US stock market returns from a long weekend: AI hype enters the "second half," these three major bottlenecks hold great opportunities.

July 6 Updates

The long weekend in the US stock market has ended, let's briefly look ahead.

I will focus on the following three things

1. TSMC Earnings Report

If TSMC's earnings report confirms the gradual expansion of Cowos capacity, I will pay close attention to $AMKR. AMKR signed a 10-year cooperation agreement with TSMC on June 16—TSMC will procure advanced packaging and testing services from AMKR, with the cooperation based in Arizona.

Additionally, if TSMC's earnings report and conference call mention expressions such as "CoWoS/advanced packaging remains a bottleneck for AI chips," "advanced packaging capacity continues to expand, with customer demand extending to 2027/2028," and "increased importance of the Arizona plant's backend capability, US packaging ecology, and partners," then as a US advanced packaging/testing partner, AMKR will gain very good narrative flexibility.

Just to mention, today Taiwan's DIGITIMES reported that due to reliance on Qatar and disruptions in the Strait of Hormuz, Taiwan’s chip manufacturers are facing a helium supply shock. This is a red flag.

2. Nvidia FUD

This has already been discussed a lot by others. In summary, demand is still there, but due to being somewhat aggressive, its supply route has become less smooth. If Nvidia shifts from Kyber to Oberon Rubin Racks, the number of systems might increase, leading to greater demand for high-power liquid cooling architectures. This would instead benefit companies like VRT/AEIS.

3. Electricity

Grid access has become one of the core bottlenecks for the expansion of AI data centers (along with an El Niño catalyst). This is somewhat related to VRT, which is why I mention it here (though VRT does not provide power).

This direction essentially has two approaches: one is grid expansion ( $XLU), and the other is self-built power ( $BE fuel cells, gas turbines). A-shares' Jerry Holding has already started to rise; it specializes in aviation modified gas turbines and the fuel cell stock $BE also rose in pre-market. $XLU has little elasticity, so I’m not very interested. Meanwhile, the gas turbine-related concept $GEV is too high, and I have some fear of heights. Currently considering a purchase of $NRG.

The biggest change for NRG is that it has acquired LS Power’s assets: 13GW of natural gas generation + CPower commercial/industrial virtual power plant platform, resulting in a total power generation capacity of about 25GW after completion. These assets are mainly quick-start gas generation, located in the Northeast and Texas, which exactly correspond to areas where data centers and grid pressures are rising, along with capacity value enhancement.

In short, NRG has the ability, through its generation assets, retail power, VPP/demand response, and capacity value, to provide more reliable power supply solutions for large load customers and earn flexibility premiums from power contracts and capacity.

Some information updates:

1. Intel

Intel has confirmed that the yield issue of its 18A advanced process has been resolved, reaching conditions for sustainable large-scale mass production. Currently, production is undertaken by two factories, with a combined monthly capacity of over 30,000 wafers.

Good news for $INTC.

2. Storage

The average selling price of HBM is expected to soar from $1.26 in Q1 2026 to $3.25 in Q4 2027, showing a significant increase. The price of non-HBM DRAM is also expected to continue climbing, with structural demand seemingly showing no signs of peaking.

Storage is king; cherish the opportunities presented by declines in storage stocks.

3. MLCC

Driven by the accelerated upgrades of AI server platforms and growth in custom ASIC chip production, as of the end of June 2026, leading MLCC suppliers Murata Manufacturing, Samsung Electro-Mechanics, and Taiyo Yuden have order shipment ratios of 1.30, 1.31, and 1.25 respectively. Murata Manufacturing’s backlog ratio has reached 1.27, surpassing the peak of 2018, indicating rapid accumulation of backlogs and increased risk of supply shortages.

I previously heard from industry chain experts that the MLCC shortage may last about a year and a half.

Additionally, I will place custom ASIC chips on my key focus list; a significant increase in demand for custom ASIC chips (currently there is no data to validate this, but the narrative has already begun) is expected to drive the narrative and performance of $AVGO $MRVL (second tier) and $ARM (third tier, currently not yet at the narrative bottleneck stage).

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。