TSMC has effectively hung a “sold out” sign on its most advanced manufacturing technology. The Taiwanese chipmaker has told investors that every 2-nanometer wafer it plans to produce through 2026 is already spoken for, and that it intends to expand N2 capacity by roughly 70% each year through 2028. The disclosure, drawn from executive commentary during recent earnings calls and investor briefings reported through TSMC’s investor relations portal, underscores just how fierce the race for cutting-edge silicon has become.
For Apple, Nvidia, AMD, Qualcomm, and the handful of other companies that design chips at the bleeding edge, the message is blunt: lock in your allocation now or risk being shut out for years. For everyone else watching the semiconductor industry, it is a signal that the AI-driven surge in chip demand is not slowing down.
What 2nm actually means
TSMC’s N2 process represents a fundamental shift in transistor architecture. The company is moving from FinFET transistors, which have powered every leading-edge chip since 2012, to gate-all-around (GAA) nanosheet transistors. The new design wraps the gate electrode around the channel on all four sides instead of three, giving engineers far more precise control over current flow. TSMC has said N2 will deliver a 10% to 15% speed improvement at the same power, or a 25% to 30% reduction in power consumption at the same speed, compared to its current 3nm process.
Those gains matter enormously for two markets driving the bulk of demand. In smartphones, where battery life is a constant constraint, a 30% power reduction translates directly into longer usage between charges. In data centers, where Nvidia’s AI accelerators and custom chips from cloud giants like Amazon, Google, and Microsoft consume staggering amounts of electricity, even modest efficiency improvements compound across thousands of servers into meaningful reductions in operating cost and heat output.
Volume production of N2 is expected to begin in the second half of 2025, with a full ramp through 2026. Apple is widely expected to be the first major customer, using the node for its A-series and M-series processors, followed closely by Nvidia, AMD, and Qualcomm for AI and mobile chips.
The Kaohsiung buildout
The physical backbone of TSMC’s 2nm expansion is a cluster of new fabrication plants rising in Kaohsiung, Taiwan’s second-largest city. The Kaohsiung City Government has approved land-use plans and coordinated infrastructure commitments for what is shaping up to be a five-fab complex, making it one of the largest concentrated semiconductor investments anywhere in the world.
Kaohsiung sits at the southern end of Taiwan’s semiconductor corridor, roughly 200 miles from TSMC’s existing mega-sites in Hsinchu and Tainan. The city government has been working with national agencies to secure the enormous water, power, and transportation resources that advanced fabs require. A single cutting-edge fabrication plant can consume as much electricity as a small city and use millions of gallons of ultra-pure water daily.
Building five such facilities in one metropolitan area creates clear efficiency advantages. Supply chains, specialized construction crews, and the highly trained workforce needed to operate advanced fabs can be shared across sites, reducing duplication and speeding up ramp timelines. TSMC’s existing experience clustering fabs in Tainan for its 3nm and 5nm nodes provides a proven template.
But concentration also introduces risk. A single disruption to Kaohsiung’s water supply, power grid, or transportation links could knock multiple production lines offline simultaneously. Taiwan’s susceptibility to earthquakes, typhoons, and geopolitical tension with China means that clustering so much critical capacity in one region is a calculated bet. TSMC has invested heavily in seismic reinforcement and backup systems at its existing sites, but the Kaohsiung expansion will test whether those safeguards scale.
The cost of staying at the frontier
TSMC’s capital expenditure plans reflect the staggering cost of building at the 2nm node. The company guided 2025 capital spending to between $38 billion and $42 billion, the vast majority of it directed at advanced process technology, according to its most recent quarterly earnings report. Industry analysts estimate that a single state-of-the-art fab now costs upward of $20 billion when equipment, construction, and qualification are included.
Those costs flow downstream. Wafer prices at the 2nm node are expected to reach $25,000 to $30,000 per wafer or higher, according to estimates from firms like TrendForce and IC Insights. That is roughly double what TSMC charges for mature 7nm wafers. For chip designers, the economics only work if they are shipping products in enormous volumes (smartphones, GPUs) or selling into markets where performance justifies premium pricing (AI training, autonomous vehicles).
This pricing dynamic helps explain why TSMC can credibly claim that 2nm capacity is sold out years in advance. The customer base willing and able to pay for leading-edge wafers is small but voracious. Apple alone accounts for roughly a quarter of TSMC’s revenue. Nvidia’s data-center GPU business has grown so rapidly that securing fab capacity has become a strategic priority discussed at the CEO level. When Jensen Huang and Tim Cook are personally negotiating wafer allocations, smaller chip designers face an uphill battle for access.
The competitive picture
TSMC’s aggressive 2nm ramp does not happen in a vacuum. Samsung Foundry is developing its own 2nm GAA process, with pilot production targeted for 2025 and volume ramp in 2026. Intel, rebranding its foundry ambitions under the Intel Foundry banner, is pushing its 18A node (roughly comparable to 2nm) on a similar timeline. Both competitors have struggled with yield issues at recent nodes, however, and neither has matched TSMC’s track record of delivering high-volume, high-yield production on schedule.
Samsung’s 3nm GAA chips, which began shipping in 2022, faced widely reported yield challenges that limited adoption. Intel’s foundry business has been losing money as it invests in new capacity while trying to win external customers. TSMC’s ability to sell out its 2nm capacity before production even begins reflects the trust gap between it and its rivals. Until Samsung or Intel demonstrate consistent, high-yield manufacturing at the 2nm node, most major chip designers are unlikely to shift meaningful volume away from TSMC.
That dynamic reinforces the supply tightness. Even if TSMC hits its 70% annual growth targets, total industry capacity at the 2nm node will remain constrained if competitors cannot pick up the slack. Companies that assumed they could diversify their foundry sourcing may find that TSMC remains the only viable option for their most advanced designs well into the late 2020s.
Beyond Taiwan: the global capacity question
TSMC’s Kaohsiung expansion is its largest 2nm commitment, but it is not the company’s only geographic bet. TSMC is building advanced fabs in Arizona under the CHIPS Act, with the first facility producing 4nm chips and later phases expected to include more advanced nodes. In Japan, TSMC’s Kumamoto fab (a joint venture with Sony and Denso) is focused on mature and specialty nodes, while a second Japanese facility is planned for more advanced production.
None of these overseas sites are expected to produce 2nm chips in the near term. The Arizona fabs face ongoing challenges with construction timelines and workforce training, and TSMC has been cautious about committing its most advanced technology to locations outside Taiwan. For the foreseeable future, 2nm production will be concentrated on the island, reinforcing both TSMC’s strategic leverage and the geopolitical concerns that have made Taiwan’s semiconductor industry a flashpoint in U.S.-China relations.
What this means for the next three years
The practical reality for the semiconductor industry through at least 2028 is straightforward: 2nm capacity will be scarce, expensive, and controlled almost entirely by TSMC. Companies that have already secured long-term supply agreements are in a strong position. Those that have not are competing for a shrinking pool of uncommitted wafers, and any delay in TSMC’s construction or qualification timeline would tighten supply further.
For investors, the key variables to watch are execution risk and yield. TSMC’s history suggests it will hit its targets, but the jump to GAA transistors is the most significant architectural change in over a decade. If early N2 yields disappoint, the revenue and margin trajectory that Wall Street is pricing in could slip. Conversely, a smooth ramp would cement TSMC’s dominance and validate the enormous capital spending.
The energy and environmental dimension also bears watching. Clustering five advanced fabs in Kaohsiung will place substantial new demands on Taiwan’s power grid at a time when the island is already navigating a complex energy transition. How Taiwan manages that challenge, balancing industrial growth with grid stability, water resources, and emissions commitments, will shape not just TSMC’s 2nm story but the long-term viability of the island as the world’s most important chipmaking hub.
As of mid-2026, the buildout is underway, the customers are lined up, and the technology is on track. The question is no longer whether TSMC can sell its 2nm chips. It is whether it can build them fast enough.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
