● AI Gold Rush, Grid Meltdown, Not Another Dotcom

Michael Burry’s Rebuttal to “Bubble On”: The 2025 AI Cycle Unlike the Dot-Com Bubble, 2026 Will Be the “Year of AI Apps”

In terms of returning to the basics of AI investing and reviewing the big picture, here is a news-style summary that extracts only the core points from Coatue’s latest report.
This article includes 1) a quantitative comparison with the Dot-Com Bubble, 2) 6 AI bubble checklist items and the current position, 3) a shift from M7 to “AI infrastructure & power,” 4) the true ROI of AI that is not clearly visible in financial statements, and 5) the 2026 “Year of AI Apps” scenario along with investment strategies.
In particular, points that other media have not extensively covered – “unlisted-centric revenues, cost-saving ROI, power & grid bottlenecks” – are organized separately.
Core keywords such as global economic outlook, interest rates, inflation, Nasdaq, and GDP are also included in context.

Coatue Report News Summary

• In conclusion: 2026 marks the era of AI apps.
• ROI is considered to be entering an improvement phase.
• The macro outlook is seen as supporting the AI rally with expectations of easing inflation and stabilized interest rates.
• However, “a tree does not grow to the sky.” Downside risk management is a prerequisite.
• A “Fantastic 40” list is provided, covering both listed and unlisted companies, which is a strong point.
• SK Hynix is the only Korean semiconductor company included.

Where Has the AI-Driven Market Moved?

AI-related stocks have continued to outperform non-AI stocks within the S&P 500.
The M7 (Mega Cap 7) has seen relative underperformance since 2025, and the leadership in the AI theme is spreading over a wider range.
The biggest beneficiaries this year have shifted to “power, energy, and grid.”
Due to power shortages relative to data center expansion demand, US power utilities like Constellation and Vistria have surged in the 50% range year-to-date.
In Korea, power facility companies such as Hyundai Electric and Hyosung Heavy Industries are in line with this context.
Within the semiconductor chain, TSMC (foundry) and Broadcom (ASIC & accelerator ecosystem) have maintained relative strength.
Although AI software (Palantir, Oracle, Snowflake) has risen, the excess return weight is tilted towards infrastructure and power.

Macro Background: The Coexistence of Easing Inflation Expectations and Valuation Concerns

JP Morgan and Citi have given positive evaluations on solid consumer spending, stable delinquency rates, and the momentum of US growth.
After a temporary rebound, CPI has been observed re-entering an easing phase.
Expectations of stable interest rates are favorable for tech and growth stocks.
However, the S&P 500’s P/E stands at approximately 22×, higher than the long-term average (about 17×), which is a burden.
Tech trades at 28× vs non-tech at 18×, showing a multiple spread, yet it is fundamentally different from the dot-com era extremes.

6 Bubble Checklist Items, Checked as of 2025

• A strong theme exists: AI clearly fulfills the role of a game-changer.
• Rapid adoption speed: Led by ChatGPT, the adoption rate is faster than that of the PC and Internet eras.
• Revenue & profit logic: Direct revenue visibility is limited, but cost savings and indirect revenue contributions are significant.
• Leverage/Capex: While data center capex is large, a significant part is funded through operating cash flows.
• Valuation: Though in an expensive range, it is not at dot-com levels.
• Speculation/IPO Overheating: Compared to past extremes, IPOs and stock issuances are lower, and the risk of excess cash is manageable.

Quantitative Comparison with the Dot-Com Bubble: Expensive, but Different

Nasdaq’s P/E is in the late 20s, compared to 80–90× at the peak of the dot-com era.
The S&P 500’s capex-to-operating cash flow ratio is about 46% currently, compared to 75% during the dot-com era, indicating more room.
The scale of IPOs and equity offerings is lower than that during the dot-com era and 2021, so it cannot be seen as an overheating indicator.
The margin loan balance stands at about 3.8% of GDP, which is similar to the 2021 peak, but when viewed relative to market capitalization, the burden is somewhat alleviated.

Is Market Concentration Necessarily Bad?

Market concentration phases have recurred in the past, and many examples of long-term “buy and hold” have yielded results.
The big tech companies, based on global revenues, have cash flows and financial health far superior to the past, providing a logical basis for a premium.

The True ROI of AI, Not Captured by Numbers

• Revenues obscured by unlisted entities: A significant portion of revenues from large AI models/tools occurs in the unlisted ecosystem and does not appear in listed indicators.
• Cost-saving effects: AI’s benefits initially appear in cost (COGS & Opex) savings and productivity improvements.
• Changes in workforce structure: Reduced new hiring means that the effect does not appear as revenue on the books.
• Example 1) Microsoft: Despite about a 6% reduction in employee count from the peak, growth was maintained.
  The top 50 tech companies have an estimated potential to save $75 billion annually (approximately 100 trillion KRW) in labor costs if adopted enterprise-wide.
• Example 2) C.H. Robinson: With AI-based dispatch/route optimization, throughput per employee increased 1.5 times, workforce reduced by 30%, and the stock surged 70% in six months.
• Example 3) Rocket Mortgage: Loan processing speeds improved sixfold.
• Example 4) Open Evidence: For AI-powered medical search, 40% of US physicians adopted it within 15 months.

Power & Grid as the New Key Variables

Power supply constraints have become bottlenecks for data center expansion.
Investments in energy and grid are elevated to essential infrastructure for the AI cycle.
While growth rates may slow, the absolute scale will continue to expand.

Risk Map and Bearish Scenario

• If AI is not a game-changer, the theme could collapse.
• Slower adoption and cuts in enterprise budgets.
• Oversupply of GPUs and decreased profitability of data centers.
• Credit risk transmission due to a breakdown in the vendor financing circular structure.
• In a high-valuation environment, deceleration in earnings (E) could lead to multiple compression.
• Simultaneous shocks from regulation, sharp increases in power rates, and supply chain issues.

2026 “Year of AI Apps”: Where the Money Flows

• Coding tools: AI IDEs such as Cursor see exponential increases in token processing, with soaring demand even at high price tiers.
• Vertical SaaS: Quick monetization in healthcare (medical literature & clinical summaries), customer support, sales automation, and financial underwriting.
• Commerce & Advertising: Companies like Meta and Google will enhance the efficiency of their existing models with AI-driven recommendations and bidding optimization.
  The full-scale launch of AI commerce by OpenAI, Amazon, and Shopify will serve as a trigger for ARPU improvement in 2026.
• Rotation from Infrastructure to Software: After initial gains in infrastructure, apps and platforms are more likely to capture margins from 2026 onward.

Investment Strategy Checklist

• Do not sell too early: The rally tends to be longer during widespread adoption phases.
• Prepare for rotation: Monitor the shift from infrastructure (semiconductors & power) to software and apps focused on profitability.
• Track unlisted companies: Keep an eye on enterprise revenues, paid conversion indicators for vertical apps from companies like OpenAI and Anthropic.
• Hedge against tail risk: Individuals should secure cash positions, and institutions should consider puts and tail hedges.
• Watchlist: Model performance updates, token pricing/inference cost trends, power rates/grid expansion, and the rate of long-term enterprise license contracts.

Coatue “Fantastic 40” Snapshot and Points for Korean Investors

A list of 40 preferred stocks covering both listed and unlisted companies is presented, and SK Hynix is included in the list.
Korean investors should also check power, transformers, and ultra-high-voltage facility chains along with memory and HBM.
In the global chain, TSMC, Broadcom, Meta, Oracle, Snowflake, and utility/grid companies form the core value chain.

Key Rebuttals to Michael Burry’s “Bubble On”

• Valuation: It is expensive, but unlike the extremes of the dot-com era, its cash flow-based absorption capacity is high.
• Financing: There is no IPO overheating, and the structure of funding capex through operating cash flows is an advantage.
• Adoption speed: It is the fastest consumer and enterprise adoption in history.
• Productivity data: Although less visible on listed financial statements, cost savings and cumulative unlisted revenues are mounting.

The Most Important Content Missed by Other Media

• The key point is that AI’s profitability appears first in cost savings rather than in revenues.
• Instead of the accounting revenue line, improvements like reduced labor costs, lower cloud expenses, and shortened operating days accumulate rapidly.
• Power and grid set the new marginal production costs; thus, the pace of power investments, rather than improvements in model performance, determines the upper end of valuations.
• While the stability of the vendor financing structure is supported by big tech cash flows, the secondary and tertiary ecosystems carry different risk profiles.

Final Check: What to Keep Watching

• Enterprise revenues, paid renewal rates, and the speed of long-term contract expansion for OpenAI and Anthropic.
• The pace of decline in inference costs vs. the rate of increase in power rates/grid expansion.
• User-to-paid conversion rates and cohort retention for vertical apps.
• How the resilience of US CPI/PCE deceleration and the policy interest rate path affect Nasdaq multiples.

< Summary >2026 is entering the era of AI apps, and ROI will first appear through cost savings and indirect revenues.
While valuations are high, unlike the dot-com era, cash flow-based capex absorption is achievable and IPO overheating is not a concern.
Power and grid constraints have become the new key variables, and the profit center is shifting from infrastructure to apps and software.
Risk management is essential, and continuous monitoring of unlisted indicators and enterprise revenues is imperative.

[Related Articles…]
2026, the Year of AI Apps – Power Grids Will Choose the Ultimate Winner
Nasdaq Valuations and Interest Rate Shift Signals: What’s Left for Tech Stocks?

*Source: [ 내일은 투자왕 – 김단테 ]

– 마이클 버리의 버블론 전격 반박. 닷컴버블과 오늘날은 다르다.

---

---