The Gap That Keeps Widening
Samsung’s semiconductor foundry division has been losing ground to TSMC for years, but the depth of that struggle is now quietly reshaping how the entire chip manufacturing industry prices its most advanced work.
Samsung’s Yield Problems Are Not a Rumor
Yield rate – the percentage of functional chips produced from a single wafer – is the foundry business’s core metric. When yield rates are low, customers absorb higher costs per working chip, delivery timelines slip, and product launches get pushed back. Samsung’s 3nm process node has reportedly suffered from yield challenges serious enough to push several major chip designers toward TSMC or at least into serious conversations about switching. The technical execution problems are not catastrophic, but in a business where margins are tight and launch windows matter, even modest yield shortfalls carry outsized commercial consequences.
The most telling signal is customer behavior. Qualcomm, which had previously split orders between Samsung and TSMC, moved a meaningful portion of its advanced chip production back toward TSMC. Google’s custom Tensor chips, initially manufactured at Samsung, have also seen shifting production conversations. These are not small design shops making experimental moves – these are high-volume, multi-year production decisions made by companies that run detailed cost-per-functional-chip analyses before committing. When those companies quietly redirect orders, the market reads it clearly.
Samsung has invested heavily in catching up. Its foundry division has announced plans to push toward 2nm and gate-all-around transistor architecture, and it has been aggressive in courting U.S. government subsidies through the CHIPS Act to fund its Texas fab expansion. The infrastructure spending is real. But semiconductor manufacturing is a discipline where announcements and execution are separated by years of process chemistry, equipment calibration, and iterative wafer testing. Promising a roadmap and delivering consistent, high-yield production at scale are different things entirely.
The competitive pressure from Intel Foundry Services adds another layer of complexity. Intel is also attempting to re-enter the merchant foundry market after years of internal manufacturing focus, which means Samsung is fighting a two-front battle – trying to close the gap with TSMC while also distinguishing itself from a new entrant with significant U.S. political backing. That combination keeps Samsung’s foundry division perpetually reactive rather than setting the terms of competition.
Why TSMC’s Pricing Power Keeps Compounding
TSMC’s position in advanced chip manufacturing is structural, not merely circumstantial. Its dominance at 3nm and now 2nm is built on decades of process learning, equipment partnerships with ASML and Applied Materials, and a manufacturing culture that treats yield optimization as a continuous discipline rather than a pre-launch checklist. That foundation creates pricing leverage that goes beyond simple supply and demand.
When a chipmaker has no credible alternative at the leading edge, price negotiation changes character entirely. TSMC can charge premium rates not because it is being opportunistic but because the cost of switching – re-designing chip architecture, re-qualifying production processes, absorbing yield uncertainty during transition – is genuinely prohibitive for most customers. Apple, Nvidia, AMD, and Qualcomm are all deeply embedded in TSMC’s production ecosystem. Moving a complex chip design to a different foundry is not a vendor switch; it is closer to rebuilding a product from the silicon layer up.
That dependency loop has a compounding effect on TSMC’s pricing trajectory. As more leading-edge volume concentrates at TSMC, the company gains more process learning per unit of time, which improves yields further, which makes it more attractive to the next wave of customers. Samsung’s difficulties accelerate this cycle rather than simply reflecting it. Every major design win that TSMC secures because Samsung’s yield performance is uncertain adds to TSMC’s process learning base and deepens its cost advantage on future nodes. The gap does not stay fixed – it actively grows when one side is stumbling.
TSMC has also raised prices at advanced nodes without triggering significant customer defection. When it increased wafer pricing for leading-edge nodes in recent years, the increases largely held because customers had no other viable option at the same performance level. That ability to maintain pricing discipline under demand pressure is the clearest evidence of true market leverage. A company with real competitive alternatives cannot sustain that kind of pricing authority. TSMC can, and does.
The geopolitical dimension amplifies all of this. The U.S. and European push to diversify chip manufacturing away from Taiwan – the so-called “friend-shoring” strategy – has led to massive subsidy packages for new TSMC fabs in Arizona and in Germany. Rather than weakening TSMC’s position, these subsidies are effectively underwriting its geographic expansion while leaving Samsung’s own U.S. expansion partially dependent on similar government support. TSMC gets to grow its capacity footprint with public funding while retaining its process technology lead. That is not a competitive disadvantage by any measure.
What This Means for Chip Customers
For the chip designers who depend on foundry services – fabless companies like Nvidia, AMD, Apple’s silicon team, and dozens of AI accelerator startups – the concentration of leading-edge capacity at a single supplier creates a structural cost problem that has no near-term fix. When TSMC sets pricing, customers negotiate at the margins but rarely on the fundamentals. The real cost pressure lands downstream: in chip prices, in product margins, and in the timelines companies can promise to their own customers. AI chip demand has made this tension sharper, because the appetite for leading-edge silicon is growing faster than the available production capacity.
The harder question is whether Samsung can stage a meaningful recovery at 2nm or whether the foundry competition at the leading edge effectively becomes a one-supplier market for the next several years. Samsung still holds significant market share in mature nodes and in memory, and its overall semiconductor business remains large and profitable. But in the segment that matters most for the AI and advanced computing era – sub-3nm logic – the gap with TSMC is wide enough that chip designers are now building their long-term product roadmaps around TSMC’s capacity schedule, not Samsung’s.
Frequently Asked Questions
Why is Samsung losing foundry business to TSMC?
Samsung has faced yield rate challenges at advanced process nodes like 3nm, leading major chip designers such as Qualcomm and Google to shift more production toward TSMC, which offers more consistent output.
How do Samsung’s struggles increase TSMC’s pricing power?
When customers have no credible alternative at the leading edge, TSMC can set premium prices because the cost of switching foundries – including chip re-design and process re-qualification – is too high to make switching practical.
