Taiwan Semiconductor Manufacturing (TSM +1.27%) is emerging as one of the smartest ways to invest in the artificial intelligence (AI) boom.
Nvidia (NVDA 0.13%) currently accounts for an estimated 85% to 92% share of the AI accelerator market. However, as large cloud companies look to reduce costs and scale faster, many are shifting toward custom chips and more specialized computing systems.
Regardless of which approach gains traction, most advanced AI chips still rely on Taiwan Semiconductor — also known as TSMC — for manufacturing.
Image source: Getty Images.
Chip manufacturing and packaging bottleneck
With AI infrastructure spending expected to reach around $660 billion in 2026 alone, TSMC is already benefiting from surging demand for advanced chip manufacturing. In the first quarter, the company reported revenue of $35.9 billion, up nearly 39% year over year, while gross margin expanded 3.9 percentage points to 66.2%.
Taiwan Semiconductor Manufacturing
Today’s Change
(1.27%) $4.67
Current Price
$372.75
Key Data Points
Market Cap
$1.9T
Day’s Range
$370.68 – $373.81
52wk Range
$156.00 – $390.20
Volume
107K
Avg Vol
13M
Gross Margin
61.02%
Dividend Yield
0.91%
AI is now at the center of TSMC’s business. High-performance computing (HPC), which includes AI-related chips, has grown from about 46% of revenue in the first quarter of 2024 to roughly 61% in the latest quarter.
The expansion of AI infrastructure is increasingly constrained by the ability to manufacture and package advanced chips at scale. CoWoS packaging technology, which connects processors with high-bandwidth memory, is also in short supply. As TSMC controls roughly 72% of the pure-play foundry market and over 90% of leading-edge chip production, it is at the center of this supply-demand imbalance. The company also enjoys significant pricing power.
A business model with Nvidia-like upside
Unlike chip designers, TSMC does not compete to build the best AI chip. Instead, it provides the manufacturing capabilities required to produce them, supplying customers such as Nvidia, Advanced Micro Devices, Broadcom, and large cloud companies building custom chips.
Designing chips for a specific manufacturing process requires close integration with the foundry. It can also require several years, making it difficult for customers to switch suppliers quickly once a chip is in production. Advanced nodes (7-nanometer and below) are already accounting for about 74% of TSMC’s wafer revenues.
TSMC is also investing heavily to maintain its lead in advanced chip manufacturing. The company expects 2026 capital expenditures to be between $52 billion and $56 billion, with a significant portion directed toward expanding leading-edge nodes and advanced packaging capacity.
Analysts expect TSMC’s revenue to rise from around $163.9 billion in 2026 to nearly $204.4 billion in 2027, $249.4 billion in 2028, and roughly $311.5 billion by 2030. This implies that the company can almost double its revenue within four years.
TSMC’s path to Nvidia-like reputation and scale, however, looks fundamentally different. While Nvidia dominates the AI accelerator market, TSMC’s growth is tied less to any single product cycle and more to the steady expansion of AI infrastructure. As a result, TSMC stock is also increasingly becoming a core way to invest in AI. If AI spending continues to expand at current rates, TSMC has a realistic path to becoming just as central to AI investing as Nvidia is today.