TSMC’s Fab 25 1.4nm Ramp Could Move to 2027

TSMC’s next leading-edge manufacturing transition may be arriving sooner than expected. Reporting from Taiwan indicates that the company’s Fab 25 complex in Taichung could move from an earlier plan of late-2027 risk production and second-half 2028 volume manufacturing to pilot output in April 2027 and mass production as early as the second half of 2027. The key word is could: the accelerated dates are reported targets, not a guarantee of commercial output. 1213

What is changing at Fab 25?

Fab 25 is TSMC’s planned four-fab 1.4nm cluster at the Phase 2 expansion of the Central Taiwan Science Park in Taichung. The first two buildings are moving fastest:

• P1: The steel structure has been completed and floor-slab work is under way. It is expected to be completed in April 2027, after which pilot production is expected to begin. 12
• P2: Foundation work is under way, with completion reportedly scheduled for October 2027. 2
• Overall pace: Local park officials have described the accelerated P1/P2 construction as roughly six months ahead of the original schedule. 1
• P3 and P4: These remain later stages of the four-fab plan. As of late August, P3 had obtained its permit while P4’s application was still being processed, indicating that neither was as far along as P1 and P2. 55

Earlier reporting had placed risk production near the end of 2027 and full-scale manufacturing in the second half of 2028. The newer reports suggest P1 might begin volume production in H2 2027, materially compressing that transition. 213

Why the earlier 1.4nm date matters

Leading-edge fabs are strategic assets because capacity—not just process design—determines which chip designers can ship the most advanced processors at scale. A successful 2027 ramp would give TSMC an earlier path to serve future AI accelerators, high-performance computing chips, CPUs and mobile processors.

The project is also unusually large. Earlier reports put the planned investment for the four-fab Taichung site at about NT$1.5 trillion, or roughly US$49 billion. A Taichung city estimate cited in reporting projected annual output of up to NT$485.7 billion once the facility is operating at scale. These are reported project estimates, rather than a detailed TSMC capital-spending breakdown. 52

A four-fab cluster gives TSMC room to expand capacity over multiple generations. That matters because an early first-fab ramp alone does not guarantee long-term supply leadership; sustained leadership requires enough capacity to support customer demand after the initial launch.

The competitive comparison: Intel and Samsung

An H2 2027 mass-production start would change the apparent timetable for the 1.4nm-class generation.

Company	Reported 1.4nm-class timeline	What it would mean if TSMC reaches H2 2027
TSMC	Fab 25 pilot production targeted for April 2027; reports say mass production could begin in H2 2027. 12	Would move ahead of TSMC’s earlier H2 2028 volume expectation.
Intel	14A risk production is planned for H2 2027, with high-volume manufacturing targeted for 2028. 33	TSMC could reach volume production before Intel’s planned high-volume ramp.
Samsung	Samsung has targeted 2029 mass production for SF1.4, after moving back its prior 2027 target. 3840	TSMC would be roughly two years ahead on announced volume-production schedules.

These labels should not be treated as directly equivalent technical measurements. “1.4nm,” “14A” and “SF1.4” are process-family names, and each company uses different transistor, density, power and performance metrics. The clearer comparison is the timing of qualified, high-volume manufacturing—not the node name alone.

AI demand is giving TSMC a stronger base for expansion

The earlier Fab 25 target comes as demand for advanced chip manufacturing remains tight. TrendForce reported that TSMC held 72.5% of pure-play foundry revenue in the second quarter of 2026, with revenue near US$40.2 billion, up 12.1% quarter over quarter. It attributed the momentum in part to AI-server GPU and XPU demand, which kept TSMC’s 5nm/4nm and 3nm capacity fully booked. 30

That utilization matters for two reasons:

1. Customers need a future capacity path. AI chips and other high-performance processors compete for the most advanced wafer capacity. Bringing 1.4nm online sooner offers prospective customers an earlier next-generation option.
2. High utilization supports continued capital investment. Strong demand at current advanced nodes helps justify the cost and execution risk of building the next generation of fabs.

Counterpoint Research separately measured TSMC’s Q2 2026 pure-play foundry share at 73%, rather than 72.5%. The small difference reflects methodology and rounding, while both measurements show TSMC holding a commanding market position. 2030

What to watch before calling the ramp a certainty

Construction progress is meaningful, but finishing a fab building is only one step in semiconductor manufacturing. The reported H2 2027 mass-production date still depends on timely equipment installation, process qualification, yield improvement and customer approval.

There is also a difference between pilot or risk production and a mature volume ramp. Earlier July reporting, for example, anticipated pilot production around the start of Q3 2027 and mass production closer to mid-2028 if installation and qualification proceeded smoothly. 8 The more recent H2 2027 reporting therefore represents an aggressive upside scenario rather than a settled industry fact. 12

Bottom line

Fab 25’s accelerated construction is strategically important because it could turn TSMC’s 1.4nm node from a 2028 volume story into a 2027 one. P1 is the immediate focal point, P2 is following behind, and P3/P4 remain earlier in the development pipeline. 1255

If TSMC delivers H2 2027 volume production, it would strengthen an advantage already supported by tight AI-driven demand and a 72.5% share of the pure-play foundry market. But the competitive conclusion remains conditional: a fab’s building schedule must still translate into reliable yields and qualified customer output before an earlier date becomes a durable manufacturing lead. 30