TSMC has decided to accelerate its next major technological leap: a new plant for “1.4 nm” processes at the Taiwan Science Park (Zhongke, Taichung). According to sources from the supply chain quoted by local media, this could reach risky production by the end of 2027 and move into mass production in 2028. The key to this shift lies in a single word that almost everything in semiconductors depends on: yield (performance). The message coming from the industry ecosystem is that yields “have been better than expected,” which is pushing the company to shorten timelines for construction and bidding processes.
The move also comes at a particularly sensitive moment for the industry: demand for chips for AI and HPC continues to strain supply chains, but also puts pressure on a less glamorous, harder-to-scalethe resource: actual manufacturing capacity at leading nodes. In this context, accelerating a 1.4 nm fab isn’t just about chasing the “next number,” but a strategic decision to protect TSMC’s leadership in the segment where value is concentrated: high-performance processors and AI accelerators.
A complex of four factories and a “nation-scale” investment
Zhongke’s plan is no small expansion. Information published in Taiwan describes a project involving four buildings for manufacturing and offices, with an estimated total investment of around 1.5 trillion Taiwanese dollars (using local numerical conventions), and initial revenue potentially exceeding 500 billion Taiwanese dollars.
Meanwhile, the projected schedule for the plant suggests a clear roadmap:
- Late 2027: risk production (pre-volume phase).
- 2028: start of mass production.
In industry jargon, “risk production” doesn’t mean final product in the market, but the period when manufacturing begins with nearly final processes and tools, refining performance and stability before the big ramp-up. Having this timeline in place indicates that TSMC isn’t just planning technology — it’s planning industrial capacity, which is the real bottleneck when demand is high.
The other key fact: 2 nm node is already underway
The push toward 1.4 nm makes more sense in the immediate context. The same Taiwanese press reports that TSMC’s 2 nm technology entered production “as scheduled” in late 2025, reinforcing the narrative that the company is meeting milestones for advanced nodes.
In other words: the leap to 1.4 nm isn’t an isolated venture but the natural next step following 2 nm, which becomes the “baseline” for the most demanding customers.
1.4 nm isn’t just “smaller”: it’s the A14 platform to stay dominant in AI
It’s important to translate nanometer marketing into practical implications. TSMC’s 1.4 nm node is typically associated with its A14 branding, and recent information suggests production will ramp up in 2028 with notable improvements in performance, power consumption, and density over previous generations.
There’s an especially intriguing nuance: some reports indicate that TSMC may not initially base this node on High-NA EUV equipment (the most advanced, expensive, and complex tools to integrate), at least at the start, suggesting a strategic evolution approach focused on industrial scalability.
For AI industry players, this matters for two reasons:
- Capacity and cost: if cutting-edge nodes require extremely limited or difficult-to-deploy tools, ramp-up becomes more complicated.
- Competitive advantage: whoever manufactures first, more, and with higher yield gains time — and in AI, time translates directly to market share.
Geopolitics and supply chain: Taiwan first, but with the big picture in mind
The Taiwanese report adds another dimension: international pressure to diversify geopolitical risk and the importance of the U.S. in planning advanced nodes. It mentions that TSMC’s facilities in Arizona would operate in the range of 2 nm to 1.6 nm processes, while 1.4 nm would be prioritized for production in Taiwan first.
This is a delicate balance: manufacturing abroad helps reduce concentration risk, but the core of technological leadership — where every percentage point of yield is invaluable — often remains where the company controls the entire ecosystem (suppliers, personnel, improvement cycles, logistics, and accumulated know-how).
What’s next: the “factory” as a competitive edge
The overall message is clear: the next phase of leadership won’t be won just through R&D, but by transforming R&D into volume production. AI demands chips, yes, but primarily needs available chips with realistic timelines and reliable supply.
In this cycle of technological change, accelerating a 1.4 nm plant indicates that by 2026, leadership will increasingly resemble a massive civil project, with equipment contracts and rising performance curves — not just an announcement.
Frequently Asked Questions
What does “risky production” mean in semiconductors?
It’s a pre-mass production phase where batches are manufactured with nearly finalized processes, to fine-tune yield, stability, and parameters before scaling volume.
When might 1.4 nm chips reach commercial products?
If the mass production timeline holds at 2028, final product adoption typically follows afterwards, depending on customer needs and design cycles (sometimes months, sometimes over a year).
Why is AI heavily driving these advanced nodes?
Because AI accelerators and HPC chips benefit significantly from improvements in density, power efficiency, and performance per watt, and they compete for wafer availability at the most advanced nodes.
Is the “50% yield” figure or concrete performance data for 1.4 nm confirmed?
Not definitively. The published information mentions “better-than-expected” yields as a reason for speeding up the project, but no specific percentages are provided.
via: Economic Daily News