An experienced trader with over 25 years at the Wall Street financial giant State Street suggests the Korean won could stage a significant rebound in the second half of the year. This outlook is based on the waning pressure from foreign investors' prolonged selling of South Korean stocks, a trend receding against the backdrop of a super bull market in South Korean equities attracting global inflows due to the "memory supercycle." Furthermore, sustained current account surpluses, driven by soaring global demand for memory chips, are expected to support the currency.
State Street's Bullish Outlook on the Won
Seo Jiwang, head of foreign exchange trading and sales for the markets division at State Street's Seoul branch, stated in a media interview that if global risk sentiment continues to improve and foreign investors return to South Korean equities on a large scale, the won could strengthen to around 1,400 per US dollar by year-end. The won traded around 1,517 per dollar on Tuesday. Such an appreciation from approximately 1,517 to around 1,400 would be a net positive for the South Korean benchmark Kospi index, which has already surged over 100% this year. It would also signal that the bull run for the two memory chip giants—Samsung Electronics and SK Hynix, which together account for over 50% of the Kospi's weighting—is far from over. Their continued strong performance is a significant positive catalyst for the global memory sector and the AI computing supply chain.
If the appreciation stems from foreign capital inflows, current account surpluses, a decline in the risk premium for South Korean assets, and heightened expectations for potential interest rate hikes by the Bank of Korea, it would reflect not a loss of export competitiveness but a sustained recovery in confidence towards South Korean assets. State Street's rationale for won strength is precisely based on the gradual easing of foreign selling pressure, persistent current account surpluses, and renewed allocation to South Korean assets, collectively driving a robust second-half rebound.
JPMorgan's Bullish Kospi Targets
Wall Street giant JPMorgan has twice significantly raised its target for the record-breaking Kospi index in less than a month. The core logic is that the bull market narrative, driven by the AI infrastructure boom and the ensuing "memory chip supercycle," is far from concluding. The bank has raised its base target for the Kospi to 9,000 points and its bull-case scenario target substantially to the historic 10,000-point level. In contrast, its base and bull targets set in late April were 7,000 and 8,500 points, respectively.
Timing the Won's Recovery
The won hit its weakest level since the 2009 global financial crisis in late March, pressured by a short-term surge in the US dollar index due to geopolitical tensions and capital outflows from foreign investors taking profits on South Korean equity assets. Seo Jiwang noted that if the super bull market in South Korean stocks persists, it could encourage global investors to quickly rebuild their positions after months of selling. He added that if the Bank of Korea proceeds with interest rate hikes as the market expects, helping to narrow the yield gap with long-term US Treasuries, the won could gain additional support. Data compiled from institutions shows the interest rate swap market is currently pricing in a policy rate of 3.84% in one year, implying expectations for more than five 25-basis-point hikes over the next 12 months.
As of June 1st, foreign investors have net sold $65.7 billion worth of South Korean stocks this year, primarily concentrated in the first quarter following massive gains in 2025. This represents the largest outflow on record. Despite this, driven by the powerful rallies of the world's two largest memory chip makers, Samsung and SK Hynix, the Kospi index has nearly doubled in less than six months, making 2025's full-year gain of nearly 80% seem modest by comparison.
Seo Jiwang emphasized that a more important positive development is not the volatility from foreign capital flows but the overall increase in market liquidity. This reflects the growing influence of global equity portfolio flows on the won exchange rate market, extending beyond South Korea's traditional export-driven dynamics. He highlighted a significant increase in market depth this year, with much larger order flows on both the buy and sell sides, indicating the market has entered a more mature phase.
The next major test for South Korea's foreign exchange market will be the introduction of near-24-hour spot trading for the dollar-won pair starting July 6th. This is a key initiative to improve market access and bolster the country's case for developed market status. Seo noted that extended trading hours signify more than just a longer session—they mark a crucial step in the won's evolution towards a more internationally traded currency. The ultimate test will be liquidity. He stated that if sufficient liquidity is available around the clock with pricing tighter and more robust than in the non-deliverable forward (NDF) market, the market could expand successfully, potentially growing significantly larger than in the past. A trial run for the extended hours will begin on June 29th.
Synergy Between Currency and Stocks
As mentioned, if won appreciation is driven by foreign capital inflows, current account surpluses, and expectations of monetary tightening, it signals a recovery in confidence towards South Korean assets, acting as a sustained positive catalyst for the memory chip giants and the stock market. Following Samsung Electronics into the trillion-dollar market capitalization club, SK Hynix has also recently surpassed this milestone, with both companies propelling the Kospi to new highs. The core driver is near-insatiable AI-driven demand for memory and expectations of rising chip prices.
For the South Korea-based memory chip leaders and the bull market, won appreciation represents a market repricing signal, shifting from a narrative of "export surplus with a weak currency" to one driven by "AI computing supply chain prosperity + foreign capital repatriation + market institutional upgrades." More precisely, appreciation is a major positive catalyst for the valuation expansion of the two memory giants, which dominate the Kospi. It signals stronger foreign willingness to buy South Korean stocks, reduces USD-denominated exchange rate risk for foreign investors, and could attract more global capital to the Kospi index and its unique, globally dominant memory semiconductor leaders.
Won appreciation could also significantly strengthen the narrative of South Korea's market internationalization and re-rating. As State Street anticipates, liquidity and market depth in the FX market are rising. The extension to near-24-hour trading from July 6th will enhance convenience for overseas investors and support the capital market reform narrative aimed at achieving developed market status. If a won rebound accompanies further Kospi gains, it could create a positive "stocks and currency rising together" feedback loop: foreign inflows push stocks higher, which in turn lowers the exchange rate risk premium, attracting more overseas capital into South Korean memory semiconductor assets.
For Samsung and SK Hynix, the short-term impact of won appreciation might be a slight negative on translated profits, but it is unlikely to reverse the current bull market trend driven by the AI infrastructure boom and the "memory supercycle." Contract prices for memory chips, especially HBM, server DRAM, DDR5, and enterprise NAND components, are being pushed higher by nearly endless demand from North American AI data centers, long-term supply agreements, capacity bottlenecks, and customer volume commitments. In this environment, the profit boost from exponential demand expansion, sustained product price increases, and product mix upgrades is likely to far outweigh the translation pressure from a currency move from 1,517 to around 1,400. If the stock rally is primarily driven by "AI memory supply shortages + HBM long-term contracts + earnings upgrades," the profit pressure from currency appreciation is merely marginal noise.
Whether for Google's massive TPU AI computing clusters or Nvidia's AI GPU clusters, comprehensive integration with HBM memory systems is essential. Coupled with tech giants' accelerated construction or expansion of AI data centers requiring massive purchases of server-grade DDR5 memory and enterprise-grade high-performance SSD/HDD storage, Samsung, SK Hynix, and Micron are uniquely positioned across the three most critical memory segments: HBM, high-performance server DRAM (including DDR5/LPDDR5X), and high-end data center SSDs. They are direct beneficiaries of the "AI memory + storage stack," capturing the "super红利" (super红利) of the AI infrastructure wave. GPUs generate intelligence, HBM/DRAM feeds data at high speed, enterprise NAND/eSSD handles hot data and caching, while HDDs manage long-term retention of massive cold/warm data. Therefore, Wall Street giant Goldman Sachs argues that the AI computing arms race led by cloud giants is transforming memory chips from cyclical commodities into scarce strategic assets. Price increases for DRAM/NAND in 2026 are not the end but potentially the initial phase of a supercycle.
From Commodity to Strategic Asset
Amid the explosion in AI computing demand, the memory chip industry is undergoing an unprecedented structural transformation. The traditional cyclical memory market, once dominated by consumer electronics and inventory cycles with highly volatile prices and demand, is being redefined. AI training and inference's persistent, near-insatiable demand for high-bandwidth memory (HBM), server DRAM, and enterprise SSDs is positioning memory as a strategic core asset within AI infrastructure. The total addressable market (TAM) is projected to reach approximately $1.7 trillion by 2028, highlighting a fundamental shift from a commodity nature to an indispensable component of AI computing infrastructure.
JPMorgan emphasizes that within AI computing systems, the CPU is becoming a key growth engine following the GPU, driving memory and NAND demand from marginal growth to deep, explosive expansion. In the coming years, the demand share from AI CPUs, ASICs, and GPUs for high-bandwidth, low-latency memory will increase significantly. The physical constraints of advanced manufacturing and the complex packaging required for HBM make rapid supply expansion difficult, likely creating a sustained supply shortage from 2026 to 2028.
Against this backdrop of global supply-demand reshaping, long-term supply agreements (LTAs) and structural demand distribution are also reconfiguring price trends. The proliferation of long-term contracts not only enhances profit certainty but also provides cash flow protection for memory makers through deferred revenue and prepayment mechanisms. This decouples pricing from being solely influenced by spot markets and inventory cycles, instead tying it more closely to the long-term capital expenditures of major corporate clients, especially cloud supergiants like Microsoft and Google. This mechanism significantly mitigates cyclical volatility risks.
This logic is even driving a fundamental change in valuation frameworks. Goldman Sachs points out that in the environment of the AI infrastructure frenzy and the growing prevalence of LTAs, memory chip makers are moving away from the old paradigm of valuation based primarily on price-to-book (P/B) ratios towards a price-to-earnings (P/E) framework based on profitability. This shift is prompting Wall Street to significantly raise target prices for leaders like Samsung, SK Hynix, and Kioxia, with their earnings profiles transitioning from "cyclical booms and busts" to "predictable and stable cash flow generation."
As Jeremy Werner, senior vice president and general manager of Micron's Compute and Networking Business Unit, recently explained from an engineering perspective, the underlying driver of this cycle is not simply that "AI needs more compute chips." The era of AI inference dominated by agents like Claude and other AI workflows places memory/storage as a system bottleneck, not just a supporting component. AI training relies heavily on massive parallel computation, while inference—especially with long contexts, multi-turn conversations, and agentic AI workflows—requires persistent storage of KV Cache, context states, and intermediate results. When memory/storage is insufficient, models must recompute historical states, leading to decreased GPU utilization and increased token generation costs. Therefore, HBM, DDR5, LPDDR, enterprise SSDs, and even HDDs/data lakes are forming an "AI memory chain" from GPU-proximate to distant storage, determining an AI system's throughput, latency, concurrency, and per-token economics. This explains why memory and data storage stocks like Micron, Samsung, SK Hynix, and others are rallying in tandem: demand is not concentrated solely on HBM but is spilling over across the entire chain of DRAM, NAND, SSD, and HDD along the AI server architecture.