Samsung Foundry Recovery Accelerates, but 2nm Yield and Texas Fab Ramp Remain Key to Profitability
nashnova research
Samsung's foundry utilization has hit a multi-year high, with select advanced-node orders repriced up to 15% — but whether this recovery converts into lasting profit hinges on 2nm yield and the ramp schedule at its new Texas fab.
What is driving the foundry recovery?
Samsung's Pyeongtaek 4nm line has run at full capacity since late 2025. Starting July, select new advanced-node orders carry a ~15% price increase.
Utilization across 8nm and below has reached its highest level in years, led by rising orders from U.S. and Chinese clients.
In the second half, Samsung plans to begin volume production of second-generation 2nm mobile chips while expanding sales of 4nm LPU and HBM base die — the logic chips that sit beneath high-bandwidth memory stacks. The revenue target: double-digit growth.
Market share is still slipping — so what actually improved?
TrendForce data: Samsung foundry revenue rose 1.8% quarter-on-quarter to $3.26 billion in Q2, yet global share actually fell to 5.9%.
TSMC held 72.5% share in the same quarter, with revenue of roughly $40.2 billion — the gap is still widening.
This means → Samsung's operations are getting better, but its competitive standing is not keeping pace. In plain terms = it is improving, but the leader is pulling away faster. Whether higher utilization can turn into sustainable margin remains the central open question.
Why is 2nm yield the make-or-break variable?
Foundry cost structure is unusual: upfront capex converts into depreciation — a fixed cost. Unit costs only come down when each wafer yields enough sellable chips.
In plain terms = even a line running at full load loses money if yield is too low — produce 100 chips but only 60 pass, and the per-unit cost stays high.
Samsung is advancing second-generation 2nm production, building on first-gen experience, and the 2nm customer pipeline now extends beyond mobile into high-performance computing. This means → whether 2nm yield hits target will directly determine if "higher utilization" actually translates into profit.
What extra pressure does the Texas Taylor fab add?
The Taylor fab is expected to come online by late 2026. It uses EUV lithography — etching chip circuits with extreme-ultraviolet light — and is designed to produce advanced chips down to 2nm.
Tesla has signed a deal worth roughly $16.5 billion to manufacture its next-generation AI6 chips at Taylor — an anchor order for the facility.
But a new fab's early phase brings heavy depreciation and operating costs. This means → until utilization climbs high enough to cover those fixed costs, Taylor will drag on overall profitability. How fast Samsung can stabilize production, clear customer qualification, and fill capacity will directly set the timeline for the foundry business to break even.
How does the new bonus structure make breakeven harder?
Foundry chief Han Jin-man said in June that full-year profitability in 2027 remains difficult; 2028 is more likely.
One reason: a new special management performance bonus introduced in May. It draws from 10.5% of the Device Solutions (DS) division's operating profit and is subject to a profit threshold.
This reflects a structural mismatch: the bonus is tied to the entire semiconductor division, not the foundry alone. Strong memory profits can trigger bonus payouts even while the foundry is still loss-making. In plain terms = profits earned on the memory side get partly redirected through the bonus mechanism, making it harder for foundry's own books to turn positive. 2nm yield and the Taylor ramp pace are the key checkpoints for judging whether breakeven can come any sooner.
市场有风险，内容仅供研究参考，不构成投资建议。