AI capital expenditure stirs South Korea's economic lifeline, central bank warns export drop could reach $100 billion! Why does Goldman Sachs remain bullish on Korean stocks and memory giants?

According to English Planet Daily, it is understood that the Bank of Korea has recently revealed in a research report a core risk to the Korean economy: if US tech giants slow down their AI capital expenditure next year, Korea’s semiconductor exports could fall short by $50-100 billion compared to earlier forecasts, dragging down the country's 2027 GDP growth rate by about 0.4 percentage points. However, this is a projection under a pessimistic scenario and does not represent an actual loss of exports, nor a baseline forecast for a $100 billion year-on-year export decline.

Meanwhile, Wall Street financial giant Goldman Sachs, although lowering its KOSPI target by 25%, still believes storage chip valuations are overly pessimistic; Korean retail investors are once again buying Samsung Electronics, SK Hynix, and highly leveraged ETF products closely tied to the chip theme. These three signals together form the current divergence in the AI computing power trading theme — on the macro level, some investors continue to worry about the series of major negative shocks that could arise if North American tech giants cool their AI investments in the face of escalating AI financing costs, while the stock market is beginning to judge whether the pessimistic risk has been excessively priced in as valuation measures for Korea’s two major storage chip giants have dropped sharply.

"Korea cannot afford to lose the AI storage competition" has real-world justification. Korea’s GDP grew 1% QoQ in the third quarter, beating the market expectation of 0.7%; the Bank of Korea estimates that semiconductors contributed about half of this economic growth, while private consumption fell by 0.2% QoQ. The chip boom has deeply impacted exports, corporate investment, and overall economic performance. This is why, in the eyes of President Lee Jae-myung and central bank officials, Korea needs to consolidate its high-end storage advantages, expand genuine AI applications and client sources, and reduce economic dependence on a single capital expenditure cycle.

The underlying mechanism behind export pressure lies in both shipment volume and average sales price jointly determining export value. Korean media reports show that even in pessimistic scenarios, chip export volumes may still increase by about 10%, but remain below initial expectations, with prices weaker than baseline assumptions. Thus, even if demand for AI computing resources continues to expand, as long as procurement growth and prices fall short of previously very high expectations, a massive export gap will result. For storage chip stocks, the key question becomes: will future profits merely slow in growth, or will they see a deep decline as implied by the current pessimistic valuation and recession cycle?

Bank of Korea Warning: If AI Spending Slows Down, Korea’s Chip Exports Could Decrease by Up to $100 Billion Compared to Forecasts

Bank of Korea data shows that if US tech giants slow down their capital spending on AI next year, Korea’s semiconductor exports could be $50-100 billion less than previously forecast.

In a written reply submitted to National Assembly member Ahn Do-jae, the Bank of Korea constructed a "pessimistic scenario," assessing the economic impact of US tech giants cutting massive infrastructure investments.

In this scenario, exports suddenly fall short of expectations, causing Korea's 2027 current account surplus to be significantly lower than the central bank’s original forecast of $430 billion, and GDP growth for next year to drop by about 0.4 percentage points.

This scenario assumes a 10% decline in the unit price of Korean semiconductors. The Bank of Korea’s latest assessment underscores that market concerns about Korea’s heavy reliance on chip exports are deepening. Chips currently account for about half of Korea’s total exports.

The report notes that the combined capital expenditure of five major US tech companies — Amazon, Google, Microsoft, Meta, and Oracle — has rapidly increased from $40 billion last year to a conservatively estimated $80 billion this year, with next year’s conservative estimate reaching a staggering $110 billion.

However, the Bank of Korea cautions that there is still considerable uncertainty regarding the development trajectory of demand for cutting-edge AI technologies and the overall business cycle for storage chips driven entirely by AI computing power requirements.

Bank of Korea Estimates Export Gap, While Goldman Sachs Attempts to Find Storage Chip Stocks' Market Expectations Divergence

Goldman Sachs’ research indicates that the institution has lowered its 12-month target for the benchmark Korea stock index — the KOSPI Composite Index — from 12,000 points to 9,000 points, reflecting a more conservative valuation assumption; at the reporting time, the new target still represents over 60% potential upside. Goldman Sachs estimates that the current semiconductor valuations imply a 65% drop in profits by 2027, a scenario which the Goldman analysts consider overly pessimistic, with market negativity being excessive.

At the time, both Samsung Electronics and SK Hynix were trading at a mere 3x forward P/E; the implied return on equity for the storage chip sector via price-to-book ratio was only around 23%, significantly lower than the market consensus of 45%. These numbers reflect the gap between market pricing and earnings expectations: the 65% implied profit decline is derived from KOSPI trading points and represents excessive pessimism, not Goldman Sachs’s own forecasts and outlook.

The Goldman Sachs analyst team states in the report that Samsung Electronics and SK Hynix are currently trading at around 3x forward P/E, a historic low. Based on the past 20 years of monthly data compiled by Goldman, which shows a 72% fit (R²) between price-to-book ratio and ROE, the current sector price-to-book ratio of 2.6x implies a 2027 ROE of only 23%, whereas consensus forecasts have ROE at 45%.

This divergence is also pronounced at the index level: KOSPI is down about 40% from its June peak, while forward 12-month earnings forecasts have risen about 22% over the same period; the KOSPI's roughly 4x record low forward P/E is at an extreme low among 18 major trillion-dollar markets covered in the relevant statistics. However, the storage industry’s low P/E also likely reflects concerns that peak-cycle profits are unsustainable. Goldman’s bullish view on Korea’s two storage chip giants and the KOSPI is based on the premise that actual profit downgrades are less than what is implied in stock prices’ pessimism, and that, combined with falling risk premiums, Korea's stock market and its two chip giants (which account for 50% of the market capitalization) may see strong valuation recovery during slower profit growth.

Goldman Sachs’ research also shows funds in the Korean stock market have begun betting on this expectations gap. From October 1 to 23, Korean individual investors made net purchases totaling 315.4 billion won across 14 high-leverage ETF products connected to Samsung Electronics and SK Hynix, with 226.3 billion won in Hynix-related products and 89.1 billion won in Samsung products. From October 1 to 22, Korean investors net purchased about $4.8 billion in US stocks, with $3.548 billion flowing into the triple-leveraged Philly Semiconductor ETF (US code "SOXL"), accounting for 73.9%. This shows bottom-fishing funds are highly concentrated in the storage chip and semiconductor supply chains driven by AI computing demand, though leverage products with daily resets amplify volatility and losses, and capital inflows alone do not prove the market has bottomed.

The industrial supply chain basis for the global rebound in the storage chip sector is that, as cutting-edge AI agents like Muse gain worldwide popularity, demand for AI high-performance computing resources is spreading to more memory and larger-capacity enterprise NAND storage levels — high-bandwidth memory (HBM) supports accelerator computation, server DRAM handles running data and tasks, enterprise SSDs house indexes, task records, and persistent states. From an AI inference system architecture perspective, complex AI tasks led by Meta Muse and OpenAI’s GPT-6 Astra often involve longer context, multiple model calls, tool operations, and result validations: prefill handles inputs, decode generates outputs, key-value cache (KV Cache) consumes more HBM memory as context and concurrency scale, requiring synchronized increases in computing throughput, memory bandwidth, and enterprise-grade high-performance storage capacity.

SK Hynix has disclosed that HBM4 began volume shipments in the second quarter and that they have established long-term supply agreements with about ten clients; Bank of America, after recent discussions with Micron’s management, continues to highlight supply constraints and the visibility provided by long-term agreements for demand. Market research firm TrendForce forecasts contract prices for conventional DRAM to rise 10-15% quarter-on-quarter in Q4 from a high base, and NAND contract prices to rise 15-20%. On the DRAM side, AI agent workloads drive general server and memory procurement, but constraints exist in advanced process, packaging, and capacity allocation; on the NAND side, real-time search, caching, and vector databases are expanding demand for enterprise SSDs. TrendForce predicts that enterprise SSD bit demand will grow by more than 80% year-on-year in 2026, with most new supply already reserved, so price increases for this category in Q4 may even be larger than those seen in Q3.

TrendForce also sharply raised its forecast for blended ASP growth of HBM in 2027 to 121% year-on-year, based on HBM and conventional DRAM competing for limited advanced process and wafer capacity, the ramp-up time needed for new product yields, and the rising proportion of higher-priced HBM4 with gradual volume growth of HBM4e.

During the October 27 morning session, Korea’s KOSPI rose more than 4%, following a 2.8% overnight gain in the Philadelphia Semiconductor Index; Anthropic's official IPO filing and Meta’s equity financing plans improved forecasts for funding sources for AI construction. The key factor for continued large-scale rebound in the US semiconductor sector on Monday will be whether fundraising results in continuously stronger procurement expectations, which may ultimately translate into profits and cash flow for storage manufacturers.