TSMC’s C.C. Wei on AI Chip Demand, Pricing Power, and Why Taiwan Keeps Its Edge

A Careful Hand on the Pricing Lever

Wei signaled that TSMC would indeed "like" to raise chip prices, acknowledging that the company is evaluating potential increases to capture more value from its constrained leading-edge capacity . However, the message was carefully calibrated. He explicitly stated the company will refrain from implementing sudden price hikes, preferring to maintain a "stable business" relationship with customers rather than imposing abrupt shocks to the supply chain .

This balanced stance leaves the door open for future increases while reassuring partners like Nvidia, AMD, and Apple that TSMC is not exploiting the shortage for short-term gain. Reports noted that TSMC had considered raising 3nm prices by 15% in late 2026 with further increases possible in 2027, though Wei’s shareholder rhetoric stayed focused on long-term partnership stability .

The High-NA EUV Puzzle: Bought, But Not Deployed

Perhaps the most revealing technical update concerned ASML’s High-NA extreme ultraviolet (EUV) lithography tools. Wei directly confronted rumors that TSMC was falling behind Intel in adopting the next-generation systems. His rebuttal was blunt: TSMC has already purchased the machines, but is deliberately holding them back from mass production .

The reason is pure economics. Each High-NA EUV scanner costs approximately $400 million, and Wei argued that productivity must improve and costs must fall before these tools make financial sense for high-volume manufacturing lines . He framed the cautious approach not as a technological lag but as financial prudence, explicitly rejecting analogies to Intel’s historic delayed EUV adoption that cost it a process leadership position .

This clarification provides crucial context for TSMC’s 2nm roadmap. While advanced nodes are in development, the lithography infrastructure for the most advanced iterations is still in an R&D phase. Wei’s logic is that forcing immature, ultra-expensive tools into production too early would erode margins without delivering proportionate performance gains.

U.S. Expansion: Land Banks, Labor Pains, and Long Timelines

Wei provided a reality check on the much-publicized effort to build a leading-edge manufacturing base in Arizona. TSMC has secured two substantial plots of land, which Wei said should satisfy the company’s physical expansion needs for a decade . However, turning that land into functioning fabs is a slow grind.

Permitting and regulatory hurdles remain ongoing, and Wei did not sugarcoat the near-term limitations . Even when the U.S. fabs reach operational status, he cautioned that domestic production alone will still fall far short of fully satisfying American customers' AI chip demand . The combination of Taiwanese and American capacity together will still not close the demand gap, which is precisely why TSMC is also expanding in Japan .

Reports have separately noted that skilled labor shortages have contributed to delays in the Arizona project, a factor acknowledged in broader coverage of the meeting . While Wei did not belabor the point in his public remarks, the subtext was clear: re-creating Taiwan’s dense semiconductor ecosystem abroad is a generational undertaking, not a quick fix.

Beyond the AI Server Rack: Robots and Autonomous Vehicles

Wei expanded TSMC’s growth narrative beyond the immediate AI training and inference boom. He identified autonomous vehicles as a "huge growth opportunity" and singled out multifunctional robots as a critical future direction for semiconductor demand . As AI models move from digital applications into physical systems, the silicon content per robot or vehicle is expected to explode, creating a long-tail demand curve that extends well beyond current hyperscaler investments .

This aligns with earlier, more dramatic forecasts from Wei, who previously speculated that the humanoid robot market could eventually dwarf the electric vehicle industry tenfold. While the shareholder meeting language was more restrained, the strategic signal remained: TSMC is betting that robotics will be the next multi-decade demand driver after the current AI infrastructure buildout .

Taiwan’s Moat Is Deeper Than It Looks

In a direct rebuttal to narratives of semiconductor supply chain diversification away from Taiwan, Wei expressed strong confidence that Taiwan will keep its competitive edge . His argument rested on three pillars: an unmatched density of supply chain partners, a deep pool of specialized engineering talent, and manufacturing operational efficiency that new sites cannot replicate quickly.

He bluntly dismissed concerns that TSMC is at risk of losing its technology leadership, calling the company’s position in the AI supply chain deeply entrenched . The message to shareholders and geopolitical observers alike was that while TSMC will build fabs wherever customers need them, the center of gravity for advanced node production—including 2nm and beyond—will remain in Taiwan for the foreseeable future .

A Record Year Rewarded

Amid the strategic briefings, Wei took care to highlight the financial results that make the capital spending and R&D bets possible. TSMC delivered a record year in 2025, with revenue and earnings per share reaching all-time highs . The company’s stock more than doubled from NT$950 to NT$2,425 per share, and cash dividends grew over 30 percent . Employee profit sharing was increased for 2025, reflecting the company’s philosophy of sharing the spoils of the AI boom with the workforce that makes it possible .

TSMC heads into the second half of 2026 with strong momentum but a clear-eyed view of the challenges: insatiable demand, expensive tools, and the slow, difficult work of building a global manufacturing footprint without losing the efficiency that made it indispensable in the first place.