The DRAM industry may be approaching one of its most important technology debates in years.
Recently, Micron reaffirmed its commitment to continue scaling traditional 6F² DRAM architectures, while CXMT has publicly discussed plans related to 4F² DRAM development and future commercialization. Meanwhile, Samsung Electronics has indicated that it is also researching 4F²-based technologies and next-generation DRAM architectures.
At first glance, it appears as if the industry is split into two opposing camps.
But is it really?
The Promise of 4F²
The attraction of 4F² DRAM is easy to understand.
Compared with conventional 6F² cells, a 4F² architecture offers:
- Higher density
- Smaller cell area
- Better long-term scaling potential
As traditional DRAM scaling becomes increasingly difficult, 4F² is often viewed as a stepping stone toward future 3D DRAM architectures.
From a roadmap perspective, it looks inevitable.
The problem is that roadmaps are not factories.
Technology Is One Thing, Manufacturing Is Another
A successful 4F² implementation requires more than simply shrinking a cell.
It typically involves:
- Vertical channel transistors
- New capacitor structures
- New integration schemes
- Significant process changes
- New manufacturing learning cycles
All of these factors increase complexity.
And in semiconductors, complexity usually translates into cost and yield risk.
As engineers know, a technically elegant solution does not always become the most profitable solution.
Why 6F² May Still Be the Right Answer
This is why Micron’s position is interesting.
Instead of rushing toward a complete architectural transition, the company appears to believe there is still substantial value in extending and optimizing the existing 6F² platform.
From a business perspective, this argument makes sense.
The current 6F² ecosystem has:
- Mature manufacturing processes
- Known yield characteristics
- Existing equipment infrastructure
- Lower execution risk
If meaningful density improvements can still be extracted from 6F², the economics may remain more attractive than introducing a completely new architecture.
What About Samsung?
Samsung may ultimately pursue both directions.
Historically, leading semiconductor companies rarely bet everything on a single path.
Instead, they tend to:
- Continue scaling current production technologies
- Simultaneously invest in next-generation architectures
This approach preserves near-term profitability while maintaining long-term optionality.
In other words, 6F² and 4F² are not necessarily mutually exclusive.
What Does This Mean for Investors?
Many investors immediately assume that a 4F² announcement represents a threat to Samsung, Micron, or SK hynix.
I disagree.
Stock markets tend to discount events that can meaningfully impact earnings within the next 12 to 36 months.
The commercial success or failure of 4F² DRAM is unlikely to materially affect industry profitability within that timeframe.
The questions investors care about today are:
- HBM market share
- AI infrastructure spending
- Memory pricing
- Supply-demand balance
Those factors will have a far greater impact on stock prices over the next few years than whether a company chooses a 6F² or 4F² roadmap.
A Lesson from the Past
This discussion reminds me of conversations I heard more than twenty years ago when FinFET and GAA technologies were first being explored.
Many people argued:
“Why work on exotic structures when we need higher yield today?”
At the time, that was a reasonable question.
But today’s leading-edge logic industry is built on concepts that once seemed impractical or uneconomic.
The challenge for technology companies is determining which future technologies deserve investment before the economics become obvious.
Conclusion
I believe that during the next two to three years, continuing to extract value from 6F² scaling remains the economically rational approach.
4F² may eventually become essential.
It may even define the future of DRAM.
But future importance and current profitability are not the same thing.
For now, Micron’s commitment to 6F² and CXMT’s push toward 4F² should be viewed less as a competition and more as two different answers to the same question:
How do we continue scaling DRAM after conventional scaling begins to run out of room?
The stock market is unlikely to determine the winner today.
That answer will emerge over the next decade, not the next quarter.