TSMC's order book as a chip manufacturer has become saturated, forcing some of the industry's largest companies to seek alternative suppliers. This might sound like a problem, but for TSMC, it has quickly become the most powerful proof of its market position. Every advanced process node is booked up years in advance, and the stock price has moved close to historic highs.

The stock closed in Frankfurt on Friday at €407, with a weekly gain of just over 11%. Year-to-date gains are approximately 49%, with the historic record of €414 within reach. The 50-day moving average of €342 highlights the strength of the recent rally—this momentum was briefly interrupted when Nikkei Asia reported that key customers were exploring alternatives.

BYD, AMD, and Google are in talks at various stages with Samsung regarding future chip foundry services. BYD needs processors for advanced driver assistance systems, AMD is evaluating Samsung's production of certain CPUs starting in 2028, and Google is exploring options for its Axion processors and TPU components. The reason is straightforward: TSMC's capacity for 3-nanometer and 2-nanometer processes is completely booked by companies like NVIDIA, Apple, AMD, and Broadcom. A manager at a Chinese automotive chip manufacturer told Nikkei Asia that TSMC prioritizes advanced process nodes because they offer higher margins, leaving little room for newcomers. Samsung's idle capacity, even with lower yields, has suddenly become more attractive.

The stock fell about 3.5% after this news broke, but quickly rebounded. Investors quickly concluded that turning away customers due to demand exceeding supply is not a competitive disadvantage—it is pricing power in its purest form.

TSMC is not taking these bottlenecks lightly. The company is directly integrating NVIDIA's artificial intelligence systems into its own fabs to accelerate simulation speeds and improve yields. Management expects full-year 2026 revenue growth exceeding 30%. On the hardware front, a new decade-long agreement with Amkor Technology secures advanced packaging capacity in Arizona, building a fully domestic U.S. supply chain.

In fact, packaging has become the next competitive battleground. CoWoS technology, essential for AI chips, still faces a capacity shortfall of about 20%, though TSMC expects to narrow this to around 10% by the end of 2026. Monthly output could reach 120,000 to 140,000 wafers, with partners adding another 50,000 to 60,000, bringing the industry's total production close to 200,000 wafers per month. TSMC's proprietary 5.5-mask CoWoS solution achieves yields exceeding 98%, compared to Intel's EMIB-T at approximately 90%, while Samsung's 2-nanometer process remains in the mid-50% range.

The next step is CoPoS—chip-on-package-on-substrate packaging. Subsidiary VisEra has been running research lines since 2025, with materials and equipment certification expected to be completed in June 2026. NVIDIA's Feynman platform is the leading candidate for the first round of mass production, no earlier than 2028.

All of this comes with a steep price tag. TSMC's 2026 capital expenditure budget is $5.2 billion to $5.6 billion, with 70-80% earmarked for advanced process nodes, approximately 10% for specialty process nodes, and the remainder for packaging, testing, and mask production. Advanced packaging accounted for approximately 8% of revenue in 2025 and is expected to exceed 10% for the first time in 2026.

An overhanging risk is a pending patent investigation by the U.S. International Trade Commission. The first ruling is expected to be released this month; a favorable outcome would remove the stock's last near-term obstacle.

The bigger picture is structural scarcity. Artificial intelligence infrastructure, high-performance computing, and automotive electronics are all pulling demand in the same direction, and this dynamic shows no signs of easing before the end of the decade. For TSMC, selling out years in advance is not a luxury problem—it is the core of the investment thesis.