thenextweb.com via Reddit

TSMC Eyes Price Hikes as AI Chip Shortage Persists

tsmc chips ai infrastructure chips ai-infrastructure supply-chain

Key insights

  • TSMC CEO CC Wei said he would 'like' to raise prices, citing persistent AI chip shortages expected to last years with fabs at full capacity.
  • TrendForce reported 15% hikes on 3nm wafers are under consideration for 2H 2026, with 5-10% additional increases projected for 2027.
  • TSMC reported $35.9 billion in Q1 2026 revenue, up 41% year-on-year, with advanced nodes representing 74% of wafer revenue.

Why this matters

TSMC's 90%-plus grip on advanced chip manufacturing means its pricing decisions act as an industry-wide cost increase for every AI accelerator produced by Nvidia, AMD, and Apple. A 15% increase on 3nm wafers in 2H 2026, followed by further hikes through 2029, compresses margins across the AI hardware supply chain with no credible alternative supplier at scale to provide pricing competition. For founders and technical leaders running GPU-intensive workloads, the cost floor for AI compute infrastructure is moving structurally higher.

Summary

TSMC CEO CC Wei told shareholders in Hsinchu that he would 'like' to raise prices, citing AI chip shortages he expects to last years. CFO Wendell Huang confirmed rising costs to the BBC, ruling out sudden large hikes. TrendForce reported 15% price increases on 3nm wafers for 2H 2026, with 5-10% more in 2027. Essentially: TSMC sets the price floor for Nvidia, AMD, and Apple via its 90%-plus grip on advanced nodes. - TSMC posted $35.9 billion in Q1 2026 revenue, up 41% year-on-year, with advanced nodes at 74% of wafer revenue. - Fabs are at full capacity with AI demand still unmet. - Earlier projections cover 3-10% hikes on sub-3nm chips through 2029. TSMC controls over 90% of advanced chip manufacturing, meaning price moves flow into the cost of every AI accelerator on the market.

Potential risks and opportunities

Risks

  • Nvidia, AMD, and Apple face margin compression if 15% 3nm wafer hikes in 2H 2026 cannot be fully passed through to data center and consumer customers
  • Cloud providers and AI labs building infrastructure around TSMC advanced nodes face rising CapEx with no credible alternative foundry controlling comparable advanced manufacturing at scale
  • Compounded price increases across 2026 and 2027 create a structurally higher and persistent cost floor for AI accelerator production, with no near-term relief from competing fabs

Opportunities

  • Samsung Foundry and Intel Foundry Services gain pricing leverage as TSMC hikes make alternative advanced node options more attractive to cost-sensitive chip designers at Nvidia, AMD, and Apple
  • AI inference efficiency vendors and custom silicon teams at cloud providers see improved ROI arguments as TSMC wafer costs rise, making optimization investments easier to justify
  • Investors and chipmakers with older-node exposure (5nm, 7nm) benefit as relative cost advantages widen against TSMC's most advanced process pricing trajectory through 2029

What we don't know yet

  • Whether Nvidia, AMD, or Apple have disclosed how projected 15% 3nm wafer hikes in 2H 2026 will affect their product pricing or gross margins
  • The per-year breakdown for sub-3nm hikes through 2029: the article cites a 3-10% range but provides no per-node or per-quarter schedule
  • Whether TSMC's overseas fabs in the United States, Germany, and Japan carry separate pricing trajectories from Taiwan production or follow the same increases