TSMC’s Q3 Earnings Mark the Start of 2nm Chip Production

The world’s largest chipmaker is quietly laying the foundation for the next wave of physical AI. TSMC is setting up new fabs in Arizona, Japan, and Germany and anchoring its 2-nanometre future in Taiwan. All this, while balancing geopolitics, capacity, and the unrelenting demand for chips that power everything from data centres to smartphones.

As revenue rises and demand outpaces supply, the company is expanding globally while keeping its most advanced technology rooted at home.

In its Q3 earnings report today, TSMC announced a net profit of NT$452.3 billion (~$14.1 billion) for the quarter ending in September 2025. This marked a 39% year-on-year increase, as sustained AI demand boosted orders for its advanced chips.

Revenue rose to $33.1 billion, surpassing guidance, driven by higher shipments from its 3nm and 5nm nodes, the company said during its earnings call. The company’s current earnings per share rose to NT$17.44 (~$0.57), compared to NT$12.54 (~$0.38) a year earlier.

CEO C C Wei noted during the call that “AI demand continues to be very strong, stronger than we thought three months ago.” The company had projected a mid-40% compound annual growth rate for AI accelerators through 2029, but signalled that the figure could be revised higher in early 2026.

On the other hand, Samsung Electronics, in its Q3 2025 guidance, projected an operating profit of 12.1 trillion won (~$9.08 billion) on 86 trillion won in sales. While Samsung’s foundry business still trails TSMC, the South Korean giant recently landed a $16.5 billion contract to supply chips to Tesla, a significant push into its contract manufacturing arm.

Wei said both front-end and back-end capacity remain “very tight” as TSMC works to increase supply for AI clients. He added that the company is “working very hard to narrow the gap between demand and supply,” with new CoWoS and advanced packaging capacity planned for 2026.

Wei confirmed that TSMC will build two advanced packaging facilities in Arizona to meet rising AI-related demand from US customers. The company is also partnering with a significant outsourced assembly and test firm that has already broken ground on its own packaging site in the state.

The facilities will together support local customers and strengthen TSMC’s role in the US semiconductor supply chain.

Shift Towards ‘Foundry 2.0’

Wei also expanded on the company’s new “Foundry 2.0” model, which integrates front-end wafer production and back-end packaging under one system. Advanced packaging now contributes close to 10% of TSMC’s revenue, reflecting its growing importance in the AI era.

“Our customers now value the whole system performance, not just chip technology,” Wei said, adding that the company’s approach aligns with how the industry is evolving beyond Moore’s Law.

This system-level strategy allows TSMC to provide integrated manufacturing solutions that combine process technology, packaging, and design collaboration. It marks a shift from traditional foundry operations to a more complete, vertically linked model.

What’s New in Manufacturing?

TSMC is close to securing land for a second fab in Arizona, with plans to develop an independent “Giga fab cluster” to support the needs of their leading-edge customers in smartphone AI and HPC applications. In Japan, a second plant in Kumamoto is under construction, and its Dresden facility in Germany is also underway, supported by local and EU incentives.

At the same time, TSMC reaffirmed that its most advanced technologies will remain anchored in Taiwan, where the company is preparing multiple phases of 2-nm facilities at the Hsinchu and Kaohsiung Science Parks. The company said these investments will “further strengthen Taiwan’s role as the hub of our leading-edge innovation.”

The company confirmed that 2nm production (N2) will begin this quarter, with the N2P node and A16 process scheduled for 2026. Together, these nodes are expected to form another “long and enduring node family,” targeting energy-efficient and high-performance computing applications.

“We expect a faster ramp in 2026, fuelled by both smartphone and AI applications,” Wei said.

Confidence Despite Export Controls

Wei said the company remains cautious amid “uncertainties from potential tariff policies.”

He added that TSMC continues to balance customer needs with geopolitical realities. “Our overseas decisions are based on customer demand and the necessary level of government support,” he said, adding that TSMC aims to be “the most efficient and cost-effective manufacturer in every region we operate.”

When asked whether tightening US-China export rules could limit AI-related growth, Wei said TSMC remains confident about long-term demand. “If the Chinese market is not available, AI support will still grow dramatically,” he said.

The company believes that global demand, particularly from US and European cloud and hyperscale customers, will offset any near-term constraints from export controls.

During this earnings call, gross margin improved to 59.5%, supported by cost optimisation and substantial capacity utilisation. “We delivered better-than-expected cost improvement efforts and utilisation,” said chief financial officer Wendell Huang.

AI Infrastructure Growth

The company raised its confidence in the long-term AI trend, saying demand for high-performance chips from both tech firms and sovereign AI initiatives continues to rise. TSMC narrowed its 2025 CapEx forecast to $40-42 billion, focusing on AI-related capacity expansion. The chip maker said its major customers estimate that each gigawatt of AI data centre capacity adds roughly $50 billion in infrastructure spending.

Wei declined to specify TSMC’s revenue share from that figure but noted that “it’s not only one chip; it’s many chips together to form a system.”

He said that whether AI computing demand comes from GPUs or ASICs, both use TSMC’s most advanced nodes, suggesting that the mix does not materially change the company’s growth or margins.

Wei said TSMC’s capital spending strategy remains disciplined even as AI expansion accelerates. “As long as we believe there are business opportunities, we will not hesitate to invest,” he said. He added that AI-related planning requires deeper coordination than previous technology cycles.

TSMC now works “two to three years in advance” with both customers and their clients to forecast capacity needs.

The post TSMC’s Q3 Earnings Mark the Start of 2nm Chip Production appeared first on Analytics India Magazine.

Follow us on Twitter, Facebook
0 0 votes
Article Rating
Subscribe
Notify of
guest
0 comments
Oldest
New Most Voted
Inline Feedbacks
View all comments

Latest stories

You might also like...