APAC Roundup: 13 March 2026
'27 WFE spend, Corning, Japanese fiber, Mipox, Nittobo, JEM, Metasurface
SemiAnalysis: In 2027, constraint shifts from power to construction and silicon.
It seems we are going to need a lot more wafer fabrication equipment in 2027.
We are at ~$135 billion in WFE spend in CY26.
What are the 2027 WFE spend estimates?
Semi is projecting $156 billion WFE spend in 2027.
GS is at $160 billion
MS at $170 billion+
UBS at $180 billion.
So how do we get to $180 billion+ next year?
HBM requires 3x more wafer capacity—and thus more equipment—than traditional DRAM.
As the industry moves to HBM4 in 2026/2027, this intensity is expected to increase to 4x, and then increase further with HBM4E.
Worsening the problem is the fact that yield drops as specs get more demanding. See Jukan’s tweet on Samsung yield.
HBM content in accelerators also high, including for ASICS. Case in point, this tweet about Meta struggling to secure HBM.
Consequently, memory suppliers must build significantly more fabs and install more equipment than in any previous cycle.
The other thing to bear in mind is that memory capex should be weighted towards shells this year, which perhaps means a larger than usual percentage is weighted towards WFE in 2027.
Despite record capex guide of between $52–$56 billion, TSMC will be unable to add enough capacity to fully meet demand for at least the next two years.
As a consequence, customers are being pushed toward Intel Foundry and Samsung Foundry. This forced diversification requires these competing foundries to also invest heavily in WFE to capture the spillover demand from Tesla, NVIDIA, and others.
Corning: BofA says scale out revenue is 4x FY25 estimate
Nice read through for Sumitomo Electric, Furukawa & Fujikura.