Foundry & Fabrication

Can Intel 18A truly shake TSMC's foundry dominance?

In-depth analysis of Intel's 18A process technology advantages, capacity bottlenecks, and its potential impact on TSMC's global foundry landscape.

Introduction

Intel's latest "18A" (1.8nm) chip design is reshaping the semiconductor industry's perception of the foundry landscape. For a long time, TSMC, as the exclusive manufacturer of global AI chips, has nearly monopolized advanced process nodes. With its Gate-All-Around (GAA) transistors and PowerVia backside power delivery technology, Intel's 18A process shows potential to compete with TSMC's N2 (2nm-class) in technical specifications. However, the massive gap in production volume—Intel's annual output of approximately 3 to 5 million wafers, a fraction of TSMC's 17 million—means Intel cannot shake TSMC's position in the short term. Yet the rapid growth in demand is creating room for a dual-giant landscape.

Technology Roadmap Showdown: GAA and Backside Power Delivery

1. GAA Transistors

One of the core innovations of 18A is the GAA (Gate-All-Around) transistor. Unlike traditional FinFETs, GAA wraps the gate completely around the channel, effectively reducing leakage current and boosting drive current. Intel is the first to mass-produce GAA on 18A, while TSMC will not introduce this technology until the N2 node. TSMC previously insisted on FinFET optimization, believing GAA would only be necessary at smaller nodes. However, the technology roadmap shift brings yield challenges: TSMC's N2 was originally scheduled for mass production in 2025, but GAA adoption may cause delays, whereas Intel's 18A is planned for production in the second half of 2024.

2. Backside Power Delivery (PowerVia vs SPR)

Intel's PowerVia moves power wiring to the backside of the chip, leaving only signal interconnects on the front, thereby alleviating routing congestion and reducing voltage drops. TSMC is concurrently developing a similar solution called SPR for N2. Backside power delivery technology is critical for advanced packaging and multi-die integration, improving energy efficiency. Intel has achieved mass production verification of this technology first, creating a certain first-mover advantage.

Technology Barriers

  • Yield Ramp: The manufacturing complexity of GAA and backside power delivery is extremely high, requiring entirely new etching, deposition, and metrology processes. Intel uses a large number of EUV lithography layers on 18A, putting pressure on the photoresist and mask supply chain.
  • Ecosystem Adaptation: EDA tools must support design rules for the new structures, and IP libraries need re-verification. TSMC's Open Innovation Platform (OIP) ecosystem is extremely mature, while Intel Foundry Services' ecosystem is still under construction.

Industry Chain Analysis

Upstream: Equipment and Materials- Lithography machines: ASML's High-NA EUV is key to GAA mass production. Intel has already purchased the first batch of High-NA equipment, while TSMC is more cautious. - Deposition/Etching: GAA requires atomic layer deposition (ALD) and atomic layer etching (ALE), benefiting Applied Materials (AMAT), Lam Research (LAM), and Tokyo Electron (TEL). - Materials: Backside power delivery requires new dielectric materials and metal interconnects. Wafer suppliers Shin-Etsu and SUMCO, as well as photoresist makers JSR and Shin-Etsu Chemical, face new demands.

Midstream: Manufacturing & Packaging

  • Intel: 18A is in trial production at facilities in Oregon and Ireland, with future expansion planned in Arizona and Ohio. However, capital expenditure is massive, and Intel's foundry business continues to incur losses.
  • TSMC: N2 is planned for production in Hsinchu and Kaohsiung, Taiwan, as well as the Arizona facility in the US. The rate of capacity expansion is constrained by equipment and talent.
  • Advanced Packaging: CoWoS and 3D stacking are critical for AI chips. Intel's Foveros and EMIB technologies are mature, but TSMC's 3D Fabric ecosystem is broader.

Downstream: End Customers

  • AI giants: NVIDIA, AMD, Google, Amazon, etc., have long relied on TSMC. Apple has shifted some orders to Intel, signaling a change in attitude.
  • Apple: Adopting Intel chips accelerates diversification in Mac and server SoCs, reducing dependence on a single supplier.- United States: Intel's domestic manufacturing aligns with the CHIPS Act goals, helping to reduce reliance on Asia. However, labor costs and construction delays remain challenges.
  • Taiwan (China): TSMC faces the risk of order diversion, but thanks to its production capacity and technological maturity, it will still maintain over 70% share for nodes below 3nm.
  • Japan: Rapidus, Sony, and others are trying to catch up, but the gap is huge.
  • South Korea: Samsung's 3nm GAA yield is poor, making it difficult to threaten the two giants in the short term.
  • Europe: Intel's European fab plans (Germany, Ireland) may benefit, but the local supply chain is weak.

Investment Perspective

Intel's stock price surged from $20 in fall 2025 to over $120 (a 500% increase), reflecting market optimism about the dual-supplier landscape. In the long term, if Intel's foundry business becomes profitable by 2027, its valuation will be reassessed; but if the 18A yield is worse than expected, the stock price faces significant downside risk. TSMC's valuation is under pressure from rising capital expenditures and increased competition, but its technological moat remains deep.

Long-term Outlook (3-10 years)

  • 2026-2027: Intel's 18A enters volume production ramp, with annual output possibly increasing to 4-6 million wafers. TSMC's N2 enters production, GAA yield gradually stabilizes. The dual-supplier landscape begins to emerge.
  • 2028-2030: "Design portability" appears for nodes below 3nm, allowing customers to use both foundries simultaneously. Intel may win some orders from NVIDIA.
  • Post-2030: Backside power delivery becomes standard. If Intel continues to invest, it could achieve a 7:3 or even 6:4 market share split with TSMC.

Conclusion

For the first time, Intel's 18A has the technological capability to directly compete with TSMC, but its capacity and ecosystem shortcomings mean that a fundamental disruption of the global foundry landscape will take another 3-5 years. The explosion in AI chip demand provides a market foundation for the dual-giant model, and Apple's orders become a turning point. In the future, the semiconductor supply chain will move from "one superpower and multiple strong players" to "two giants standing side by side," but caution is needed regarding uncertainties from Intel's execution and TSMC's counterattack.

Desk context · semiconreport

semiconreport frames this note through Semicon Report tracks chip design, fabrication, AI compute demand, supply-chain shifts, market cycles, and.... dates, names and status changes still need checking: Source links should be opened before the summary is reused. Chip Industry / Industry brief / Focus explains the local editorial angle.

Source links

  1. https://www.forbes.com/sites/johnwerner/2026/06/27/will-intel-boom-fundamentally-challenge-tsmc/Primary

Related articles

Back to channel