TSMC Sees AI Surge

Taiwan Semiconductor Manufacturing said on April 16 that second-quarter revenue could reach $40.2 billion as demand for artificial intelligence chips keeps climbing. (investor.tsmc.com) The company reported first-quarter revenue of $35.9 billion, gross margin of 66.2%, and operating margin of 58.1%. It also guided second-quarter revenue to $39.0 billion to $40.2 billion, up from its prior quarter guidance range of $34.6 billion to $35.8 billion. (investor.tsmc.com) TSMC said first-quarter net income rose 58% from a year earlier to NT$572.48 billion, while revenue reached NT$1.134 trillion. Chief executive C.C. Wei said on the earnings call that “AI-related demand continues to be extremely robust” and that the company’s conviction in a multi-year AI trend remains high. (cnbc.com; money.usnews.com) A foundry is a chip factory that manufactures processors designed by other companies, and TSMC is the main producer for many of the most advanced AI chips. Reuters reported that TSMC is expanding 3-nanometer capacity in Taiwan, the United States and Japan as customers push for more supply in 2027 and 2028. (money.usnews.com) That demand sits at the center of Nvidia chief executive Jensen Huang’s latest warning about China’s AI stack. In an April 15 interview on Dwarkesh Patel’s podcast, Huang discussed chip sales to China, Nvidia’s supply chain moat and how the company makes investments across the AI market. (podcasts.apple.com; www.dwarkesh.com) Reports published on April 16 and April 17 said Huang warned that a DeepSeek V4 model running entirely on Huawei Ascend chips would be bad for U.S. interests, because it would show a leading Chinese model can run on a Chinese chip stack instead of American hardware. Those reports cite the Dwarkesh interview and frame the remark as part of Huang’s argument that export controls cannot freeze China’s AI progress. (msn.com; techflowpost.com; huaweicentral.com) Huawei’s Ascend chips are China’s best-known alternative to Nvidia’s AI processors, and DeepSeek has become a closely watched Chinese model developer since its earlier releases drew global attention in 2025. Separate recent reports said DeepSeek’s next model is expected in late April and is being tested for Huawei-based deployment, though those launch details remain reports, not a public DeepSeek announcement. (msn.com; msn.com) Huang also said Nvidia invests broadly instead of trying to guess which single AI company will win. Coverage of the interview said he argued that Nvidia’s role is to support the wider ecosystem, a strategy that fits a market where chip suppliers, model developers and cloud companies are all scaling at once. (msn.com; tekedia.com) TSMC’s forecast and Huang’s comments point to the same pressure point: more AI models require more computing power, and more computing power requires more advanced chip manufacturing. On April 16, TSMC also raised its 2026 revenue growth outlook to more than 30% in U.S. dollar terms and said capital spending would land at the high end of its $52 billion to $56 billion range. (money.usnews.com; cnbc.com) The next test is whether TSMC can add enough leading-edge capacity while U.S.-China technology restrictions keep pushing customers toward separate supply chains. For now, the clearest number in the market is TSMC’s own forecast: as much as $40.2 billion in revenue for one quarter. (investor.tsmc.com; money.usnews.com)