● Treasury Yield Shock

Why do memory semiconductors jump first when Treasury yields turn down: a one-shot summary of Nasdaq, AI data centers, and Micron’s long-term contracts

The truly important point in this material is not simply that “the stock market rises in October.”

The core point is that the possibility of a decline in U.S. Treasury yields, pressure for Nasdaq to hit new highs, undervaluation of memory semiconductors, and expanding AI data center investment are all aligning at the same time.

In particular, what the market has not yet fully priced in is Micron’s long-term supply contracts and the earnings leverage of optical communications and network infrastructure in 2027~2028.

In other words, the next AI investment cycle is likely to favor not the big tech companies that spend the money, but the companies that receive it, rather than big tech stock prices.

1. The case for a market rebound after October: why seasonality is getting attention again

According to Citadel’s hedge fund data, in past midterm-election cycles, the market often bottomed in late September and the S&P 500 rebounded from October.

The data indicated a bottom on September 30, followed by a strong average upward trend in the fourth quarter.

• The average fourth-quarter return was cited at about 4.2%, making it the strongest quarter of the year.
• The average return for October in midterm-election years was given as about 3%.
• The average return for November in midterm-election years was cited at about 2.7%.
• Combined, October and November suggest a possible gain of about 5.7%.
• If the December Santa rally is added, a scenario of roughly 10% by year-end was also mentioned.

Of course, seasonality is not an absolute investment rule.

But if market participants share the expectation that a rebound may begin in October, actual capital flows can tilt in that direction.

In particular, with the possibility that individual investor funds could re-enter the market from October, expectations are growing that risk appetite across global equities could revive.

2. The key variable is the U.S. Treasury yield: if the 10-year yield turns down, growth stocks revive

The most important variable in this market interpretation is the U.S. 10-year Treasury yield.

The material explains that the 10-year yield briefly rose to around 5.38%, threatening the highest level in 20 years.

But what mattered most was that the yield later eased and closed around 5.24%.

It is true that yields are still high.

Frankly, yields at this level are a burden for the stock market, especially for growth stocks and technology stocks.

Even so, the Nasdaq and S&P 500 holding near their highs suggests market resilience is quite strong.

More importantly, if upward yield pressure eases from here, the valuation of suppressed technology stocks can reopen.

If the Nasdaq held up even while rates were weighing on stocks, then the moment that rate pressure eases, upside momentum could become even stronger.

3. The bond market has become too one-sided: the possibility of a yield reversal

One particularly interesting part of the material is the positioning in the bond market.

It explains that put-option exposure on TLT, the U.S. long-term Treasury ETF, has risen to its highest level in six months.

In other words, the put-call skew is tilted extremely far in one direction.

Put simply, the market is betting too heavily that “bond prices will keep falling and yields will keep rising.”

But in financial markets, when positioning becomes too concentrated on one side, mean reversion often occurs.

• Treasury yields and bond prices move in opposite directions.
• Higher yields mean bond prices have become too cheap.
• If bonds are oversold, buying can come in and yields may fall instead.
• Short positioning in 10-year Treasuries was also described as near historical highs.

The material explains that, in similar past situations, the 10-year yield tended to fall by about 87bp three months later.

The reason this matters is simple.

If Treasury yields turn down, valuation pressure on technology stocks, semiconductors, and AI infrastructure names eases.

4. The real reason Nasdaq is strong: earnings are stronger than yields

The reason Nasdaq is strong is not just because of liquidity.

The material emphasizes that technology earnings are very strong, and that some companies are posting annual profit growth of more than 40%.

The current forward P/E of the S&P 500 was cited at around 19x.

That may seem high, but considering the earnings growth rate of technology stocks, it is interpreted as a historically manageable range.

To summarize:

• Yields are high.
• Oil prices are unstable.
• War risks remain.
• Even so, Nasdaq and the S&P 500 remain near their highs.
• The reason is that AI and technology earnings are overpowering the burden of rates.

Therefore, if yields fall, this could go beyond a simple relief rally and trigger a valuation re-rating.

At that point, the sector most likely to react first is semiconductors.

5. Why Micron matters: the structure of memory semiconductors is changing

The most important stock-specific news in this material is Micron.

Micron disclosed 10 new long-term supply contracts in its earnings release.

The scale of these contracts was described as equivalent to about 35% of estimated revenue.

Even more important, when existing long-term contracts are included, a substantial portion of future revenue has been secured through long-term agreements.

This means Micron is no longer simply a company that sells DRAM and HBM based only on spot prices.

The traditional memory semiconductor industry was a cyclical industry.

In good times, it could generate huge profits, but in bad times operating profit could collapse or turn negative.

Samsung Electronics, SK hynix, and Micron all struggled to receive high valuations because of this cycle.

But expanding long-term supply contracts changes the picture.

Price volatility falls, revenue visibility rises, and earnings stability improves.

In that case, the market may assign memory companies a higher P/E multiple than before.

6. Memory semiconductor valuation: the argument that they are too cheap

The material explains that the current forward P/Es of memory-related companies are historically low.

• Micron was cited at about 6x.
• SanDisk was cited at about 7x.
• Samsung Electronics and SK hynix were cited at about 4x.
• By contrast, Nvidia was mentioned at about 15x, despite debate over whether it is undervalued.
• The semiconductor sector average P/E was compared with the 20x range.

The conclusion from this comparison is clear.

Memory companies at the center of the AI semiconductor cycle are still being valued as if they were traditional cyclical industries.

But if HBM demand rises structurally, DRAM supply remains tight, and long-term contracts increase, the possibility of the kind of sharp downcycle seen in the past becomes smaller.

In that case, memory semiconductors can be re-rated not as simple cyclical stocks, but as core AI infrastructure assets.

7. Why the idea that 2028 is the semiconductor peak is getting weaker

Previously, many analysts saw 2028 as the peak of the semiconductor cycle.

But the material says that outlook is now changing.

Demand and supply imbalances may not be easily resolved even after 2030.

The reason is the AI agent era.

If AI so far has been centered on chatbots, the next phase is likely to see AI agents that actually perform tasks spread widely.

As AI agents increase, inference workloads will surge, and as inference workloads rise, demand for HBM and DRAM will rise together.

Ultimately, AI data centers need more GPUs, more memory, and faster networks.

In this structure, memory semiconductor demand can become not a temporary fad but a long-term investment cycle.

8. Why a sharp near-term surge is still difficult: supply-demand and sentiment issues

That said, the material does not see Micron or memory stocks surging immediately.

The reason is supply and demand.

It explains that liquidity in the market is not yet abundant enough, and investor sentiment in the Korean market has not fully recovered.

For Micron to rise strongly, Samsung Electronics and SK hynix would also need to rise together.

Because memory semiconductors are a global sector, it is hard for just one company to be re-rated on its own.

That is why the long-term view matters more than short-term trading.

If long-term supply contracts are confirmed in actual results, HBM demand continues, and yields fall as well, then a valuation re-rating for memory companies becomes possible.

9. AI data center investment: where the real money is flowing

Another important data point in the material is data center construction spending.

Data center construction spending was said to have increased by 7.5% in just one month.

By contrast, the office sector increased by only 0.1%.

This gap is very significant.

It shows where capital expenditures are going.

The money is flowing not into traditional offices, but into AI data centers.

This trend leads to increased equipment orders.

As a result, memory, optical communications, network, and power infrastructure stocks can react strongly.

10. Optical communications and networks: the hidden leaders of 2027~2028

The material says the optical communications sector could post its strongest results in 2027 or 2028.

It explains that Goldman Sachs released a report on the optical sector, and that the report pointed to strong growth potential in optical communications.

In the AI era, GPUs are not the only important thing.

Chips must be connected to chips, servers to servers, and data centers to data centers.

What is needed for that is optical communications and network equipment.

The material mentioned Lumentum as a key company in the optical sector.

It explains that Lumentum’s stock rose to around 104.6 dollars, moving close to a new high.

This is still a relatively underappreciated theme in the market.

When most people talk about AI investment, the conversation ends with Nvidia, Broadcom, Microsoft, and Google.

But as data centers expand, bottlenecks arise not only in compute chips but also in connectivity.

That is why optical communications can become a key late-stage beneficiary of the AI infrastructure investment cycle.

11. The defeat of the data center power-cost bill: easing concerns about construction delays

The material also mentioned that a bill in the U.S. Senate that would have made hyperscalers bear data center power costs did not pass.

This news is more important than it may seem.

If large cloud companies had been forced to shoulder a bigger share of data center power-grid costs, that could have slowed data center construction.

But with the bill failing, concerns about delays in data center construction can be seen as easing.

This trend is positive for AI semiconductors, servers, networks, power equipment, and cooling equipment companies.

Ultimately, the likelihood that AI data center construction will continue has increased.

12. Nvidia and Broadcom: even if they take a breather, the structural advantage remains

The material explains that Nvidia rose 1.9%, and AI semiconductors and CPU-related stocks were generally strong.

At the same time, it says Nvidia and Broadcom have been taking a bit of a breather recently.

Even so, the fundamentals and competitive advantages of both companies remain strong.

Nvidia has the GPU ecosystem and the CUDA platform, while Broadcom has a powerful position in AI ASICs and network semiconductors.

From a stock-price perspective, however, infrastructure supply-chain companies may be more favorable in the short term than big tech.

That is because big tech is the side that has to spend enormous amounts on data centers, while infrastructure companies are the side that receives that money as revenue.

13. Accenture’s rebound and the AI replacement debate: the idea that “AI cannot take jobs” is too simplistic

The material also mentioned the share-price rebound of IT consulting company Accenture.

There were concerns that AI would replace IT services work, but Accenture’s results came in better than expected, and the broader software sector rebounded as a result.

Some interpret this to mean that “the idea that AI replaces people is exaggerated.”

But the material takes a different view.

The AI replacement thesis is not just hype.

In fact, the pace of AI advancement is very fast.

What did not work yesterday works today, and tasks that were impossible a month ago are now automated.

One to two years from now, automation of work at a level that is hard to imagine today may be possible.

Therefore, just because Accenture rose 15% in one day does not mean concerns about AI replacement have disappeared.

On the contrary, it is entirely possible that many S&P 500 companies could face business-model pressure from AI.

14. Why the “shovels and jeans” strategy is more favorable than big tech

The most important conclusion from the investment perspective in this material is this.

Who the winners of the AI era will be is still not clear.

But who is making money is relatively clear.

It is the companies supplying the equipment and infrastructure needed to build AI data centers.

• Memory semiconductor companies
• HBM suppliers
• Network semiconductor companies
• Optical communications companies
• Power infrastructure companies
• Cooling equipment companies
• Server and data center equipment companies

This is the so-called shovels and jeans strategy from the gold rush era.

Big tech companies operating AI services may be excellent businesses in the long run.

But in the short term, they have to bear enormous data center investment costs.

By contrast, infrastructure supply companies receive that spending as revenue.

So from 2026 through early 2027, memory, networks, and power infrastructure may have an advantage in stock-price terms over big tech.

15. Two years to complete a data center: the real earnings impact may only become visible after 2028

Data centers take time to build.

The material explains that data center construction has roughly a two-year lag.

If data centers are being built on a large scale now, actual operations and monetization are likely to accelerate after 2028.

Because stock prices usually move before earnings, related expectations could be reflected more strongly starting in 2027.

From this perspective, the market is now in the middle phase of the AI investment cycle.

In the early stage, Nvidia GPUs led the way; in the next stage, HBM and memory semiconductors are gaining attention.

In the stage after that, optical communications, networks, and power infrastructure could come into sharper focus.

The most important point that other news often misses

First, a decline in rates is not just a positive catalyst, but a trigger for valuation re-rating.

What matters is that Nasdaq has already held up despite high rates.

If yields turn down, the P/E multiples of growth stocks and semiconductors that had been compressed could expand again.

Second, Micron’s long-term supply contracts could change the nature of the memory industry.

If memory semiconductors no longer rely entirely on spot-price cycles, the market could grant higher valuations to Samsung Electronics, SK hynix, and Micron.

Third, the bottleneck in AI data centers is not only GPUs.

Optical communications and networks that connect chips to chips, servers to servers, and data centers to data centers may become the next bottleneck.

This area still receives less public attention, but earnings momentum in 2027~2028 could be very strong.

Fourth, the fact that big tech is a good business is not the same as its stock price rising immediately.

Big tech has to spend massive amounts on AI data centers.

By contrast, memory, network, and power infrastructure companies receive that money.

Right now, it is more important to look closely at the AI infrastructure supply chain than at AI service companies.

Fifth, concluding that AI cannot replace jobs is too premature.

Accenture’s rebound is only a short-term earnings relief, and it is hard to treat it as evidence against the long-term shock of AI automation.

Given the speed of AI advancement, a substantial portion of enterprise services, consulting, and software work is likely to keep being reshaped.

Key checklist from an investment perspective

• Check whether the U.S. 10-year Treasury yield fails to rise further around the 5.3% level and turns down.
• Watch whether the concentration in TLT put options and the short position in 10-year Treasuries unwind.
• It is important to see whether the Nasdaq and S&P 500 break to new highs together with falling yields.
• Check whether Micron’s long-term supply contracts translate into actual revenue stability.
• Watch whether Samsung Electronics and SK hynix receive a valuation re-rating together with Micron.
• Track whether AI data center construction spending keeps rising.
• Check whether orders and earnings guidance improve for optical communications, network, and power infrastructure companies.
• Also important is how much of big tech’s data center CAPEX burden is already reflected in stock prices.

< Summary >

The core point of this material is that if U.S. Treasury yields turn down from high levels, Nasdaq and semiconductors could react strongly.

The bond market has become too crowded with short positions and put-option bets, increasing the possibility of a downward reversal in yields.

Micron’s long-term supply contracts are an important change that could re-rate memory semiconductors from a traditional cyclical industry into a core AI infrastructure industry.

As AI data center construction spending rises rapidly, memory, optical communications, networks, and power infrastructure are likely to receive real benefits.

Big tech companies are good businesses, but for now they face huge CAPEX burdens, and attention should be focused more on the AI infrastructure supply chain that receives the money rather than the companies that spend it.

[Related Articles…]

• AI Data Center Spending and the Next Growth Wave
• Memory Chips and the Semiconductors Re-rating Story

*Source: [ 월텍남 – 월스트리트 테크남 ]

– 국채금리 꺾이고 메모리위주 폭등장 온다는 자료 4가지

---