TSMC in a tight spot: demand for high-end chips exceeds capacity by a factor of 3

It is no longer a secret that we are in the middle of a semiconductor supercycle. But what TSMC CEO C.C. Wei is now publicly admitting brings a whole new level of drama into play: according to Wei, the demand for advanced manufacturing processes at TSMC is currently three times the actual production capacity. This is not only an understatement of “scarcity”, but also reveals a systemic failure of the global manufacturing infrastructure in the age of AI, 5G, automotive digitalization and cloud-based high-performance computing.

Monopoly through technological leadership

TSMC has become by far the most important foundry in the high-end sector over the last ten years through smart investments, technological leadership and excellent partnerships with customers such as Apple and NVIDIA. While Samsung Foundry is keeping pace technologically, but is struggling with yield problems, and Intel is just about to revamp its foundry services, TSMC is the de facto sole supplier for 5nm, 3nm and now also for the upcoming 2nm process.

The problem is that production in these nodes is extremely capital-intensive, technically highly complex and logistically challenging. EUV lithography, packaging innovations such as CoWoS and 3D stacking consume billions per site. Building a new fab takes years, not months.

Demand: “Explosive” would be an understatement

With the rise of generative AI, LLMs and edge AI for vehicles and industrial applications, the demand for high-performance chips has multiplied. NVIDIA, AMD, Broadcom, Marvell, MediaTek, even Google with its Tensor SoCs, are all vying for TSMC’s capacity. Many secure long-term supply contracts, but these prefer high-volume customers such as Apple or NVIDIA. Smaller or new players, start-ups or second-tier providers, on the other hand, are often left out in the cold or end up at the back of the queue. The bottleneck is therefore not being felt evenly, and is mainly affecting those who do not have long-term supply agreements. What does that mean? Delayed product launches, detours via older nodes, with consequences for performance and competitiveness.

Geopolitics meets technology

The bottleneck is not just an economic problem, but has long been a geopolitical one. Taiwan is a strategic hotspot between China and the West and TSMC is at the heart of this tension. The US is investing billions in local manufacturing (CHIPS Act), but the reality remains: The most advanced capacity continues to be in Hsinchu, not Arizona. Even TSMC’s new US sites are not yet in regular operation and will not be able to deliver significant quantities until 2026/27 at the earliest.

What’s more, Japan (with Rapidus), South Korea (Samsung) and Europe (Intel Magdeburg, STMicro/NXP in France) are also trying to build up their own capacities with subsidies. But this is a marathon, not a sprint. And until then, the dependency on a single supplier will remain, a supplier that is currently several quarters behind.

Market distortions: From oligopoly to dependency

When the world’s largest contract manufacturer admits that it can no longer keep pace with growth, this is not a sign of a healthy market mechanism, but of structural instability. Manufacturing prices are rising, smaller customers are being squeezed out, innovation cycles are stalling. And although competitors such as Intel Foundry Services and Samsung Foundry could theoretically step in, confidence remains limited. Many companies are reluctant to qualify their designs for alternative manufacturers – the risks in terms of yield, compatibility, clocking or thermal behavior are too great.

Strategic question: Who will win the next decade?

The long-term implications are far-reaching. Whoever is able to supply highly developed chips reliably and in quantities will dictate not only technology but also policy. The Western world has realized that technological sovereignty depends not just on IP, but on manufacturing. And this is where the bottleneck lies.

What C.C. Wei says is therefore more than an economic comment, it is a strategic wake-up call to politics, industry and research: either you solve the production dilemma through massive, coordinated investment, or you live with the risk of remaining technologically dependent and vulnerable.

Source: Wccftech