TSMC just handed its competitors a gift. The world’s dominant semiconductor foundry is slashing 28-nanometer wafer production at one of its key facilities by more than 25%, and the ripple effects are already showing up in stock prices across Asia’s chip sector.
Monthly output at TSMC’s Fab 15A has dropped from roughly 200,000 wafers to 150,000, a reduction that accelerated through early 2026 and became official with the company’s June announcement. The move isn’t a sign of weakness. It’s a deliberate pivot toward the technologies TSMC believes will define the next decade of computing.
Why TSMC is walking away from easy money
Here’s the thing about 28nm chips: they’re workhorses. They power OLED display drivers, Wi-Fi hardware, automotive electronics, and countless consumer devices.
The capacity being freed up at Fab 15A is being redirected in two directions. Some of it goes toward manufacturing silicon interposers, the critical connective layers used in advanced packaging for AI accelerators and high-performance computing chips. The rest supports TSMC’s aggressive push into sub-12nm and 2nm process nodes.
TSMC has been clear that it isn’t abandoning 28nm production entirely. The company maintains capacity at other facilities. But the reduction at Fab 15A is substantial enough to reshape competitive dynamics in the mature-node market.
UMC and VIS step into the gap
United Microelectronics and Vanguard International Semiconductor are the most obvious beneficiaries. Both companies have built their businesses around the exact process nodes TSMC is deprioritizing, and the market noticed immediately.
Following the June 22 announcement, shares of both UMC and VIS surged to their daily price limits.
UMC is particularly well-positioned. The company operates a broad 28nm platform and has been advancing into 22nm process technology, targeting applications like display driver ICs and consumer electronics.
VIS, majority-owned by TSMC itself, occupies a similar niche in mature semiconductor manufacturing. The company focuses on power management, display drivers, and sensor chips, all categories where 28nm and larger process nodes remain the standard.
The bigger picture: AI reshapes the foundry food chain
The explosion in demand for AI training and inference hardware has created an almost insatiable appetite for advanced packaging technologies. Silicon interposers, the components TSMC is now prioritizing at Fab 15A, are essential building blocks for chip-on-wafer-on-substrate (CoWoS) packages used in products like Nvidia’s data center GPUs. Every square meter of clean room space dedicated to interposer production is a square meter that can’t make 28nm logic chips.
This is the trade-off TSMC is making explicitly. The company is betting that the margins and strategic value of enabling AI infrastructure outweigh the steady revenue from mature-node manufacturing.
What this means for investors
For UMC specifically, capturing displaced TSMC orders could meaningfully improve capacity utilization rates. The key metric to watch is whether UMC can convert this windfall into sustained customer relationships rather than one-time fills.
There’s risk here too. TSMC’s reallocation could be partially reversed if AI demand softens or if interposer production yields improve enough to require less dedicated capacity. And TSMC explicitly said it isn’t exiting 28nm. It’s adjusting.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.

1 hour ago
21








English (US) ·