The two global superpowers of memory chips, South Korea's Samsung Electronics Co., Ltd. and SK Hynix, are making their largest-ever bets on frontier technology with investments worth hundreds of billions of dollars in response to the unprecedented AI super wave. However, the proposed capacity build-out also raises concerns: if AI spending cools, it could lead to a painful reckoning for the global stock market's "AI computing power trade" theme.
For global investors, the unprecedented investment boom around AI has undoubtedly been a core driver of the robust bull market in global equities in recent years. For Wall Street giants like Goldman Sachs and JPMorgan Chase, who are bullish on the continuation of the AI megabull market, this move by Samsung and SK Hynix significantly strengthens the logic of a "Circular AI Boom," thereby propelling the bull market surrounding the AI computing infrastructure frenzy to continue sweeping the globe.
As the phenomenon of the "Circular AI Boom" begins to spread from North America to every corner of the world, the massive windfall reaped in recent years by Asia's AI computing power supply chain has even begun to provide a financial support system for the "hyperscale tech giants" (Meta Platforms, Inc., Amazon.com, and Alphabet, etc.) in North America.
The global boom in artificial intelligence development is generating unprecedented cash flow for Asian computing hardware manufacturers. Some of these funds are flowing back into dollar-denominated assets, potentially indirectly supporting the financial environment—characterized by the low borrowing costs favored by former President Trump—that facilitates unprecedented AI capital expenditure by large tech companies and AI application leaders like Anthropic.
The windfall profits earned by Asia's top-tier AI computing infrastructure manufacturers like SK Hynix, Samsung Electronics Co., Ltd., Taiwan Semiconductor Manufacturing, and Foxconn/Wistron are circulating rapidly in the global economy, reflecting the internal flow of funds within the AI ecosystem on a worldwide scale.
The AI frenzy is creating an unprecedented amount of cash for smaller Asian economies like South Korea and Taiwan. This, in turn, is forming a narrower version of the Asian savings glut. Meanwhile, these leaders of Asia's AI computing supply chain are also becoming a key force for a long-term and healthy bull market in the US and global stock markets. In the late 1990s and early 2000s, the Asian savings glut had previously kept US benchmark borrowing costs at historically low levels.
Within the "Circular AI Investment Boom," the logic for the US stock bull market is arguably more solid than simply "tech giants burning money on GPUs" because it is forming a global closed loop: US AI applications and cloud vendors create immense demand and achieve strong AI revenue, the Asian chip/memory/server supply chain gains export benefits, Asian surpluses partially flow back into dollar assets, thereby strongly supporting the equity/debt financing environment and AI capital expenditure ambitions of US tech giants.
Samsung and SK Hynix are providing "hard evidence" from the supply side and national industrial capital expenditure: South Korea aims to double its memory chip production capacity within five years, with Samsung and SK Hynix jointly committing to invest 3200 trillion won, approximately $2.07 trillion, including an 800 trillion won chip cluster in the southwest region.
The significance of the "Circular AI Investment Boom" framework lies in addressing the other half of the market's greatest concern: who pays for this round of computing power capital expenditure, and can AI applications truly convert computing power consumption into cash flow? Anthropic recently projected its Q2 revenue to surge to $10.9 billion, a sequential increase of about 130%, and achieve its first quarter of operating profit. This directly undermines the pessimistic narrative that "AI only burns money with no commercial return."
Under this framework, the capacity expansion by Samsung and SK Hynix should not be simplistically interpreted as a "risk of overcapacity," but rather as a real bet by the industrial chain on AI-driven global economic prosperity.
From Caution to Full-Throttle Expansion: Samsung and SK Hynix Jointly Bet on an AI-Driven Memory Supercycle, a Decade-Long Global Chip Supply and Demand Gamble
After these two South Korean tech giants fully supported the government's push for semiconductor and large-scale AI data center construction, President Lee Jae-myung praised them and bowed deeply. This is clearly different from the previously more restrained approach to chip capacity expansion, a restraint shaped by decades of painful "boom-bust cycles" in the memory chip industry.
According to the plan, the South Korean government hopes the country can double its memory chip production capacity within five years. Samsung Electronics Co., Ltd. and SK Hynix will also accelerate construction of wafer fabrication plants within the existing Yongin semiconductor cluster, significantly shortening the typical 7 to 12-year construction cycle and bringing new capacity online earlier.
The world's two largest memory chip manufacturers have jointly committed to investing around 3200 trillion won (approximately $2.07 trillion) in the semiconductor sector. This includes a new 800 trillion won chip cluster in southwestern South Korea, along with previously announced projects.
As global AI investment surges, South Korea is emerging as a core winning force in AI investment, driven by the absolute dominance of Samsung Electronics Co., Ltd. and SK Hynix in high-bandwidth memory (HBM) and the broader DRAM and NAND memory chip markets. HBM is crucial for the production of advanced AI GPUs/AI ASIC processors led by Nvidia, Alphabet, and Advanced Micro Devices.
Year-to-date, shares of SK Hynix and Samsung Electronics Co., Ltd. have risen 307% and 179%, respectively. This announcement comes as super-customers, from AI hyperscale cloud vendors to the world's largest consumer electronics manufacturers like Apple, are demanding more memory chips against a backdrop of global chip supply shortages.
However, some analysts point out that chip fabrication plants take years to build and ramp up, meaning most of the new capacity will not arrive until deeper into the next decade. Morningstar analyst Jing Jie Yu stated, "We believe memory chip pricing will remain a function of demand and supply, and the accelerated capex over the next decade further raises the risk of long-term overcapacity." He added that the memory chip boom still depends on whether AI hyperscalers continue to expand at the current strong pace; if the speed remains this high, all "boom-bust cycles" will continue to be absent.
Professor Lee Jong-ho from Seoul National University's Department of Electrical and Computer Engineering said the investments seem to be moving too fast. He stated, "This is the kind of investment that could determine a company's future." "No one knows what the situation will be like in three years. We need to respond quickly when demand is strong, but after that, there is uncertainty in demand, and decisions should be made carefully."
The soaring profit margins for Samsung and SK Hynix's memory businesses are a relatively recent phenomenon, driven by the global AI frenzy and severe memory chip shortages, with prices for all types of memory chips nearly doubling in the first quarter alone. Past chip price downturns have severely impacted these companies, pushing SK Hynix to the brink of bankruptcy in 2001 and causing both firms to record huge losses in 2023.
But in recent months, as the global wave of AI investment has grown stronger, ruling party lawmakers and government officials began proposing the idea of relocating some chip fabrication plants to the country's southwest region. This area is a crucial support base for them but was largely bypassed during South Korea's industrial rise.
According to scenes captured by journalists, South Korean President Lee Jae-myung and SK Group Chairman Chey Tae-won bowed to each other at a government briefing on three super-projects held at the Blue House guesthouse in Seoul, with Samsung Electronics Co., Ltd. Chairman Lee Jae-yong looking on. Just two months ago, when asked by a lawmaker about plans to build a chip fabrication plant in the southwest, SK Hynix Chairman Chey Tae-won expressed skepticism. He said, "I'm not sure semiconductors are necessarily an area you have to expand into."
Neither of the South Korean tech giants explained what prompted their change of heart. Both Samsung and SK Hynix said that with the government's promise to expedite approvals, they will now be able to accelerate the Yongin project as part of efforts to meet surging demand.
Nomura analyst CW Chung said, "Investing in other regions could be a way to hedge uncertainty around the Yongin semiconductor cluster." However, the analyst added that he believes the companies have learned from countless past memory chip cycles.
CLSA senior analyst Sanjeev Rana stated, "A memory industry downturn is clearly a risk to the plan." But "if signs of overcapacity emerge, memory chip producers will still retain the flexibility to adjust the pace of investment."
Samsung's plan envisions investing 2100 trillion won in chip production by 2040, though the company noted spending could be adjusted based on market supply-demand conditions and actual global business needs. In 2024, during a sustained downturn in global chip demand, Samsung paused construction of its P5 chip fabrication plant in Pyeongtaek, south of Seoul, for nearly two years before resuming work late last year.
At Monday's presentation attended by President Lee, SK Hynix's Chey Tae-won said the company would monitor demand conditions when determining the scale of investment, but he added that it remains difficult to fully resolve supply shortages for now.
However, for the South Korean government, maintaining momentum is a strategic priority. Presidential Chief of Staff Kang Hoon-sik said on Monday the government would expedite approvals and seek to complete the chip cluster construction before President Lee's term ends in 2030.
iM Securities research head Koh Taebong said he sees the move as a bet in the right direction for South Korea as the world enters a historic, AI-driven transformation era. He stated, "The message from yesterday's landmark presentation is that South Korea is seriously considering becoming one of the world's top three AI superpowers. This isn't just a slogan—the government is committing real money to achieve it."
From Claude's Profit Outlook to HBM/DRAM/NAND Capacity Expansion: South Korea's $2.07 Trillion Bet Rewrites the AI Megabull Market from a "Cash-Burning Narrative" to a Global Cash Flow Closed Loop
For AI revenue prospects, Anthropic, which launched a series of heavyweight AI agents in February that rattled global software stocks, recently revealed a strong profit trajectory. This robust revenue data and better-than-expected profit path will likely accelerate tech giants' efforts to build their "AI computing power super-empires."
Anthropic's latest developments, projecting Q2 revenue to double and achieve its first operating profit, highlight how the AI industry chain is shifting from a "cash-burning narrative" to a "cash flow cycle narrative." Most significantly, Anthropic's monthly ARR surged by $11 billion, equivalent to the combined scale of SaaS giants Palantir, Snowflake, and DataBricks over a decade—an unprecedented miracle in capitalist history.
Anthropic also expects Q2 revenue to jump from $4.8 billion in Q1 to $10.9 billion, with operating profit around $559 million. This indicates that frontier AI applications are not merely consuming computing power but are beginning to convert enterprise programming, agent workflows, cybersecurity, and data analysis into high-value token revenue.
Exploding demand for Claude and AI tools is pushing Anthropic close to its first profitable quarter. Furthermore, its computing power cost per dollar of revenue decreased from about 71 cents in Q1 to about 56 cents in Q2, showing that scale effects and inference efficiency are improving the economic model of AI applications.
Additionally, a research report suggests that for every $100 increase in US imports of AI-related hardware, the GDP of major Asian suppliers may increase by $35, accelerating the formation of an "AI-driven super surplus" in Asian economies. This also means the AI boom is no longer just a story of valuation expansion at the application layer for OpenAI/Anthropic, but a global capital cycle of "US AI application profit expansion—large cloud vendor Capex—Asian AI computing power infrastructure hardware exports—dollar asset repatriation."
Anthropic's strong performance outlook, coupled with South Korea's $2.07 trillion chip capacity expansion bet—a figure representing the combined committed investment by Samsung and SK Hynix—thoroughly reinforces the closed-loop investment logic of the "AI megabull market": "US AI application revenue generation—Asian computing hardware windfall profits—dollar asset repatriation—continued support for AI capital expenditure." It also strengthens the "cash flow re-circulation cycle" surrounding the AI megabull market.
Therefore, the capacity expansion by Samsung Electronics Co., Ltd. and SK Hynix should not be simplistically interpreted as a "risk of overcapacity," but rather as a real bet by the industrial chain on AI-driven global economic prosperity. If AI agents, enterprise programming, inference services, data analysis, AI search, and multimodal applications continue to drive token consumption, memory demand could show nonlinear growth: HBM determines GPU/ASIC performance release, DRAM supports inference and caching, and NAND/enterprise SSDs carry massive data, vector retrieval, model context, and AI agent long-term memory.
Since wafer fabrication plant construction and capacity ramp-up take years, the large-scale release of South Korea's new capacity will likely occur deeper into the next decade. In the short to medium term, this could instead strengthen the pricing logic of "supply struggling to keep up with AI demand."
The grand narrative of the "Circular AI Boom" is perhaps the only one capable of linking the US AI stock bull market, Asian export surpluses, the semiconductor equipment supercycle, and the memory price hike cycle: the US is responsible for creating AI demand and software monetization, Asia provides the physical computing power foundation, and capital markets are responsible for re-circulating cash flow, surpluses, and valuations.
Observing global capital flows, it currently appears more like the AI megabull market has entered its second phase. The first phase was driven by hyperscale training clusters dominated by AI GPUs and ASICs. The second phase is beginning to diffuse comprehensively towards "new AI centers" including the data center power chain, HBM/DRAM/NAND, advanced packaging, liquid cooling, data center CPUs, optical communication/interconnects, high-performance Ethernet network infrastructure/data center DCI high-speed interconnects, as well as AI application terminals like PCs, wearables, humanoid robots, autonomous driving, and space AI computing power infrastructure represented by SpaceX.
Wall Street giant Goldman Sachs has repeatedly raised its target for South Korea's KOSPI benchmark index this year, betting that the Korean stock market—which has surged about 100% year-to-date with high volatility—can continue its rally. Goldman Sachs has lifted its target for the KOSPI from 8000, to 9000, and now to 12000 points (compared to a close around 8470 on Tuesday). The core logic is not macro stimulus or monetary policy factors, but rather the long-term memory upcycle and the continued massive explosion in AI memory/storage chip demand, with Korean leaders like SK Hynix and Samsung Electronics Co., Ltd. being the primary beneficiaries of the global AI computing chain demand explosion.
In Goldman Sachs' view, the global bull market around the AI computing chain is far from over. The market's main theme has evolved from the "programming/code-driven software light-asset valuation expansion" prevalent since 2008 to a "re-pricing of a series of physical assets for AI computing infrastructure."
Goldman Sachs' latest calculations show the global AI capital expenditure baseline model is expected to grow from $765 billion annually in 2026 to $1.6 trillion annually by 2031, with cumulative capital expenditure from 2026 to 2031 estimated at around $7.6 trillion. US data center power demand is projected to rise from 31 GW in 2025 to 66 GW in 2027. This will directly spill over AI computing infrastructure investment into areas like server CPUs, DRAM/NAND/HBM, advanced packaging, liquid cooling, power equipment, transformers, gas turbines, grid connection equipment, data center REITs, and construction engineering.
Comments