Taiwan Semiconductor Manufacturing Co. reported first-quarter 2026 revenue of $35.9 billion and a 58% jump in profit, confirming that the company’s role as the world’s most important chipmaker is only growing. Artificial intelligence processors now account for 61% of TSMC’s total sales, up from roughly 53% a quarter earlier, according to figures CEO C.C. Wei shared on the April earnings call. And in a detail that sent ripples through the semiconductor supply chain, Wei disclosed that every 2nm wafer TSMC can produce through the end of 2026 has already been claimed by customers.
The numbers behind the quarter
TSMC’s Form 6-K filed with the SEC puts first-quarter revenue at US$35.90 billion (NT$1,134.10 billion), with gross margin holding above 58%. The filing is a legal document submitted under U.S. securities law, making it the most reliable public record of the quarter’s performance.
The Associated Press independently confirmed the 58% year-over-year profit increase, a figure that reflects both surging AI demand and the pricing power TSMC commands at its most advanced process nodes. For context, the company’s revenue in the same quarter a year ago was roughly $25.8 billion, meaning TSMC added more than $10 billion in quarterly sales in just twelve months.
Management guided second-quarter 2026 revenue to between $28.4 billion and $29.2 billion, a sequential dip that aligns with typical seasonal patterns in the smartphone segment. But the full-year trajectory still points sharply upward: analysts polled by Refinitiv expect TSMC to surpass $140 billion in annual revenue for 2026, driven almost entirely by AI-related orders.
Why 2nm capacity matters so much
TSMC’s 2nm node, known internally as N2, represents the company’s biggest generational leap in years. It introduces gate-all-around transistor architecture, which improves power efficiency and performance enough to matter for the massive chips powering AI training clusters. Apple, Nvidia, AMD, and Qualcomm have all been reported as early adopters of the node, and the competition for wafer starts has been fierce.
Wei’s statement that every 2nm wafer through 2026 is already allocated is unusual even by semiconductor standards, where 12- to 18-month lead times are routine. It means that any chip designer without a confirmed booking today cannot get its product into 2nm production this year. For smaller firms or startups hoping to compete at the leading edge, the door is effectively closed until 2027 at the earliest.
The sell-out also gives TSMC extraordinary leverage over pricing. When demand outstrips supply this dramatically, customers accept higher wafer costs and commit capital years before their products ship. That dynamic benefits TSMC’s margins but raises costs across the AI hardware stack, from data-center GPUs to the cloud computing services built on top of them.
AI’s growing share of the revenue mix
The 61% AI revenue figure, cited by Wei during the earnings call and reported by multiple outlets including Reuters, marks a milestone. As recently as the first quarter of 2025, AI-related chips accounted for less than half of TSMC’s sales. The rapid shift reflects both the sheer volume of silicon that hyperscale cloud providers are ordering and the premium pricing attached to the advanced packaging and process technologies those chips require.
TSMC’s Chip-on-Wafer-on-Substrate (CoWoS) advanced packaging lines, which are critical for assembling Nvidia’s Blackwell and next-generation Rubin GPUs, have been running at full utilization for over a year. The company has been expanding CoWoS capacity aggressively, but demand continues to outpace supply. Microsoft, Google, Amazon, and Meta have all publicly discussed their AI infrastructure spending plans for 2026, and much of that spending flows directly to TSMC through chip orders placed by their silicon partners.
The concentration does carry risk. If AI spending were to slow, whether because of a macroeconomic downturn, a shift in enterprise budgets, or diminishing returns from large language models, TSMC’s revenue mix would leave it more exposed than in previous cycles. Executives have acknowledged this in prior calls but maintain that the structural demand for AI compute is durable, not speculative.
The Iran conflict and supply-chain risk
During the earnings call, TSMC executives raised the ongoing conflict involving Iran as a potential risk to operations, a disclosure the Associated Press highlighted in its coverage. The concern is not that TSMC’s fabs in Taiwan, Arizona, or Japan face direct threat, but that the conflict’s secondary effects could ripple through the supply chain.
Energy prices are the most immediate transmission channel. A sustained disruption to oil and gas flows from the Persian Gulf would raise electricity costs in Taiwan, where TSMC’s largest and most advanced fabrication complexes are located. Shipping insurance premiums for routes through the Strait of Hormuz have already climbed in recent months, adding cost to the movement of chemicals, gases, and equipment that fabs consume in enormous quantities.
There is also the question of sanctions. Depending on how the conflict evolves, new restrictions on trade with certain countries could complicate procurement of specialty materials or limit TSMC’s ability to serve specific customers. The company did not detail specific scenarios on the call, but the fact that management chose to flag the issue during a quarterly earnings discussion, rather than burying it in a routine risk-factor update, suggests it views the situation as more than hypothetical.
What this means for TSMC’s customers and competitors
For Nvidia, AMD, Apple, and the growing roster of companies designing custom AI accelerators, TSMC’s results reinforce a reality they already know: there is no viable alternative at the leading edge. Samsung Foundry continues to struggle with yields on its own gate-all-around process, and Intel Foundry Services, despite billions in U.S. government subsidies through the CHIPS Act, is still years away from competing for the highest-volume AI chip orders.
That leaves TSMC as the single point of failure, and the single point of leverage, for the global AI hardware buildout. Companies that locked in 2nm capacity early are positioned to ship next-generation products on schedule. Those that did not may face delays that cost them market share in a sector where six months can determine who wins a hyperscaler’s next procurement cycle.
TSMC’s expanding U.S. footprint offers some geographic diversification. The company’s Arizona fabs are expected to begin volume production on the N4 process in 2025 and ramp toward more advanced nodes in subsequent years, partly in response to U.S. government incentives and partly to reassure American customers wary of Taiwan Strait tensions. But Arizona alone cannot absorb the demand that Taiwan’s mega-fabs handle today, and 2nm production will remain concentrated in Taiwan for the foreseeable future.
Where the hard data ends and the forecasts begin
Investors and analysts should draw a clear line between what TSMC’s SEC filing confirms and what the earnings call implies. The $35.9 billion revenue figure and the 58% profit growth are anchored in regulatory disclosure. The 61% AI revenue share and the 2nm sell-out are management statements from the call, widely reported but not yet documented in a formal filing. Both are likely accurate, but they carry the caveats that accompany any forward-looking commentary: execution risk, geopolitical uncertainty, and the possibility that customer commitments shift if market conditions change.
What is not in dispute is the trajectory. TSMC is converting the AI boom into financial results at a pace that no other chipmaker can match. Its technology lead, its manufacturing scale, and its customer relationships form a moat that competitors have spent tens of billions of dollars trying to cross without success. The open question is not whether TSMC will dominate advanced chip manufacturing in 2026, but whether the concentrated demand, geopolitical friction, and capacity constraints that define this moment will eventually force the industry to find alternatives, or simply make TSMC more indispensable than ever.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
