AI Cycle at Year 4: Is This a Dot-Com Bubble or an Industrial Revolution? Korea's Stake in the Debate

The bull and bear cases for AI have never been more sharply polarized. On one side, Meritz Securities in Seoul projects the KOSPI at 11,500 based on a structural transformation thesis that draws explicit parallels to the industrial revolution. On the other, Bank of America's chief investment strategist Michael Hartnett warns that the narrow market breadth — just 20 of the S&P 500's 500 constituents hit new highs on May 29 — mirrors the dot-com peak of March 2000 with alarming precision. I think both sides present intellectually honest arguments backed by real data, and the truth probably lies somewhere in between. But the evidence tilts toward the structural shift camp — with important caveats that every investor in Korean equities needs to understand before making allocation decisions.

" alt="AI Cycle Year 4 Bull Bear Debate Infographic: S&P500 8-day rally at all-time highs, KOSPI 11,500 year-end target by Meritz Securities, Bank of America Michael Hartnett dot-com bubble warning citing only 20 of S&P 500 stocks hitting new highs, AI data center investment projected 0.9% of GDP in 2024 rising to 2.5% by 2030, Nvidia market cap $5.4 trillion world number 1. Sources: Meritz Securities, Bank of America, Bloomberg." style="max-width:100%;height:auto;border-radius:4px;">

The Structural Shift Thesis: Meritz's Bold Call and Historical Framework

Meritz Securities' research center published a report last week that sent genuine shockwaves through Seoul's normally cautious investment community. Their KOSPI year-end target of 11,500 implies another 30% upside from the current 8,800 level and would represent a new all-time high by a wide margin. The basis for this call is not speculative exuberance but a detailed bottom-up earnings projection. Meritz forecasts that 2027 net profits for Korea's listed companies will reach 989.8 trillion won — 11.8% above the current market consensus of 885.7 trillion won. With an estimated KOSPI return on equity of 24.1% and a target price-to-book ratio of 2.22x, they calculate the fair KOSPI index level at 11,763.

The core of Meritz's argument is that the AI cycle, now entering its fourth year since the ChatGPT launch in November 2022, is still in its early stages when measured against historical industrial revolutions. The report draws explicit comparisons to three transformative eras. The electrification era of the 1920s saw US manufacturing investment in electricity-related infrastructure exceed 3% of GDP at its peak. The 19th-century railroad construction boom absorbed over 5% of US GDP at its zenith, transforming American commerce and settlement patterns. The 1990s internet infrastructure investment cycle peaked at approximately 1.5-2% of GDP as fiber optic networks were laid and data centers were built. Today's AI data center investment — projected to rise from 0.9% of global GDP in 2024 to 2.5% by 2030 — follows a trajectory that is strikingly similar in scale and duration to these historical precedents.

I've been tracking this historical comparison framework since early 2025, when a handful of investment banks first began circulating it, and I think it is the most intellectually honest and useful framework for evaluating AI's true economic significance. If you accept that AI represents a general-purpose technology comparable to electricity, railroads, or the internet — a plausible thesis given AI's potential to transform productivity across every sector — then we are currently in the infrastructure build-out phase of the cycle. This is analogous to laying railroad tracks across the American continent or building the electrical grid in the 1920s. Under this framework, the most transformative applications and productivity gains are still ahead of us, not behind us. On Bluesky, the Financial Times reported that "US convertible bonds set for record year as issuers harness AI boom," suggesting that corporate America is placing real capital bets on this thesis — hundreds of billions in convertible issuance to fund AI infrastructure expansion.

Meritz's research center believes the AI rally "is not a simple tech stock rally limited to a few AI beneficiaries, but the initial phase of a structural transformation across the entire Korean economy." In my view, the critical phrase here is "initial phase." If Meritz's historical framework is correct, the current cycle has years, not months, of runway remaining. The question is whether the market's pricing already reflects that potential or still discounts it.

" alt="AI Infrastructure Investment Historical Comparison Infographic: AI data center investment projected 0.9% of global GDP in 2024 rising to 2.5% by 2030 compared to 1920s electrification investment 3%+ of GDP, 19th century railroad construction boom 5%+ of GDP, 1990s internet infrastructure 1.5-2% of GDP. Korea 2027 net profit forecast 989.8 trillion won by Meritz versus consensus 885.7 trillion won, KOSPI fair value estimate 11,763 based on 24.1% ROE and 2.22x PBR. Sources: Meritz Securities, BofA Global Research, BLS historical data." style="max-width:100%;height:auto;border-radius:4px;">

The Bubble Warning: Hartnett's Dot-Com Parallel

Bank of America's Michael Hartnett is the most prominent and credible voice warning that the AI rally has entered dangerous territory. His May 29 research note pointed out that only 20 of 500 S&P 500 stocks hit new highs that week — a market breadth reading that precisely mirrors the behavior of the NASDAQ composite at the dot-com peak of March 2000. When the market's gains are concentrated in a shrinking handful of names while the vast majority of stocks are failing to participate, it suggests that the rally is being driven by momentum and narrative rather than broad-based economic strength.

The data supporting Hartnett's bearish view is substantial and cannot be dismissed as mere alarmism. The NYSE advance-decline line — a cumulative measure of the number of advancing stocks minus declining stocks — peaked in late March 2026 and has been in a steady decline since mid-April, creating a classic bearish divergence with the price index. The CBOE put-call ratio's 5-day moving average dropped to 0.452 on May 29, the lowest level since March 30, 2022, and consistent with readings that have historically preceded major market tops. Market participants are almost unanimously betting on continued upside — a level of consensus that has historically been a contrarian signal.

The S&P 500's equity risk premium — the excess yield that stocks offer over risk-free government bonds — has effectively converged to zero for the first time since the dot-com era. This means investors are receiving no additional compensation for the risk of owning equities versus holding supposedly risk-free Treasury bonds. BCA Research strategists have warned that "narrow market breadth often signals underlying vulnerability in equity markets," noting that only 55% of S&P 500 constituents are trading above their 200-day moving average — a remarkably low proportion for an index trading at all-time highs. On the June 1 trading session, only two sectors — technology and energy — traded higher, while declining stocks outnumbered advancing ones across the broader market.

I think Hartnett's warning deserves serious consideration but does not invalidate the structural shift thesis for three reasons. First, the dot-com comparison overlooks a critical difference in fundamentals: in 1999-2000, the market cap leaders (Cisco at 200x earnings, JDS Uniphase at 300x sales, Pets.com with no earnings) were detached from any reasonable valuation framework. Today's AI leaders — Nvidia at $5.4 trillion with a P/E of approximately 55x, Microsoft at $3.4 trillion at 35x, Alphabet at $4.5 trillion at 25x — are expensive but not absurd relative to their earnings growth rates of 30-100% annually. Second, AI may genuinely be a winner-take-most technology where concentration is structurally justified — not all 500 S&P 500 companies should participate equally in an AI-driven transformation. Third, the narrow breadth reflects the fact that AI is disrupting many legacy business models, and the laggards are being correctly priced for obsolescence.

The Korea Dimension: Why This Debate Matters for Seoul

This global AI debate has direct and profound implications for Korean equities. Korea's HBM memory duopoly means Samsung and SK Hynix are leveraged plays on the infrastructure build-out phase of the AI cycle. If the structural shift thesis prevails, Korean memory makers are among the most direct beneficiaries globally — every new AI data center needs HBM memory, and Korea controls over 90% of the global HBM market. If the bubble thesis prevails and AI capital spending contracts, Korean memory stocks would be among the hardest hit globally.

Business Insider reported on Bluesky on June 2 that "Anthropic's IPO filing adds momentum to a blockbuster lineup of potential tech offerings. With SpaceX preparing to go public and OpenAI widely expected to follow, investors are debating what comes next." A wave of marquee tech IPOs could absorb massive liquidity and pressure the valuations of existing AI beneficiaries. Korea's semiconductor stocks would not be immune to such a rotation. The second half of 2026 could see the most significant IPO pipeline in technology history — a potential liquidity drain that the bulls are not adequately pricing in.

Jensen Huang, Nvidia's CEO, is scheduled to visit Korea in June, and his meetings with Samsung and SK Hynix executives will be closely watched for signals about HBM4 development timelines and supply agreements. Nvidia CEO Huang stated on June 2 that the company "has capacity to supply robust AI growth despite constraints," as reported by Reuters on Bluesky — a constructive signal for the entire AI supply chain including Korean memory makers.

" alt="AI Bubble Debate Market Breadth Infographic: Bank of America Hartnett warning only 20 of 500 S&P 500 stocks hit new highs mirroring dot-com March 2000 peak, NYSE advance-decline line bearish divergence since mid-April, CBOE put-call ratio 5-day MA 0.452 lowest since March 2022, S&P 500 equity risk premium near zero first time since dot-com, only 55% of S&P 500 stocks above 200-day moving average. Anthropic IPO SpaceX OpenAI IPO pipeline 2026 liquidity concern. Sources: BofA, CBOE, Business Insider." style="max-width:100%;height:auto;border-radius:4px;">

The debate between structural shift and bubble is not merely academic for Korean investors - it has direct portfolio implications. If the structural shift thesis is correct, Korea's HBM duopoly represents a generational investment opportunity that could deliver multiples of the current market cap over the next decade as AI permeates every industry. If the bubble thesis is correct, the current elevated valuations for semiconductor stocks represent a selling opportunity that may not recur for years. My framework for resolving this debate focuses on three leading indicators: capital expenditure commitments by hyperscale cloud providers, HBM pricing trends, and the pace of AI application deployment. If these three indicators remain constructive through Q3 2026, the structural shift thesis gains credibility. If they show signs of peaking, it is time to reduce exposure regardless of valuation arguments.

What This Means for Global Portfolios — My Take

Here is how I reconcile the two competing narratives into an actionable framework. I assign approximately a 60% probability to the structural shift thesis and 40% to the bubble thesis — tilted toward the bulls but with substantial risk awareness.

My base case: The AI cycle has further to run, driven by real and accelerating earnings growth from AI infrastructure deployment. Korea's semiconductor sector is uniquely positioned as the dominant supplier of the essential memory technology that every AI system requires. The KOSPI reaches 10,000-10,500 by year-end 2026 — below Meritz's bold 11,500 target but still a healthy 15-20% return from current levels. Samsung re-rates to 16-18x forward earnings as the market prices in sustained HBM demand through 2027-2028.

I would advise holding Korean semiconductor names as core long-term positions with a disciplined risk management framework. The asymmetry of returns is favorable: if the structural shift thesis prevails, Samsung and SK Hynix could deliver another 30-50% upside as they re-rate toward global semiconductor peer valuation multiples. If the bubble thesis prevails and a correction comes, the maximum drawdown is probably 20-30% — painful but survivable with a predetermined stop-loss strategy. I use a 15% trailing stop on individual positions as my risk control mechanism.

The key risk that keeps me cautious despite my generally constructive outlook is the tech IPO pipeline. If OpenAI (potential valuation $300 billion+), SpaceX ($200 billion+), Anthropic ($100 billion+), and other AI-related IPOs all come to market within a 12-month window, the liquidity absorption could be substantial. This is not a reason to sell today, but it is a reason to maintain some cash reserves and be selective about entry points. The second half of 2026 could see the most concentrated pipeline of high-profile tech IPOs in market history — the modern equivalent of the 1999-2000 IPO wave that ultimately drained liquidity from the market.

🔍 Related Keywords

• AI stock market bubble debate 2026 structural shift vs dot-com comparison
• Meritz Securities KOSPI 11,500 target year-end 2026 forecast
• Michael Hartnett Bank of America narrow market breadth warning
• Korea semiconductor HBM memory AI infrastructure investment cycle
• Samsung Electronics SK Hynix HBM3E valuation forward earnings
• Anthropic IPO SpaceX OpenAI tech IPO pipeline 2026 liquidity
• Nvidia Jensen Huang Korea visit HBM4 supply agreement June 2026