To The Moon

The memory trade just got a new risk to consider.

On Monday, $SanDisk(SNDK)$ fell 5.97%, $Lumentum(LITE)$ dropped 9.78%, $Micron Technology(MU)$ lost 3.88%, and $SK Hynix(000660)$ declined 4.02%.

The trigger? China’s ChangXin Memory Technologies (CXMT) announced progress on a new DRAM architecture that could increase memory density and improve manufacturing efficiency. The company plans to introduce a server-memory product using its 4F² design by the end of 2026. 

But I think there are two separate stories here.

🔍 1. The threat is pricing power, not just market share

The 4F² architecture aims to fit more memory cells into a smaller area than the conventional 6F² design.

If CXMT can successfully scale this technology, it could eventually become a more serious competitor in server DRAM, putting pressure on established suppliers such as Micron and SK Hynix.

However, announcing a new architecture is not the same as producing reliable chips at scale. Manufacturing yields, performance, customer qualification and production capacity will all matter.

That distinction is important. The competitive threat is real, but its commercial impact remains uncertain.

💰 2. AI demand doesn’t guarantee rising profits forever

AI data centres need enormous amounts of memory. High-bandwidth memory (HBM), conventional DRAM and enterprise storage all play different roles in the infrastructure buildout.

Strong demand can support pricing and profitability, but semiconductor memory remains a cyclical business.

If supply expands faster than demand, prices and margins can come under pressure even while the industry continues to grow.

The market may therefore be reassessing not whether AI needs memory, but how long today’s favourable pricing conditions can last.

⚡ 3. Why are optical stocks falling too?

Lumentum’s decline highlights another concern.

Optical networking companies depend heavily on investment in AI data centres. Broadcom CEO Hock Tan recently warned that access to capital could become a major constraint on AI infrastructure expansion.

That introduces a separate risk: even if demand for AI computing remains strong, financing costs and spending decisions could affect the pace of new projects. 

In other words, the entire AI supply chain isn’t exposed to the same risks. Memory manufacturers face competition and pricing pressure, while optical suppliers also depend on continued infrastructure investment.

📊 My watchlist

🔹 $Micron Technology(MU)$ — Can it maintain pricing power and profitability as Chinese competition develops?

🔹 $SK Hynix(000660)$ — Can its position in advanced memory and HBM help differentiate it from competitors?

🔹 $SanDisk(SNDK)$ — Will demand for enterprise and AI-related storage continue to support growth?

🔹 $Lumentum(LITE)$ — Can optical networking demand justify continued investment in AI infrastructure?

These companies operate in different parts of the supply chain, so I wouldn’t treat their recent declines as identical buying opportunities.

🎯 My conclusion

I wouldn’t assume CXMT will immediately disrupt the global memory market. The technology still needs to prove itself commercially.

But I also wouldn’t dismiss the announcement simply because existing suppliers currently have strong AI demand.

The bigger investment question is whether today’s valuations assume that high margins and limited competition will continue for longer than they realistically can.

If you were looking at the memory sector today, would you buy the dip in Micron or SK Hynix, wait for more evidence on CXMT’s production capabilities, or look elsewhere in the AI supply chain?

This post shares my personal market observations and investment ideas for discussion purposes only.