Two Major Headwinds Weaker Than Expected! AI Hardware Stocks Supported as SOX Rises for Second Consecutive Day

According to a report by 'Wall Street Insights,' the market believes that headwinds related to data center construction and the impact of open-source models have been overinterpreted, and the fundamental demand for semiconductors remains solid.

Semiconductor stocks, led by memory chip manufacturers, surged over 5% on Tuesday (21st), with the Philadelphia Semiconductor Index outperforming the broader market for two consecutive trading days.

Bloomberg analysis suggests that while short-covering provided short-term support, a more compelling reason is that neither local opposition to data center construction nor the rapid rise of open-source large models is sufficient to signal the end of the AI capital expenditure cycle.

Recent Asian export data, along with price hikes by firms such as TSMC, indicate that underlying demand in computing, memory, networking, and power remains robust, suggesting that the market’s prior pessimism may have been excessive.

As these two concerns gradually fade from market sentiment, the rebound from last Friday’s lows may be more than just a technical oversold correction.

Data center construction does face local resistance.

According to Morgan Stanley, citing third-party data, approximately $156 billion in data center projects are canceled or delayed in 2025, with another $130 billion affected in Q1 2026.

Morgan Stanley estimates that AI capital expenditure will reach $877 billion by 2026, but still faces downside risks such as grid capacity constraints, construction bans, and stricter electricity and water usage regulations.

However, given AI’s strategic importance, the most likely outcome remains project delays or redesigns, rather than outright cancellations.

On the other hand, construction restrictions could actually boost semiconductor procurement demand at existing data centers.

Operators can increase overall computing capacity within the same building space and power supply by replacing inefficient servers, increasing rack density, and adopting more advanced cooling systems.

Additionally, on-site power generation, battery energy storage, and grid support services offer alternative solutions to bypass transmission bottlenecks.

These constraints objectively shorten the lifespan of older equipment and further drive replacement demand for next-generation chips, memory, and cooling systems.

Meanwhile, the latest wave of open-source large models launched in China has a relatively limited impact on hardware demand.

Take Kimi K3 as an example: reports indicate that each service instance still requires substantial high-bandwidth memory (HBM) capacity and numerous AI accelerators.

While Kimi K3 can complete individual tasks with lower computational resources, its overall memory capacity demand remains considerable.

Bank of America Research also points out that falling API prices in China should not be interpreted as a simultaneous decline in hardware costs.

Lower prices primarily reflect improved model architecture efficiency, China’s advantages in electricity, labor, and land costs, and aggressive market share acquisition strategies—not a systemic collapse in semiconductor hardware costs.

In the longer term, the proliferation of open-source models could even expand the overall semiconductor market.

Closed models centralize hardware in a few cloud computing facilities, while open models can be downloaded and deployed independently by enterprises, governments, and sovereign clouds. Each new deployment node may generate fresh demand for high-bandwidth memory (HBM), DRAM, and NAND flash memory, further driving hardware procurement.

Overall, the rebound of AI capex beneficiary stocks from recent lows appears to be driven by more than just technical oversold corrections.

While corporate earnings expectations may have temporarily peaked, current data suggest that the underlying economic logic of AI infrastructure development is far more resilient than previously feared.

With market positioning already adjusted to relatively bearish levels while fundamentals show no clear deterioration, this provides support for the continuation of the current rebound rally.