies’ share prices plunged 9.48% and 11.06%, respectively. They edged up slightly on October 16 but did not recoup their losses.

DeepSeek’s new model was released over the weekend, so the market has yet to react. However, there are concerns that its release could fuel skepticism about AI investment and extend the downturn in share prices.

Others point to Jevons’ paradox, arguing that lower AI costs could instead boost demand and computing workloads. In other words, lower-cost AI would not necessarily lead directly to a decline in semiconductor demand.

This content was produced with the assistance of AI translation services.