Profit Soars 77 Percent as the Artificial Intelligence Boom Tightens Its Grip
TAIPEI, Taiwan — Taiwan Semiconductor Manufacturing Company reported another record quarter on Thursday, underscoring how the global race to build artificial intelligence infrastructure has strengthened the position of the world’s dominant contract chipmaker.
TSMC’s net profit climbed 77 percent from a year earlier to NT$706.6 billion, or about $22 billion and €19.1 billion. Analysts surveyed through LSEG had expected approximately NT$632.6 billion. Revenue rose 36 percent to NT$1.27 trillion, equivalent to $40.2 billion. The company also recorded a gross margin of 67.7 percent and an operating margin of 60.3 percent, both unusually high for large-scale manufacturing.
The result reflects more than an increase in the number of chips sold. TSMC is producing a greater proportion of its wafers with its most advanced—and generally most profitable—manufacturing technologies.
AI Chips Move From a Side Business to the Center of TSMC
The most important source of growth is spending by operators of data centers, which are buying enormous quantities of processors designed by Nvidia, AMD, Broadcom and other TSMC customers. These chips are used to train and operate generative-AI systems.
High-performance computing, a broad category that includes AI accelerators, server processors and other powerful chips, generated 66 percent of TSMC’s quarterly revenue. Smartphones accounted for 22 percent. Meanwhile, the company’s advanced 2-, 3- and 5-nanometer processes together supplied 77 percent of wafer revenue. The new 2-nanometer process contributed 3 percent and 3-nanometer chips 30 percent.
The statement in the original German report that “modern high-performance chips” contribute roughly one-quarter of revenue requires clarification. That narrower figure most likely refers to dedicated AI accelerators. TSMC’s broader high-performance-computing division is already much larger, accounting for nearly two-thirds of sales.
Why Profit Grew Much Faster Than Revenue
Several forces explain why earnings advanced 77 percent while revenue increased by 36 percent:
- A richer product mix: Advanced AI and high-performance chips command higher wafer prices than many mature-node products used in appliances, basic electronics and industrial equipment.
- High factory utilization: Demand from Nvidia and other AI-chip designers has kept TSMC’s advanced production lines and packaging facilities heavily loaded. Fixed costs are therefore spread across more wafers.
- Improving manufacturing yields: As production processes mature, a larger share of the chips cut from each wafer meets performance requirements and can be sold. This is particularly important for large, expensive AI processors.
- Advanced-packaging demand: AI systems require technologies such as CoWoS, which combine processors and high-bandwidth memory in the same package. TSMC’s strength in both fabrication and packaging gives it another source of revenue and customer dependence.
- Pricing power: Capacity at advanced nodes remains scarce, giving TSMC greater leverage in customer negotiations. Reports of premium-priced, accelerated production orders also suggest that some customers are paying more to secure capacity.
- Economies of scale: TSMC’s enormous volume allows it to distribute research, equipment and development costs across a much larger revenue base than its foundry competitors.
The resulting 67.7 percent gross margin shows that the earnings increase was not produced by sales growth alone. It also came from a pronounced expansion in profitability.
There are countervailing pressures. The early ramp-up of 2-nanometer production is expected to reduce gross margin by roughly three to four percentage points before yields and volumes improve. New factories in the United States and Japan are also more expensive to operate than comparable facilities in Taiwan. TSMC expects its third-quarter gross margin to moderate to between 65 and 67 percent.
TSMC Has Competitors—but No Equivalent Rival
TSMC does not have a legal or technological monopoly. It nevertheless occupies a position that, for the most advanced outsourced chips, increasingly resembles one.
Counterpoint Research estimated that TSMC held approximately 73 percent of the global pure-play foundry market in the first quarter of 2026. TrendForce’s comparable estimate was 72.3 percent. Samsung remained a distant second; its share was about 7 percent at the end of 2025.
| Competitor | Principal strength | Main limitation relative to TSMC |
|---|---|---|
| Samsung Foundry | Advanced 2-nanometer technology, large investment capacity and integrated memory expertise | Smaller customer base and continuing questions about yields, consistency and external-customer trust |
| Intel Foundry | Intel 18A process, advanced packaging, U.S. government support and domestic manufacturing | Still establishing itself as a dependable large-scale manufacturer for outside customers |
| SMIC | China’s largest foundry, strong domestic demand and government backing | Restricted access to leading lithography equipment and several generations behind at the frontier |
| GlobalFoundries | Strong in automotive, communications, industrial and specialty chips | Does not compete in the most advanced 2- or 3-nanometer segment |
| UMC | Efficient, established production of mature-node chips | Focuses mainly on older processes rather than frontier AI chips |
| Rapidus | Japanese government and corporate backing for a planned 2-nanometer business | Early-stage production, limited scale and no long commercial manufacturing record |
Samsung and Intel are the only immediate contenders with credible leading-edge processes. Samsung’s SF2 and Intel’s 18A technology compete broadly with TSMC’s N2 generation, although node names are marketing categories and do not provide a perfect technical comparison.
Intel has demonstrated progress with 18A and is expanding its external foundry business. Samsung is developing several 2-nanometer variants for mobile, high-performance and automotive chips. Japan’s Rapidus could become a longer-term challenger, but it must first prove that it can achieve competitive yields and mass production.
TSMC’s advantage is consequently not confined to transistor dimensions. It rests on four interconnected assets: dependable yields, immense production capacity, a mature design-software ecosystem and a long record of not competing directly with the companies whose chips it manufactures. Replicating that system would require years and tens of billions of dollars.
Which Products Could Become More Expensive?
Industry reports indicate that TSMC is seeking price increases of roughly 5 to 10 percent across advanced processes, potentially covering nodes from 2 to 7 nanometers. These figures have been reported by the technology press and supply-chain sources; they should not be treated as a universally confirmed public price list.
The most directly affected products would include:
- Nvidia, AMD and custom AI accelerators used in data centers
- Server CPUs and networking processors
- Apple-designed processors for future iPhones, iPads and Mac computers
- Qualcomm and MediaTek chips for premium Android phones
- AMD and Apple personal-computer processors
- High-end gaming graphics cards and consoles
- Advanced automotive computing and driver-assistance systems
- Networking equipment used in AI clusters and cloud data centers
A 5-to-10-percent increase in wafer prices would not automatically produce an equivalent rise in retail prices. The processor is only one component of a phone or computer, and large customers may absorb some costs, negotiate different terms or offset them through design improvements.
The effect could be more visible in AI servers and graphics cards. Their processors are exceptionally large, often require advanced packaging and high-bandwidth memory, and are being sold into a market where demand already exceeds supply. Cloud providers may pass those higher costs on through more expensive AI-computing subscriptions rather than through a conventional retail price.
Products made with older manufacturing processes—basic household electronics, inexpensive microcontrollers and many conventional automotive chips—may initially experience less direct pressure. But capacity reallocations, energy costs and separate price increases at mature-node foundries could still affect them.
A Dominant Position, With Expensive Risks
TSMC raised its expected 2026 capital spending to between $60 billion and $64 billion and now forecasts full-year revenue growth of slightly more than 40 percent. It is also expanding in Arizona, where its total planned investment has reached $265 billion, partly in response to customers and governments seeking production outside Taiwan.
That expansion may make the global chip supply more resilient, but it will be costly. Overseas factories, the rapid transition to 2 nanometers and the possibility of slower AI investment could eventually weigh on margins.
For now, however, the central fact of the semiconductor industry remains unchanged: nearly every major designer of advanced processors can draw its own blueprints, but most still depend on TSMC to turn those designs into chips—at sufficient scale, with acceptable yields and on schedule.