"AI Bottleneck Trade" Ignites Upstream Sector—Who’s Raking in the Profits?

牛牛ETF研究所

joined discussion · ·

Live Stream Highlights | Is the AI Dividend Still Alive? CIO of CSOP Asset Management Explores New Opportunities in the Storage Industry!

As July draws to a close, the memory storage sector has begun sounding a counterattack—but remains locked in volatile back-and-forth trading.

Currently, the market’s focal point centers squarely on validation of AI-related capital expenditures, shifts in global supply-demand dynamics, and the latest moves by tech giants.

After the sharp sell-off, has the core investment thesis for the storage sector changed? What risks and contrarian opportunities are hidden in this market volatility?

Given current market conditions, Niuniu specially invited@南方东英资产管理Chief Investment Officer Wang Yi in our live session, offering fellow investors exclusive, in-depth insights! Join us as we cut through the noise, uncover the truth behind recent market moves, and identify strategic opportunities in this turbulent market.

This session was packed with insights—we’ve distilled the key takeaways for you:

[Live Stream Highlights Summary]

Live Stream Topic: Investment Opportunities at the AI Bottleneck and in Asia’s Storage Industry

Guest: @南方东英资产管理Chief Investment Officer Wang Yi

Host: Senior Tech Stock Influencer from Niuniu Circle@桃桃淘金币

1. What is the situation regarding the leverage adjustment of 7709?

Q: Starting August 3, 7709 will be adjusted to 'maximum 2x daily leverage'—what does this mean?

A:The market and investors are highly concerned about this.We issued the announcement because the Securities and Futures Commission of Hong Kong introduced new regulatory rules on July 24, prompting us to update the relevant leveraged and inverse products accordingly.

Only whenUnder extreme market conditions(e.g., when product scale becomes extremely large, making trading volume or size difficult to track), the so-called 'flexible leverage' mechanism would be triggered, as clearly stated in the announcement.

But frankly speaking, our product scale once reached approximately HK$130 billion, and even then we didn’t encounter such bottlenecks. Therefore,We are still tracking twice the daily performance.。

Investors certainly expect consistency from the product—understanding that 'two times' means it remains consistently at two times, without needing to check daily. Of course, we publish the tracking leverage ratio daily on our company website and via HKEX, where you’ll see the investment exposure generally maintained around 200%, 201%, or 199.95%.

In summary, thisis not a significant change to the product’s investment strategy, but rather a supplementary amendment announcement made in response to regulatory guidance.

2. How do you view the recent AI-driven market correction in July?

Q: SK Hynix has corrected more than 50% from its June high to its low—how can we assess the current position of the memory industry cycle?

A:From an industry perspective, there hasn’t actually been much change.

The main shift is at the narrative level—both Wall Street and Asian investors are starting to question whether the logic that cloud service providers’ profits are shifting to hardware suppliers remains sustainable.

Looking at the world’s most profitable companies, Samsung, SK Hynix, Taiwan Semiconductor, and Nvidia—the providers of AI infrastructure—have captured the lion’s share of profits, driven by investments from IDCs (internet data centers).

From a token usage standpoint, demand has grown exponentially, and the overall market remains relatively tight. In our conversations with many industry participants, they continue to complain about insufficient computing power and difficulty securing services. The fundamental industry dynamics haven’t changed significantly this cycle; however, the macro narrative has faced considerable skepticism.

The dominant narrative concern is whether cloud service providers can sustain their spending. Google’s free cash flow has recently tightened, and Oracle’s CDS spreads have widened—raising doubts about the sustainability of capex over the next two to three years.

On the capital side, funds invested in memory stocks in Japan, Korea, and Taiwan are relatively concentrated and highly leveraged. Last week, the well-known U.S. AI fund 'Situation Awareness' announced its closure, with its entire position unwound—a fairly symbolic event.This 50% pullback is rare for the sector. What we’re primarily observing are issues related to capital flows and market momentum rather than signals of a demand inflection point or price decline.

The summer holiday effect also plays a role. For European and U.S. capital, summer is a vulnerable period—many large funds go on vacation, and 'Sell in May, Go Away' is a common practice. Once this period passes, we hope to see more allocation-driven, fundamentals-oriented institutional buyers return to the market.

3. How is the valuation of the memory industry?

Q: Companies like Micron mentioned long-term agreements (LTAs) in their earnings reports, with approximately 20% to 30% of capacity already locked in through such agreements. Could this shift the valuation framework for the memory industry?

A:This is a very interesting question.Currently, the market has not fully priced in these long-term agreements, nor has it truly moved beyond the 'consumer electronics cycle' valuation framework.

However, from a business logic standpoint, this cycle is indeed different: the LTAs now come from IDC (Internet Data Center) procurement, which itself stems from large AI model companies seeking to secure computing power. These compute providers lock in chip supply, and GPU/TPU vendors then place memory orders accordingly. The entire supply chain structure has fundamentally changed. Previously, smartphone makers purchased memory based on how many phones they sold that quarter—even under three-year contracts, orders could be canceled if phones didn’t sell.

But IDC construction is inherently long-term—building a compute center isn’t a one- or two-year plan. Moreover, customer diversification itself serves as risk mitigation: IDC clients are not the same as consumer electronics clients, effectively creating a degree of cycle mismatch that smooths out volatility. That said, there’s a fair concern: this LTA system hasn’t yet been tested during a downturn in IDC demand.

If IDCs cancel memory orders just as aggressively during a downturn, why should memory stocks deserve higher valuations? This is a reasonable market skepticism.

Stepping back, even if the market continues to price memory stocks using traditional cyclical frameworks, company actions will directly impact share price performance.At a price-to-book (PB) ratio of around 1.3x–1.4x, the announcement of a large-scale shareholder return program—such as buybacks or dividends—provides immediate support to the stock price. For example, Micron’s recent announcement of an ~$8 billion buyback plan immediately triggered a strong inflection point in its share price, which is a key development worth watching closely.

4. Capacity expansion status and supply-demand outlook

Q: Memory manufacturers continue to raise their capital expenditures—will the tight supply-demand situation ease by 2027?

A:Memory needs to be segmented by market: HBM primarily serves GPUs/TPUs and has the highest technical barriers; DRAM is experiencing shortages due to production lines shifting to HBM, driving parallel price increases; NAND is benefiting from overall IDC demand. Currently, DRAM prices are even rising faster than HBM long-term contract prices, resulting in a 'price inversion' phenomenon.

Why is it difficult to expand high-end capacity? Capacity expansion isn't just about 'buying lithography machines'—it requires an entire integrated production line: wafers, ASML lithography tools, etchers, cleaning systems, inspection equipment, and materials—none of which can be missing. There are only so many lithography machines available; even though ASML says it will increase capacity by 30% next year, the actual number of additional units won’t be significant.

SK Hynix originally planned to start constructing new production lines only in 2027, and even after completion, those lines would require tuning, testing, and yield ramp-up—it’s not as simple as 'installing machines and starting production.' The tight supply-demand balance for high-end HBM is expected to persist into next year.HighHigh-end memory faces significant technical barriers, so its profitability and scarcity may not decline as quickly as commonly assumed, given persistently strong demand and limited ability to rapidly scale supply.

Nvidia's chip iteration cycle has also changed dramatically—it used to be every two to three years, but now it’s every one to two years. Customers are willing to pay more for newer-generation chips because they deliver lower cost per unit of performance: even if the price rises fourfold, efficiency improves tenfold, making each token cheaper—a highly counterintuitive outcome. This drives continuous upgrades in IDCs, sustaining strong demand for the most advanced memory.

5. How does SK Hynix compare with Samsung Electronics?

Q: The ticker 7709 tracks SK Hynix, and CSOP also offers a 2x leveraged product on Samsung Electronics—how should these two companies be compared?

A:Broadly speaking, SK Hynix has a relatively focused business, while Samsung Electronics operates across a more diversified range of businesses.

SK Hynix is focused on a high-end memory strategy and has deeply integrated into Nvidia's HBM supply chain. Its business is highly concentrated, directly linked to the demand for AI infrastructure and offering strong operational flexibility. Additionally, SK Group now has an ADR—the first Korean company to list an ADR in the U.S.—which provides a certain premium effect.

Samsung hit near-term lows two years ago precisely because it wasn't included on Nvidia’s procurement list—Micron and Hynix were. Samsung’s valuation includes a 'free option' premium: if its other businesses improve, there’s additional upside potential, especially in its Foundry segment. In terms of scale, Samsung’s wafer output exceeds Hynix’s, and Nvidia is actively inviting Samsung to participate in testing next-generation chips, similar to Apple’s dual-supplier strategy.

If you want pure exposure to the AI memory theme, Hynix offers a more direct play; if you also want memory exposure plus optionality from other businesses, Samsung presents another angle.

6. How should leveraged ETF products be used properly?

Q: Following a significant pullback, how should investors properly use a 2x leveraged ETF?

A:Inverse and leveraged single-stock ETFs only became available in Hong Kong after the new regulatory framework was introduced. In principle, all leveraged ETFs areproducts intended more for trading than for buy-and-hold strategies, as they are designed to trackdailyDouble leverage performance. It’s not reasonable to calculate 'double' based on a single month’s cumulative return; it’s primarily intended as a trading instrument.

From an investment perspective, always refer to the underlying stock’s performance—not the prior high of the 2x leveraged product.Because for a leveraged product to recover to its prior high, the underlying stock needs to achieve a very substantial cumulative gain.

After a pullback, investors should take a rational view—for example, assess whether SK Hynix can rebound to KRW 250 or 300 per share. Once the price reaches the underlying stock’s previous high, don’t keep waiting for ticker 7709 to rebound back to its own prior high—many investors confuse these two concepts.

This differs from the underlying stock’s performance due to leverage. During rallies, compounding effects occur (e.g., when SK Hynix rose 250%, the product gained nearly 700%); however, during pullbacks, the 2x leveraged trading products decline much faster than the underlying stock.

7. Other Korea-related ETF Products

Q: What other ETF products are available for investors with different investment objectives?

A:This year, given heightened activity in the Japanese and Korean markets, we’ve launched new Korea-related products in addition to our existing Nikkei leveraged/inverse and 1x offerings:

$CSOP KOSPI 200 ETF (03121.HK)$(KOSPI 200 ETF) is a 1x index ETF tracking Korea's leading large-cap benchmark.Its key constituents include SK Hynix and Samsung, which together account for roughly half the index weight; the remainder includes companies like Samsung Electro-Mechanics, LG, and LG Chem. The ETF has delivered solid year-to-date (YTD) performance.

$CSOP KOSPI 200 Covered Call Active ETF (03537.HK)$KOSPI Covered Call ETF: Recently, whether surging higher or pulling back, the Korean market has exhibited very high volatility. This covered call product generates income by selling call options while still retaining significant upside potential, thereby delivering index enhancement effects. The annualized yield from selling calls is close to 20%–30%, offering highly attractive returns, and its covered call yield is notably higher than in other markets.

If you're interested in participating in the Korean market, we recommend starting with a 1x leveraged product.

8. How far along is deleveraging in the Korean market?

Q: What is the current state of deleveraging in the Korean market?

A:Given recent heightened attention on leverage in Korea, we conducted an in-depth review of the overall leverage levels in the country.

To be honest, Korea’s leverage isn’t as high as many imagine. There are two main aspects to consider:

First, at the broker margin (financing) level, Korea’s overall leverage ratio is actually lower than that of China’s A-share market and even the U.S., and remains relatively modest. From the household credit perspective, leverage hasn’t shown significant increases either. While some investors did use leverage and faced margin calls, many responded by depositing additional funds rather than being forced into liquidation.

Another perspective: influenced by media reports, many assume the recent market decline resulted from rapid leverage buildup followed by abrupt deleveraging in Korea. However, in reality, Korea’s leverage levels weren’t that elevated to begin with.

The primary driver of the recent decline was actually substantial outflows from foreign institutional investors. This wasn’t speculative hot money rushing in and out—it was long-standing foreign capital already present in the Korean market rebalancing portfolios due to Korea’s significant gains this year, leading them to reduce their allocation.

Of course, recently the issue surrounding leveraged ETFs has thrust South Korea's domestic regulators into the spotlight, with political factors also at play—such as opposition parties using it as a point of attack.

We examined whether South Korea has developed any 'forced selling' positions, but we haven’t actually observed the typical 'deleveraging' dynamic playing out in the South Korean market. Although retail investors have indeed sold off significantly lately, we still need deeper analysis to understand exactly who is captured under the statistical definition of 'retail.' Therefore, the situation may differ somewhat from media reports, and the reality isn't as severe as it might appear.

Disclaimer: This content is compiled from a live-stream replay and is provided for reference purposes only. It does not constitute investment advice. The information presented does not recommend any specific industry or indicator. Any mentioned stocks and related commentary reflect solely the personal views of the guest speaker and do not constitute stock recommendations. Past performance of individual stocks does not guarantee future results. The stock market involves risk; please invest cautiously.

246K Views