Foundry shifts and margins
- Analysts say AI demand will soon represent about one-third of TSMC's business, reshaping foundry economics. - Intel is boosting foundry investment, hired Shawn Han, and increased equipment orders to pursue new customers. - Changes in supplier mix and capacity tighten the need for inventory-ageing by technology node and rigorous reserve reviews. ( )
Artificial intelligence chips are starting to redraw the economics of contract chipmaking, with Taiwan Semiconductor Manufacturing Co. pulling in more revenue per wafer as demand shifts to advanced nodes. (nextplatform.com) The basic business is simple: customers like Nvidia and Apple design chips, and foundries like TSMC and Intel sell factory capacity to make them. TSMC reported first-quarter 2026 revenue of $35.9 billion, gross margin of 66.2%, and second-quarter guidance of $39.0 billion to $40.2 billion. (investor.tsmc.com) The Next Platform estimated TSMC’s artificial intelligence chip revenue reached $33.38 billion in 2025 and said AI demand could account for about one-third of TSMC’s business in 2026. It also said TSMC processed a record 4.17 million 12-inch wafer equivalents in the first quarter of 2026. (nextplatform.com) That mix shift changes margins because a wafer made on a leading-edge process can carry far more value than one made on an older node. TSMC said on April 16 that it was raising its full-year 2026 revenue growth outlook to above 30% year over year as AI demand stayed strong. (investor.tsmc.com, cnbc.com) Intel is trying to use the same AI buildout to win more outside manufacturing customers. TrendForce reported on April 20 that Intel had increased chipmaking equipment orders by more than 50% from a year earlier, citing supply-chain reports tied to its foundry push. (trendforce.com) Intel also changed management. GovCon Wire reported on April 20 that Intel named former Samsung executive Shawn Han as senior vice president and general manager of foundry services, reporting to Naga Chandrasekaran, executive vice president and general manager of foundry technology and manufacturing. (govconwire.com) Intel has been telling customers its next manufacturing step, called 14A, is aimed at better performance and lower power than 18A. At Intel Foundry Direct Connect in April 2025, Intel said 14A should deliver a 15% to 20% performance gain and 25% to 35% lower power than 18A, while multiple customers planned test chips on the node. (trendforce.com, newsroom.intel.com) For finance teams, the manufacturing shift creates a quieter problem inside the balance sheet. When supplier mix and factory capacity move toward newer nodes, companies have to review inventory by technology generation because older wafers, tools, and materials can lose value faster if demand moves on. (investor.tsmc.com, nextplatform.com) That is why inventory-ageing by node and tougher reserve reviews matter more in 2026 than they did when smartphone and personal computer chips dominated fab loading. If AI keeps taking a larger share of leading-edge capacity, foundries will have more incentive to favor the newest, highest-yielding business and less room for mistakes on older inventory. (nextplatform.com, trendforce.com)
