TSMC’s Place in the AI Stack
TradingKey - Taiwan Semiconductor Manufacturing (TSM) has established itself as a crucial player in the Artificial Intelligence (AI) supply chain and this has begun to manifest itself in the performance of TSMC’s shares.
TSMC’s scale and process leadership have resulted in a quasi-monopoly of advanced semiconductor manufacturing, with its offerings including GPUs, AI ASICs, high-performance CPUs and various types of logic chips that power today’s data centers.
For semiconductor designers wanting to manufacture their most groundbreaking chips in high volume with robust yields, TSMC is the only viable option at the moment.
This necessity causes customers to do more than just book time in TSMC’s factories because they also enter multi-year agreements with TSMC during which their technology roadmaps and production capacities are forecasted for multiple years, enabling TSMC to have very clear visibility into customer demand and significant pricing leverage.
On April 7, TSMC's stock price was up 13% for the year to date in spite of a general tech decline due to strong numbers versus mere hype.
In Q4 of 2025, TSMC reported net income and earnings per share of 35% increases from the prior year, and revenue of 20.5%. This pattern has existed for years and is reflected by TSMC being a cyclical business; however, the current AI cycle has created a different kind of demand and growth rate than other cycles within TSMC's history.
Due to a number of reasons, many current valuations are considered "crazy," especially because of the tremendous growth achieved over the last few quarters. However, this quarter is only part of the evidence for why TSMC's revenues will continue at these levels for the remainder of the year; the case for TSMC's ongoing growth is much stronger than just this past quarter.
Currently, TSMC has a 72% market share in the pure-play foundry industry. This is an increase of six percentage points versus when they had 66% in Q3/24. This is why TSMC has a solid position for continuing to dominate the pure-play foundry market beyond this period because they have a solid manufacturing facility, an ecosystem of locked-in customers, and the yield advantage that no other foundry can provide.
Because of the nature of co-designs and long-term commitments, the ability to switch becomes costly, and TSMC has already locked in additional revenues for the future that will be less likely to have short-term variability.
AI Accelerators Push the Growth Runway Further
The major change during TSMC's recent quarterly earnings call was that C.C. Wei, TSMC's CEO, raised TSMC's forecast for AI accelerators.
Currently, AI accelerators make up approximately 18% of TSMC's total annual revenue in 2025, but TSMC is now projecting this percentage will grow at an annual compound growth rate (CAGR) of 55% to 60% from 2025 through 2029, which is above TSMC's prior estimate of a 45% CAGR for AI accelerators. This increase implies that demand for AI construction ability is stronger than TSMC had previously estimated and that the amount of spending on AI-related infrastructure is far from its peak.
Nvidia’s Lead, and How the Game Is Changing
With its early vision and execution, NVIDIA (NVDA) has become a leader in the AI revolution.
NVIDIA dominates the global GPU (Graphics Processing Unit) market with a market share of over 90%.
Through the creation and distribution of its CUDA platform (a tool used by engineers and developers to create AI applications), NVIDIA helped to shape the AI industry and its growth.
After launching the CUDA platform, NVIDIA created numerous tools that drove AI development and made them available for free. Some of these tools included the original CUDA toolkit that inspired many of the first AI developers.
NVIDIA also acquired Mellanox Technologies, a leading provider of state-of-the-art data centre networking products, which gave NVIDIA the ability to continue to enhance its offerings for customers wanting to integrate AI applications and solutions into their data centre.
This history of technology acquisitions, combined with NVIDIA's acquisition of Groq and SchedMD, clearly shows that NVIDIA is willing to invest in technologies early on as an avenue for future investment. This licensing arrangement makes it easier for NVIDIA to use Groq's technology in the actual deployment of AI solutions into the CUDA ecosystem. SchedMD provides critical software capabilities that will be extremely useful for agentic AI (e.g., reinforcement learning) workflows.
Why a Broader Chip Race Can Benefit TSMC
The market is changing even if Nvidia continues to be an ongoing success. Many major customers are developing their own custom AI ASICs to give them greater returns on performance relative to their costs; plus, they have entered into agreements with companies such as AMD for additional options to replace certain Nvidia GPUs.
By this very change, Nvidia will gradually lose market share.
However, this same issue is also beneficial to TSMC because as things change, more chip designers and architects will begin working together and the need for a neutral manufacturer with large scale will become more important than before. New and custom designs need an operator that can manufacture them at very high volumes, which is exactly what TSMC does best.
Furthermore, as agentic AI workloads grow and new technologies like autonomous driving come into play, TSMC will continue to benefit from the increasing demand for CPUs in data centers.
Cyclical Worries, Valuation, and What Could Go Wrong
There's some cause for concern regarding the cyclical nature and long-term valuations of TSM. Historically, TSM's revenues have shown both positive and negative correlations with total semiconductor company revenues, and it is clear that the most recent price run-up has increased the expectations for management execution.
If there is a slowdown in AI infrastructure spending sooner than expected or a competitor achieves a significant improvement in advanced manufacturing, TSM's growth rates may slow substantially. In addition, the pricing of TSM's stock currently reflects market expectations for the expansion of the mix of AI accelerators, expecting an increase at an average annual growth rate of 50+% from now through year-end 2029.
Factors that would likely negatively impact the growth outlook could have major impacts on valuation, especially following a strong start to 2023. Although the negatives are certainly not "deal-breakers" for the fundamental thesis regarding TSM, they are certainly part of the analysis investors should be performing.
The Long-term Case for TSMC Stock Over Nvidia
There is no clear choice between TSMC and Nvidia because while both have sizable upside potential, their risk premiums could differ significantly.
TSMC is uniquely positioned to take advantage of the fact that, as AI demand continues to rise and more competitors enter the market with their own branded solutions, there will be continued demand for low-cost, custom chip solutions.
Since TSMC manufactures on behalf of all OEMs as part of its 72% share of the foundry business, it will capture additional data center CPU and automotive-related orders as well as the expanded opportunity for AI accelerators, since it announced its next major product cycle would be much larger than previous cycles.
The combination of those factors gives TSMC significantly more time and opportunity to grow than Nvidia due to the fact that TSMC is also approximately 12 times smaller than Nvidia on a market cap basis.
Bottom Line for Investors
Historically, AI will be remembered as one of the largest technological changes that has occurred in decades, and AI investment continues to rise.
As a company that is pioneering AI, Nvidia is expected to continue delivering outstanding, long-term returns to its shareholders; however, if one's goal is to invest in companies that will provide strong, long-term returns as the AI ecosystem becomes more diverse, then TSMC may be the healthier investment for participating in the AI supercycle.
Given TSMC's rapidly growing accelerator product lines, increased penetration into the marketplace through strong customer relationships, commitments from customers for many years, and unique manufacturing capabilities, the stock of TSMC looks set up very well to profit from the ongoing global expansion of AI across many types of chips and customers.
Find out more
Comments