Author: Godzilla, Ge Long

01

If one word could describe the global market in the last two days, the most fitting would be“chain explosion.”

First, last night's U.S. stocks, AI semiconductor collectively plummeted.

NVIDIA dropped nearly5%, with its market value being surpassed by Apple overnight, while the Philadelphia Semiconductor Index fell5% at the open.

SK Hynix was even harsher, crashing 7.47%, directly falling below the IPO issue price of U.S. stocks, leaving new investors buried;

Today, this stock panic has spread to the Asian market.

The Korean KOSPI opened with a direct plunge of 8%, triggering a circuit breaker, and continued to crash after trading resumed, finally closing down10.84%; the Nikkei225 once plunged over 2700 points, closing down nearly 4%.

Among the heavyweight stocks, SK Hynix plummeted 14.65%, losing over 47 billion USD in market value in just over a month;

Japan's Kioxia fared worse, crashing in a single day18.33%, having seen six single-day drops exceeding 10% since July, mocked as "the drop machine."

The domestic market didn’t escape either.

The ChiNext Index plunged7.35%, the Sci-Tech 50 fell 6.33%, with the storage sector and CPO sector as disaster areas; the Southern Double Long Hynix saw a crash of29%, while the Southern Double Long Samsung dropped over 25%.

This is just the performance of one day.

If we look from the peak of this round, many AI stocks have been halved, and for more aggressive ones, like the double long SK Hynix, the drop has reached as much as 80%.

A drop of 80% requires a rise of 4 times to break even.

In the first half of the year, A-shares had as many as 246 “double earnings funds” that skyrocketed due to heavy holdings in electronics and telecommunications, but since July, the average return of these 246 “double earnings funds” has been -23.87%, with an average drawdown of 52.8%! In less than a month, the number of “double earnings funds” has abruptly dropped to 19, with 227 funds “disappearing,” accounting for 92%!

The drop today can definitely be considered an extreme market crash in a global sense.

02

It must be said that these sudden negative news in the AI sector over the past few days occurring simultaneously indeed feels strange.

First, regarding NVIDIA, it was reported that NVIDIA reached a 500 billion USD AI partnership with SK Group, and is planning to provide 250 billion USD in financing guarantees for OpenAI.

Although this is not new, but during the time of severe fluctuations in the AI semiconductor sector, any non-positive factors are easily interpreted as negative.

While AI cloud vendors' capital expenditures remain at high levels, their consumption of corporate profits and cash flow has already heavily pressured the valuations of cloud vendors. Meanwhile, the revenue growth from the AI application end is weak, and the “killer app” has yet to be widely implemented, causing a serious disconnect between ROI and trillion-dollar-level capital investment.

Although NVIDIA remains the most perfect “shovel seller” in AI, if the return on investment for AI cloud computing vendors and the killer applications downstream of AI continues to fall short of expectations, then NVIDIA will transition from being the perfect “shovel seller” to an “infinite liability guarantor” in an industry marked by substantial debts.

Once this domino falls, the entire investment logic of the AI industry will change significantly.

This was also evident last night when news broke about NVIDIA's debts starting to have issues. The five-year credit default swap (CDS) for NVIDIA saw its guarantee cost rise by 0.14 percentage points, with an annual yield reaching 0.82 percentage points, marking the largest intraday increase since November 2025.

At the same time, according to data from the London Stock Exchange Group, the prices of five-year credit default swaps related to companies like Oracle, SpaceX, Alphabet, Amazon, Meta, Broadcom, have recently reached new historical highs.

Secondly, the narrative about China’s semiconductor supply chain reshaping the global competitive landscape is also notable.

The soaring market value of AI giants in the U.S. stocks and Korean stocks over the past two years is largely due to their monopolistic (scarcity) premium in the AI hardware and software industry chain.

However, this narrative logic has shaken since the beginning of this year.

Since this year, China's large AI models have continuously impacted globally. Recently, high-performance open-source models like Kimi K3 have rapidly iterated, significantly narrowing the technological gap with overseas closed-source models, further challenging the technological moats of European and American AI companies.

Open-source forces a price war, compressing profit margins, putting pressure on downstream AI companies, which further suppresses upstream computing power demand.

Next is high-end storage, which serves as the second critical bottleneck under high-end computing chips in AI hardware, historically dominated by the three giants: Samsung, SK Hynix, and Micron. However, yesterday, domestic storage kingChangxin Technology went public on A-shares, not only filling the critical gap in domestic storage but also accelerating the replacement of overseas products, breaking these international giants' long-term monopoly.

Moreover, there are reports today that domestic companies have begun mass production of immersion deep ultraviolet (DUV) lithography machines. The market interpreted this as breaking the long-term monopoly expectations of established semiconductor equipment giants, triggering a brutal sell-off in the global chip manufacturing equipment sector.

ASML saw a drop of over 8% during the day due to this news, while Tokyo Electron and Nikon both plummeted over 10%, likely as a result of this information.

As of now, in core areas such as computing chips, storage chips, and lithography machines, we have made significant progress; though we are not currently the very top tier, and even lag behind by two or three generations, it is sufficient to reverse the situation and form a new narrative logic. However, for corresponding AI giants, this is not good news.

03

It must be said that the biggest driver of the global AI stock plunge is the series of unprecedented, suffocating super maneuvers by the South Korean government:

First, they let the entire population heavily leverage, then, when everyone was going crazy, they abruptly slammed the brakes and violently deleveraged.

On May 27 of this year, the Korean Exchange approved the first batch of 16 leveraged ETFs that track single stocks of SK Hynix and Samsung at 2x leverage, saying it was a good idea to let the public share in the economic benefits of AI storage.

But later, as the stock market surged, the Korean government hastily slammed the brakes by raising the minimum margin requirement for chip leveraged ETFs and prohibiting providers from listing new single stock leveraged products, resulting in enforced deleveraging and ultimately cutting off the bull market process.

By mid-July, the cumulative forced liquidation scale in the Korean stock market had reached 344.2 trillion won, with over 1.2 million leveraged retail accounts hitting the margin call line. Among them, 320,000 to 360,000 accounts were fully liquidated by brokers, with some investors even owing brokers money. Market sentiment was already extremely fragile, easily tipping with the slightest pressure.

However, unexpectedly, the Korean government not only did not timely ease up but also worked with the Korean Central Bank to further tighten funds, squeezing leveraged accounts.

As early as July 16, the Korea Financial Commission introduced the first round of regulatory measures, halting new listings of single stock leveraged ETFs, prohibiting related marketing promotions, and raising the cash margin for leveraged ETFs to30 million won. Subsequently, regulation accelerated again, bringing forward the originally planned margin reinforcement regulations from August to be implemented on July 31.

On the same day, the Korean Central Bank announced it would raise the seven-day repurchase agreement rate from2.50% to2.75%.

On July 28, Korean regulators once again signaled extreme regulatory control. The head of the Financial Commission, Lee Eok-yun, publicly stated that if the current policy could not cool the market, new rounds of regulatory measures would be introduced.

It is reported that the core plan includes strict control of leveraged investment ratios, limiting personal single stock leveraged investments to20% of total financial investment assets, preventing retail investors from heavily betting on a single track from the source.

This is essentially a rhythm meant to eradicate leveraged accounts.

Thus, we have witnessed this epic plunge in the Korean stock market.

In just seven months this year, the number of circuit breakers in the Korean stock market has already exceeded the total of the past 26 years.

It must be said, this is indeed a shoddy setup!

Additionally, besides the events in the AI field, there are looming heavyweight nodes approaching that are also causing market declines.

On July 28, a new round of nuclear talks between the U.S. and Iran will take place, and if the outcome is below expectations, it could disrupt already weak energy and risk assets;

Following closely, on the 30th, the Federal Reserve's monetary policy meeting will commence, and if the new Federal Reserve Chair Waller continues to release hawkish signals, it will once again stir global capital markets.

04

Many times, the market works like this: once it falls significantly, various ghost stories suddenly emerge.

However, few people genuinely contemplate whether these ghost stories are truly reliable, or whether they change the market logic.

The current AI sector's plunge is fundamentally due to the extremely rapid gains in the short term, encountering a backlash now.

From last year to mid-year this year, the short span of two years has seen an exaggerated accumulation of gains in the global market, particularly in光 modules, storage, and computing chips.

Not only have SK Hynix, Samsung, Micron, SanDisk, and other storage stocks seen gains of dozens or even hundreds of times, but there are also numerous AI concept stocks in A-shares that have skyrocketed by dozens of times.

However, from January 2025 to mid-2026, A-shares alone will have as many as 2750 stocks whose maximum gain from their lowest point exceeds 100%, accounting for half of the total A-share count.

Out of these, 65 stocks have increased by over 10 times, with the most gaining stocks like Honghe Technology, Shangwei New Material, Dingtai High-tech, and Yuanjie Technology exceeding 30 times, while Hongjing Technology, Tongguan Copper Foil, Nanya New Materials, Changfei Optics, and Zhongji Xuchuang saw increases exceeding 20 times!

If we focus more, examining only AI-related beneficiary concept stocks in the industry chain (components, semiconductors, chips, electronics, optical electronics, computer equipment, communication equipment, etc.), there are over 600 of them, making up more than one-tenth of the total A-shares.

Alone, the market value increase of these global AI sector stocks potentially exceeds 50 trillion USD.

Such reckless market behavior resembles the internet bubble era of 2000.

Therefore, the rapid retraction is also an inevitable fate.

However, we must profoundly and clearly recognize that this round of plummet is essentially just a correction of valuation bubbles and leverage, not an attack on growth logic.

In reality, if you are leading your investments with a long-term approach, you will certainly be much more at ease during this downturn.

Confidence comes from two core bases:

1, global AI narrative certainty and vast market space. Everyone now believes that AI is a new productive foundation for the development of human society, and it is not just the business of a few tech companies but has risen to a technological competition among global nations; it will undoubtedly be the “super main line” of future global economy and technology investments.

From last year, just in terms of enterprises, the global AI overall spending has been soaring at a terrifying speed and scale, breaking through 25 trillion USD this year alone (including the combined capital expenditures of the four major cloud vendors projected to be 670 billion USD in 2026, with a year-on-year increase of nearly 80%), and is expected to maintain a compound growth rate close to30% in the future.

These are solid expenditures, and such a large scale can drive the investment attendance of other companies across the entire industry chain, creating countless huge investment opportunities in the stock market.

As long as this number is not revised downward, the story of AI hardware demand remains unrefuted.

2, investment aspects of the value law. Any valuable investment must have its recognized valuation anchoring level. Taking SK Hynix as an example,

In the first quarter of 2026, SK Hynix's revenue nearly doubled year-on-year, and net profit increased nearly 4 times; market expectations are for the upcoming second-quarter results to reach new highs, with operating profit expected to increase by nearly 600%, with a profit margin maintained above 75%.

Institutions' benchmark expectations suggest that SK Hynix is projected to achieve a year-on-year increase of around 420% to 480% in net profit for the whole year of 2026; entering 2027, even after the base is lifted several times, net profit year-on-year growth is still expected to be 40% to 55%.

The growth rate of performance can be considered frightening.

However, SK Hynix's forward P/E ratio is only 7.8 to 8.5 times, and the forward P/E ratio for 2027 is only 4.6 to 5.3 times.

Regardless of which valuation model is used, it is clear that its valuation is at a historically low level.

If it weren't for the selling pressure brought on by deleveraging, it probably would not have been able to drop to such a valuation level.

Similarly, for AI hardware giants like NVIDIA, AMD, Micron, and Samsung, while the current valuation level cannot be deemed undervalued, their strong growth momentum in the next few years means that the valuations are certainly not overvalued.

Of course, relative to the larger A narrative, the valuation levels are indeed generally much higher than those of these international giants, and still cannot be considered low.

05

In simple terms, the recent deleveraging in South Korea is very similar to our 2015 experience; first, the government encouraged a bull market through the large issuance of leveraged ETFs, then when regulators became concerned about excessive leverage, they tightened measures, leading to a crash.

Currently, the Korean index has retraced 35% from its peak, which corresponds to the A-share market's first round of deleveraging retracement in 2015, and valuations are now quite cheap.

If we reference our 2015 script, whether Korean stocks can regain confidence in the market after stopping the decline will be worth watching.

Investment opportunities emerge from declines and require time.

Objectively, the current global AI revolution is still in its infancy.

Short-term leverage bubbles bursting, geopolitical disturbances, and valuation corrections— these are all routine pains in the growth cycle of the technology industry and cannot stop the wave of industrial transformation driven by AI.

On the contrary, this is precisely the best time to sift out those with genuine core advantages and those likely to grow into great companies.

Once this wave of leverage is cleared out, more opportunities will still emerge.

No need to hurry, let the bullets fly a bit longer.

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。