TSMC raises its 2026 forecast, confirming steady demand for AI chips

Taiwan Semiconductor Manufacturing Co. has sent one of the strongest signals yet to the entire AI infrastructure market: the world's largest contract chipmaker raised its 2026 revenue forecast, showing that even with the conflict in the Middle East and worries about the global economy, demand for the most advanced semiconductors is not weakening. For the industry, this is an important marker: if TSMC is confidently raising expectations, big customers are still investing heavily in data centers, accelerators and server hardware for AI.

For the first quarter, which ended on March 31, 2026, TSMC reported revenue of $35.9 billion, up 40.6% year over year in dollar terms. In New Taiwan dollars, revenue grew 35.1% to NT$1.134 trillion, and net profit rose 58.3% to NT$572.48 billion. Diluted earnings per share came to NT$22.08, gross margin to 66.2% and operating margin to 58.1%. The numbers are not just strong on their own: the company also said quarterly revenue slightly exceeded its own guidance, and its second-quarter outlook points to further growth, to $39-40.2 billion.

The main source of this momentum is not smartphones but high-performance computing, which TSMC uses to classify a large share of AI workloads. In the first quarter, the HPC segment generated 61% of the company's revenue, versus 55% in the previous quarter, while smartphones made up 26%. The breakdown by process node is just as telling: 3nm chips accounted for 25% of wafer revenue, 5nm for 36% and 7nm for 13%, and the combined share of advanced technologies at 7nm and below reached 74%. This means money in the sector is increasingly concentrated in the most expensive and technologically complex nodes, which are critical for accelerators, server processors and AI infrastructure as a whole.

Even more important, TSMC raised its full-year guidance. It now expects 2026 revenue growth of more than 30% in dollar terms, compared with roughly 30% previously. At the same time, management indicated that capital expenditure is shifting toward the upper end of the previously announced range of $52-56 billion.

The logic is simple: demand is so strong that the company is trying to obtain and install equipment as quickly as possible, primarily for 3nm. Management separately acknowledged that capacity is still short and bottlenecks persist. Against this backdrop, TSMC also highlighted progress on its next generation: the 2nm family entered volume production back in the fourth quarter of 2025 and is ramping successfully to meet demand from smartphones as well as HPC and AI orders.

Geopolitics is an important nuance here. The forecast increase came as investors were weighing the possible consequences of the Middle East conflict for the global economy and supply chains. TSMC is not ignoring that risk, but it is effectively showing that so far it has not broken the AI investment cycle. Moreover, the revenue mix points to the company's heavy dependence on North American customers, who accounted for 76% of first-quarter sales. It is one more sign that American cloud providers, accelerator designers and large tech companies continue to set the pace for the whole market.

For the market, this means the following: talk of a possible cooling of the AI boom is not yet supported by figures from the very center of the global supply chain. TSMC sees not a one-off spike but multi-year demand for leading-edge nodes, and is ready to accelerate investment so as not to lose customers. If the trend holds, the winners will include not only AI chip designers but the whole stack around them, from memory and packaging suppliers to data center operators. Conversely, any disruption to TSMC's capacity expansion will be felt immediately across the industry.