TSMC widened its lead over Samsung Foundry in the second quarter of 2026. TSMC held 72.5% of the global foundry market, while Samsung Foundry ranked second at 5.9%. The top 10 foundries generated nearly $53.49 billion in combined revenue as demand for AI and high-performance-computing processors expanded the market.
The result is a clear win for TSMC in scale and current advanced-node demand. Samsung is responding with fuller 4nm utilization, additional 2nm projects, reported yield improvements and a more aggressive push for new customers, but those measures have not translated into a comparable share position.
The latest market snapshot
The quarter-over-quarter movement tells the story. TSMC’s share rose from 72.3% in 1Q26 to 72.5% in 2Q26. Samsung Foundry’s share fell from 6.5% to 5.9% during the same period.
| Metric | TSMC | Samsung Foundry |
| 2Q26 market share | 72.5% | 5.9% |
| 1Q26 market share | 72.3% | 6.5% |
| 2Q26 foundry revenue | Nearly $40.2 billion | $3.26 billion |
| Change in market share | +0.2 percentage points | −0.6 percentage points |
The direct gap was 66.6 percentage points in 2Q26, calculated from 72.5% minus 5.9%. In 1Q26, it was 65.8 points, so the gap widened by 0.8 percentage points.
Samsung’s revenue still increased 1.8% from the previous quarter to $3.26 billion. That is the uncomfortable part of the comparison: Samsung grew, but the market around it grew faster.
Why the market is growing without closing the gap
AI server GPUs, custom AI processors, XPUs and advanced smartphone chips are consuming large amounts of leading-edge manufacturing capacity. TSMC’s 5nm and 4nm capacity was fully booked by demand from AI and high-performance-computing customers, while 3nm capacity also remained heavily utilized.
TSMC’s process mix shows where much of that business is concentrated. The 3nm process represented 30% of its 2Q26 wafer revenue, and 5nm represented 33%. Together, those two nodes accounted for 63% of wafer revenue. The 2nm process contributed 3%.
This is why the headline market-share gap does not tell the whole operational story. Foundries compete not only for percentage points, but also for the capacity, yields and production schedules required by chip designers building expensive products around a specific process.
Samsung’s recovery plan
Samsung Foundry’s response has several parts. Its Pyeongtaek 4nm line was reportedly operating near full capacity amid demand associated with Qualcomm and HBM base-die customers. Samsung also reportedly raised SF4 wafer prices by 10–15% for customers in the United States and China and by 5–10% for customers in Taiwan.
Those price figures are reported industry estimates rather than a published global price list. They point to tight demand for selected capacity, not to a reversal of the overall market-share ranking.
Yield is another part of Samsung’s strategy. Kim Dong-won, head of the Research Center at KB Securities, said Samsung’s 4nm yield had exceeded 80% and that its 2nm GAA yield was improving beyond 70%. The figures are an analyst assessment of Samsung’s manufacturing progress, not an independently audited production metric.
Samsung is also pursuing more 2nm projects and expanding sales efforts aimed at hyperscalers developing custom AI accelerators and at fabless semiconductor companies. That approach could help it capture work where customers value a second manufacturing partner, but the current figures do not show a substantial transfer of production from TSMC to Samsung Foundry.
What the numbers mean for chip customers
For a chip designer, a foundry comparison is a capacity and execution decision as much as a market-share contest. TSMC offers the much larger current production base and the dominant share of the measured market. Samsung offers a second leading-edge manufacturing option and is trying to improve utilization, yields and customer coverage.
The practical choice depends on the process node, production schedule and qualification work required for a particular chip. Moving a design between foundries is costly because manufacturing processes are not interchangeable; a customer may need substantial redesign and validation before changing suppliers.
That creates room for a multi-vendor strategy, in which a company uses both TSMC and Samsung Foundry, or assigns older-generation products to a different supplier. It does not automatically mean that major production volumes have already moved.
The foundry race from here
Samsung’s measures address the main pressure points visible in the 2Q26 comparison: more productive 4nm capacity, improving 2nm manufacturing, higher reported yields and a broader sales effort. TSMC, meanwhile, enters the next phase with a commanding share and strong demand for its 3nm, 5nm and 4nm capacity.
Samsung’s expanded outreach to hyperscalers and fabless companies is therefore an attempt to turn a capacity-constrained market into new customer opportunities. The next meaningful test will be whether those projects become enough revenue to change the share gap, which stood at 66.6 percentage points in 2Q26.
