TSMC's Unmet AI Orders Are Reshaping the Global Chip Landscape
Hello everyone, I'm your financial journalist. Today, we're going to discuss a very complex but profoundly influential topic: Amid the AI boom, TSMC, the global leader in chip manufacturing, is struggling to keep up with the demand. So, where are all those orders going that it can't handle? And what does this mean for understanding the future landscape of technological competition?
To put it simply, the current situation is this: The demand for AI chips is surging like a flood. Although TSMC is a powerhouse, its production capacity has reached its limits. As a result, orders are beginning to spill over to other manufacturers such as Samsung, Intel, and ASE (Asia Semiconductor Engineering).
Let me break down this "spillover effect" into five key points to help you understand the logic behind it.
---
1. The Core Conflict: Surging AI Demand Hits TSMC's Production Cap
First, we need to understand why this spillover is happening. TSMC's CEO, Mark Liu, put it quite plainly: Both the capacity for manufacturing the core parts of chips and for packaging them are under intense pressure. AI chips, such as NVIDIA's GPUs and Apple's new chips, require extremely advanced manufacturing processes and are in high demand. TSMC predicts that its revenue will grow at a compound annual rate of over 45% in the next five years, meaning the orders are overwhelming.
In plain language:
It's like a Michelin-starred restaurant with the best chefs and the freshest ingredients, but the kitchen is too small and the stoves aren't enough. Customers are queuing all the way out into the street. Even though the restaurant owner (TSMC) is desperately expanding the kitchen (with plans to increase capacity by 2026), the expansion speed isn't fast enough to keep up with the orders. So, the owner has to direct some customers to neighboring restaurants (Samsung, Intel) or outsource to specialized packaging companies (ASE, Amkor).
The key point is that this isn't due to TSMC's lack of technology; it's simply a physical limitation on production capacity. As long as AI demand remains high, this supply-demand imbalance will continue.
---
2. Advanced Manufacturing Spillover: Samsung and Intel Benefit, but Don't Celebrate Too Soon
The orders that TSMC can't handle for the core manufacturing of chips are mainly going to Samsung and Intel.
- Samsung's Big Wins:
- Tesla FSD Chips: Samsung has secured the contract to manufacture Tesla's next-generation AI chip, "AI6," worth $16.5 billion, the largest order in its history.
- Reason: It's not because Samsung's technology has suddenly surpassed TSMC's; rather, TSMC's 2-nanometer production capacity is full, and Tesla needed to find another supplier.
- Other Customers: Samsung is also in talks with AMD for the next generation of server CPUs, and Qualcomm is considering using Samsung's 2-nanometer process for some mid-range mobile chips.
- Key Data: Samsung's 2-nanometer chip yield has risen from less than 50% at the beginning of the year to over 70%. This shows that Samsung's technology is improving, but it's still catching up, not leading.
- Intel's Breakthrough:
- Microsoft's Order: Microsoft has placed an order with Intel to manufacture its next-generation AI accelerator, Maia 2, using Intel's 18A or 18A-P process. This is the first major external order for Intel's 18A technology.
- Positioning: Intel's 18A technology is roughly on par with TSMC's 3-nanometer level. Microsoft chose Intel for "supply chain security" reasons, not because Intel's technology is superior.
In plain language:
Samsung and Intel are now acting as "substitutes." The main player (TSMC) is struggling, so they step in to help. While the substitutes can handle some of the work, customers know that TSMC is still the core. These orders are a result of capacity constraints and do not challenge TSMC's dominant position in high-end chip manufacturing.
---
3. Advanced Packaging Spillover: The Real "Structural Opportunity" for Companies Like ASE
If the spillover in advanced manufacturing is a passive redistribution, then the spillover in advanced packaging (such as CoWoS) is a proactive outsourcing trend that's happening faster and more significantly.
- Why More Packaging Work?
AI chips require stacking multiple small components (processing cores, memory, etc.) together, a process called advanced packaging. TSMC's CoWoS packaging capacity is fully utilized, even for companies like NVIDIA and Apple.
- TSMC's Strategy:
TSMC is focusing on the most critical and high-value packaging technologies (like CoWoS-L, SoIC) and outsourcing the less complex, standardized packaging steps to specialized packaging companies (OSATs).
- Who's Benefiting?
- ASE and UMC: These long-time partners of TSMC have invested billions to expand their production capacity to handle the overflow orders. ASE's advanced packaging revenue is expected to grow by over 60% in 2026.
- Amkor: It's investing $7 billion to build a new factory in Arizona to provide packaging services for TSMC's U.S. wafer plants.
- Equipment Manufacturers: Companies like Hon Hai and Shin Yuon, which supply equipment to packaging companies, are also profiting.
In plain language:
TSMC acts as the chief designer, creating the complex blueprints, while the actual packaging is done by specialized companies like ASE and Amkor. These companies, having worked with TSMC for a long time, have the expertise to handle this work. As AI chips become larger and more complex, the importance of packaging will only increase, and this outsourcing trend is long-term and sustainable.
---
4. Chain Reactions: Even Mature Process Manufacturers Are Benefiting
You might think that 2-nanometer and 3-nanometer technologies are far from everyday use, but they actually affect the entire chip industry. TSMC is gradually phasing out some lower-end production lines (such as 6-inch and some 8-inch wafers) to focus on more profitable advanced processes.
- Who's Taking Over?
Companies like World Advanced and UMC, which specialize in mature processes.
- Results:
In June 2025, World Advanced's revenue hit a record high, and UMC also saw its highest revenue in nearly 47 months.
In plain language:
TSMC is shifting its focus to higher-end products, leaving lower-end production to companies like World Advanced and UMC. These companies have seen a surge in business as they take over these orders. This shows that TSMC's strategy has helped each level of the supply chain find its place.
---
5. Summary and Outlook: Benefits Are Not Easy to Come By; Technology Accumulation Matters
Let's summarize the essence of this spillover pattern:
- Clear Hierarchies:
- Advanced Manufacturing (Samsung, Intel): They're getting orders due to capacity constraints, not because they've surpassed TSMC in technology or customer trust. They serve as secondary suppliers to enhance supply chain resilience.
- Advanced Packaging (ASE, Amkor, etc.): They're receiving structural outsourcing orders because packaging requires relatively lower technical barriers, and OSAT companies have accumulated the necessary expertise. This is a result of the industry's growing specialization.
- Benefits Are Conditional:
Although orders are flowing over, not everyone can handle them.
- Technical Accumulation: Samsung's yield improvement (70%) was necessary to secure the Tesla order; ASE's experience is crucial for getting packaging contracts.
- Production Expansion: Expansion takes time; Amkor's new factory won't be fully operational until 2027, so it can't handle many orders yet.
- Supply Chain Support: Equipment, materials, and talent are all essential.
Implications for Everyone:
- Investment Perspective: Don't focus solely on TSMC. As AI demand continues to grow, companies in advanced packaging (ASE, Amkor) and mature processes (World Advanced, UMC) are likely to benefit more significantly due to the structural nature of their businesses. The opportunities for Samsung and Intel depend on TSMC's capacity recovery, which is more volatile.
- Industry Trend: Chip manufacturing is moving towards a more specialized division of labor. TSMC remains the core, but it no longer handles all steps. Future competition will involve not just technology but also supply chain collaboration.
In conclusion, the AI trend will continue, and TSMC's spillover effects will persist for some time. However, who will truly benefit depends on which company has the stronger technology, more stable production lines, and a more reliable supply chain.