How does the AI computing power bubble burst transmit to the crypto market? From the Korean stock market circuit breaker to the linkage of risk assets

On June 23, 2026, the global capital markets experienced a trading day widely known as "Black Tuesday." The Korea KOSPI index closed down by 910.71 points, a 9.99% drop, at 8,203.84 points. During trading, a drop of over 8% triggered a circuit breaker, pausing trading for 20 minutes. Samsung Electronics and SK Hynix both fell more than 12%, becoming the focal points of this plunge. The Nikkei 225 index declined by 3.55%, closing at 69,788.38 points. Nasdaq 100 futures dropped 2.5% in pre-market trading. Bitcoin fell to $62,000 amid a broad retreat of risk assets.

This turbulence was not an isolated market event but a concentrated release of long-standing structural pressures in the secondary market.

Why South Korea’s Stock Market Became the "Epicenter" of the AI Computing Power Bubble Burst

The Korea KOSPI index recorded an unprecedented single-day decline of 9.99%. This drop ranks among the top in KOSPI history—in percentage terms, it is the fifth-largest single-day decline ever.

The reason South Korea’s stock market became the "epicenter" of this sell-off is closely related to its index structure. Samsung Electronics and SK Hynix, two giants in storage chips, together account for about 50% of KOSPI’s weight. These two companies are the most core suppliers of storage chips in the global AI infrastructure supply chain, and their stock performance directly reflects the global AI capital expenditure outlook. When the market begins to reassess the return logic of AI investments, these two stocks naturally become the primary targets for capital withdrawal.

Samsung Electronics plummeted 12.31% that day, and SK Hynix fell 12.47%. South Korean financial regulators harshly criticized the leveraged ETFs tracking Samsung Electronics and SK Hynix, stating these products "not only profit at the expense of retail investors but also serve little other purpose." The leverage effect further amplified the market’s downward slope.

Why the $725 billion Capital Expenditure by the Four Major Cloud Providers Sparks Serious Market Doubt

The underlying driver of this sell-off is a collective skepticism about the return on the massive capital expenditures of the four major cloud providers.

According to Goldman Sachs’s June 2026 updated forecast, the four largest hyperscale data center operators—Alphabet (Google), Amazon, Microsoft, and Meta—will spend a total of $725 billion on capital expenditures in 2026, a 77% increase from $410 billion in 2025. In 2024, this figure was only about $250 billion. Over three years, these four cloud giants’ capital spending has nearly tripled.

However, the pace of capital expenditure expansion has diverged sharply from revenue growth. Bank of America estimates that in 2026, about 90% of the operating cash flow of hyperscale cloud service providers will be consumed by capital expenditures, up from 65% in 2025. Barclays projects that this year, Alphabet and Meta’s free cash flow will plunge nearly 90%.

What further alarms the market is the fragility of their financing structures. In 2025, bond issuance by hyperscale cloud companies exceeded $100 billion, four times the annual average of $28 billion from 2020 to 2024. By October 2025, debt related to AI had ballooned to $1.2 trillion, accounting for 14% of the JPMorgan US Liquidity Index, surpassing the major US banks.

When a company needs to borrow an amount equivalent to 90% of its operating cash flow to sustain capital expenditures, the market’s valuation logic must shift from a "growth narrative" to a "return validation" perspective.

How Falling Prices for Computing Power Leasing Break the "Computing Power Scarcity" Belief

Over the past three years, the core narrative of the AI industry was built on a simple logical chain: the scarcer the computing power, the more reasonable the capital expenditure; the larger the expenditure, the higher the valuation; the higher the valuation, the easier the financing. This self-reinforcing cycle was almost unquestioned.

But by mid-2026, every link in this logical chain is undergoing stress testing. The most direct signal comes from the computing power leasing market—Nvidia’s flagship AI chip B200’s hourly operation price has fallen from $6.11 on May 30 to $4.22 as of June 21.

The continuous decline in leasing prices indicates that the "computing power scarcity" narrative is being undermined by supply and demand changes in the physical market. If scarcity persists, prices should remain firm and support ongoing capital expenditures. When prices start trending downward, the fundamental logic supporting high valuations begins to crack.

Meanwhile, spot prices for computing power and forward contract prices have diverged in a rare manner, revealing a deep contradiction between short-term supply surplus and long-term demand expectations. Tech giants are tightening AI budgets, power and engineering delivery capacities are exposing physical limits, and capital markets are beginning to evaluate each AI company based on ROI.

Why Bitcoin’s Decline Is Synchronized with Nasdaq Futures

On June 23, Bitcoin briefly surged to $65,500 in the morning but then quickly dropped, closing at around $62,900, down about 2% over 24 hours. According to Gate data, Bitcoin traded between $63,900 and $64,300 that day, with intraday volatility between $62,000 and $65,500.

The synchronized decline of Bitcoin and Nasdaq futures reflects the risk asset nature of cryptocurrencies in the current macro environment. Although by early June 2026, Bitcoin’s 40-day correlation with Nasdaq had fallen to zero, from a broader perspective, Bitcoin’s correlation with the Nasdaq 100 remains around 0.45—above its 10-year average. This indicates that Bitcoin still struggles to detach from tech stocks during systemic risk events.

The transmission path of this decline is clear: large US tech stocks weakened significantly during Monday’s trading; during Asian trading hours, Japanese and Korean markets followed suit; pre-market, Nasdaq 100 futures fell 2.5%; Bitcoin retreated to around $62,000. This is a typical transmission chain from "AI computing stocks → global risk assets → Bitcoin."

It’s worth noting the shift in market drivers. For weeks, Bitcoin’s movements have mainly been influenced by Middle East geopolitical events. Now, with the release of the US-Iran peace roadmap, the dominant market force has shifted to the same AI-driven tech trading that has driven stock prices to record highs—yet as this trade wobbles, cryptocurrencies are also falling.

How Crypto Mining Companies Face Risks and Opportunities Amid AI Overcapacity

The supply-demand reversal in the AI computing power market is profoundly changing the survival logic of crypto mining companies.

Since May, Bitcoin network hash rate has declined by 145 EH/s, the first such contraction in six years, as miners shift electricity and sell BTC to fund AI data center construction. In Q2 2026, hash rate decreased by 5.8% to 1,004 EH/s. Currently, electricity accounts for 70% to 90% of mining operational costs, and competition from AI data centers makes cheap power harder to access.

In terms of revenue, AI data centers generate $200 to $500 per megawatt, while Bitcoin mining yields only $57 to $129 per megawatt. This revenue gap makes it financially logical for miners to shift electricity toward AI workloads.

However, this transition faces significant capital constraints. Bitcoin miners face an estimated short-term funding gap of about $50 billion when converting electricity assets into AI data centers. Meanwhile, the falling prices of AI computing power leasing are compressing profit expectations for this transition. If leasing prices continue to decline, the profitability of miners’ AI transformation will be further squeezed, creating a "front-loaded investment, delayed returns, and uncertain ROI" dilemma.

Why Micron’s Earnings Report Is a Key Test for Risk Assets

Market focus is on Micron Technology’s upcoming quarterly earnings report, scheduled for June 24 (Wednesday). This report is widely seen as a critical test of whether AI spending can sustain the current market rally.

Micron’s earnings are crucial because they directly relate to the demand for memory chips—the core hardware component of AI infrastructure. The sharp decline in SK Hynix and Samsung Electronics’ stocks is essentially a pre-emptive pricing of Micron’s earnings. If Micron’s results fall short of expectations, it will reinforce market doubts about the return on AI capital expenditures; if better than expected, it could provide a short-term respite for the oversold semiconductor sector.

Bloomberg strategists note: "Recent risks for regional chip stocks include increasingly unstable market structures and Micron’s upcoming earnings report after US markets close Wednesday. Concerns about whether US hyperscale cloud providers’ unprecedented AI infrastructure investments are prudent continue to rise."

From "Unlimited Drinking" to "Quantitative Rationing": Structural Turning Point in the AI Industry

June 23, 2026, "Black Tuesday," may be viewed as a dividing line marking the transition of the AI industry from its first half to its second half.

The first half was characterized by "unlimited drinking"—unlimited expansion of capital expenditure, soaring valuations, and boundless financing. The second half is characterized by "quantitative rationing"—capital demands returns, valuations are scrutinized, and financing faces constraints.

The decline in computing power leasing prices from their peak, tech giants collectively tightening AI budgets, and physical limits exposed in power and engineering delivery—these three cracks open simultaneously, pushing the AI industry into a new phase. Capital markets now evaluate each AI company based on ROI rather than solely relying on the "computing power scarcity" narrative to justify valuations.

For the crypto market, this means Bitcoin’s attribute as a "risk asset" will continue to be tested during the digestion of this AI computing power bubble. Whether Bitcoin can maintain its "digital gold" narrative amid a risk-off environment will largely depend on macro liquidity shifts and the crypto market’s fundamentals—including ETF capital flows, stablecoin regulation progress, and the supply structure post-halving.

Summary

The market turmoil on June 23, 2026, is essentially a collective reassessment of the return on over a trillion dollars of AI capital expenditure accumulated over four years. The circuit breaker in Korea’s KOSPI, the sharp plunge in Nasdaq futures, and Bitcoin’s retreat all reflect the same underlying logic across different asset classes. As leasing prices for computing power begin to fall, free cash flow of cloud giants comes under pressure, and AI-related debt expands, the "computing power scarcity" narrative can no longer sustain the existing valuation system.

For crypto market participants, understanding the evolution of the AI computing power bubble is fundamentally about understanding Bitcoin’s pricing logic as a risk asset. Until the return on AI capital expenditure is validated, Bitcoin’s high correlation with tech stocks may persist.

FAQ

Q: Why does the burst of the AI computing power bubble affect Bitcoin prices?

Bitcoin still participates in pricing as a risk asset in the current market environment. When doubts about the return on AI capital expenditures arise, funds withdraw from high-beta tech stocks and risk assets, and Bitcoin, as part of risk assets, naturally faces similar selling pressure. The clear transmission chain—"US tech stocks weaken → Asia markets plunge → Nasdaq futures fall → Bitcoin retreats"—reflects this linkage.

Q: Which companies are included in the $725 billion capital expenditure by the four cloud giants?

Refers to Alphabet (Google), Amazon, Microsoft, and Meta, four US hyperscale cloud service providers. According to Goldman Sachs’s June 2026 forecast, these four companies will spend a total of $725 billion on capital expenditures in 2026, up 77% from $410 billion in 2025.

Q: What does the decline in computing power leasing prices imply?

Leasing prices are the most direct indicator of supply-demand balance for AI computing power. Nvidia’s B200 chip hourly leasing price has fallen from $6.11 at the end of May to $4.22 in late June, indicating that supply is outpacing demand growth. This directly undermines the "computing power scarcity" narrative, which has been the core logic supporting high valuations of AI-related assets over the past three years.

Q: What impact does AI overcapacity have on crypto mining companies?

On one hand, competition from AI data centers drives up electricity costs for miners, with electricity accounting for 70% to 90% of operational costs. On the other hand, converting electricity assets into AI data centers faces an estimated short-term funding gap of about $50 billion. If leasing prices continue to decline, miners’ profitability expectations for AI transformation will be further compressed.

Q: Will Bitcoin continue to fall alongside tech stocks?

Currently, Bitcoin’s correlation with the Nasdaq 100 is about 0.45, above its 10-year average. Before the market validates the return on AI capital expenditures, this high correlation may persist. Key variables to watch include Micron’s earnings, Federal Reserve monetary policy, and capital flows within the crypto market itself.