Daeduck, Simmtech and TLB Profit Jumped Together as AI Substrate Orders Find New Capacity

An AI server is not complete when a GPU is plugged in. Substrates connect the GPU and CPU to the main board, while multilayer printed circuit boards carry signals and power across the server. That makes board suppliers part of the AI-server buildout, not just GPU makers. Jukan (@jukan05) summarized Korean press reports on August 8. Daeduck Electronics posted second-quarter revenue of KRW 401.0 billion and operating profit of KRW 70.3 billion, up 63% and 3,599% year over year. Simmtech reported revenue of KRW 514.6 billion and operating profit of KRW 62.9 billion, up 51% and 1,035%. TLB reported revenue of KRW 88.2 billion and operating profit of KRW 12.8 billion, up 38% and 86%. The Daeduck figures also appear in a filing summary.

The three companies make different parts of the same supply chain. Daeduck makes FC-BGA and multilayer boards that connect high-performance chips to main boards. Simmtech makes package substrates and memory-module PCBs, while TLB makes PCBs for server DDR5 and enterprise SSDs.

Why substrate orders are moving now

Daeduck said in July that it would invest KRW 710 billion in semiconductor facilities this year. It expects high-value substrates for powerful chips to remain short and is also expanding the multilayer boards used in AI accelerators.

Daeduck Electronics to invest KRW 710 billion in semiconductor substrate facilities this year (Seoul Economic Daily, 2026-07-21)

출처: en.sedaily.com

Jukan argues that some memory and system-semiconductor substrate orders could move to Daeduck and Simmtech as large Taiwanese suppliers direct more capacity toward high-value ABF substrates. This is a possible shift, not a confirmed order.

Why Vera Rubin and SOCAMM enter the story

The timing point is Nvidia's next AI platform, Vera Rubin. Nvidia said on May 31 that it was ramping into full production and that production shipments would begin this fall.

Nvidia says Vera Rubin is ramping into full production, with production shipments starting this fall (NVIDIA, 2026-05-31)

출처: investor.nvidia.com

Vera Rubin is a server platform combining the Vera CPU with the Rubin GPU. It uses a small, low-power memory card called SOCAMM beside the CPU to keep the server's data close to the processor.

Jukan expects Simmtech and TLB substrate orders to benefit because both supply related SOCAMM products. But Nvidia's schedule does not automatically become Korean revenue: customer orders, production yields and delivery dates must line up.

What to watch in the third quarter

The third quarter will show whether this is a one-quarter rebound or a durable demand signal. Watch three things: whether all three companies hold their second-quarter profit, whether output of high-value substrates and SOCAMM products rises, and whether Vera Rubin shipments start on schedule and become supplier revenue. If Daeduck, Simmtech and TLB keep improving, the AI buildout is reaching the less visible layer that connects chips. If orders rise but capacity or delivery does not, the second-quarter results may have been a temporary peak created by a supply-chain bottleneck.