TSMC Strength, Nvidia Design AI
TSMC reported record quarterly revenue above NT$1 trillion and heads into Q1 results under heavy expectations tied to AI demand. At the same time Nvidia said AI can compress a GPU-design job that once took eight engineers ten months into an overnight task, pointing to AI reshaping both demand and the chip‑design process. (tradingkey.com, tomshardware.com)
Taiwan Semiconductor Manufacturing Company just posted quarterly revenue above NT$1.13 trillion, days before its April 16 earnings call, as demand for artificial intelligence chips keeps climbing. (tsmc.com) The company said March 2026 revenue reached about NT$415.19 billion, up 30.7% from February and 45.2% from a year earlier. Revenue for January through March totaled NT$1.1341 trillion, up 35.1% year over year. (tsmc.com) Taiwan Semiconductor Manufacturing Company’s investor calendar lists its first-quarter 2026 earnings conference for Thursday, April 16, at 2:00 p.m. Taiwan time, or 2:00 a.m. Eastern Time. Analysts surveyed by London Stock Exchange Group had expected about NT$1.12 trillion in first-quarter revenue. (tsmc.com, cnbc.com) A foundry is the factory side of the chip business: companies like Nvidia design processors, and Taiwan Semiconductor Manufacturing Company manufactures them. That division has turned Taiwan Semiconductor Manufacturing Company into the main producer of the advanced chips used in the current buildout of artificial intelligence servers. (tsmc.com, cnbc.com) At the same time, Nvidia says artificial intelligence is changing how chips get designed inside the industry, not just how many get ordered. Tom’s Hardware reported this week that Nvidia chief scientist Bill Dally said one design task that once took eight engineers 10 months can now be done overnight on a single graphics processor. (tomshardware.com) The task is called porting a standard cell library, which is the catalog of tiny prebuilt logic blocks engineers use to assemble a chip. Tom’s Hardware said Dally described a library of roughly 2,500 to 3,000 cells, and said Nvidia’s reinforcement-learning tool, called NB-Cell, now handles that work overnight on one graphics processor. (tech.yahoo.com, tomshardware.com) Dally also said Nvidia is still “a long way” from having artificial intelligence design whole chips without human input. The current use is narrower: speeding up labor-heavy steps as chip teams move designs to each new manufacturing process. (tomshardware.com) That puts the two developments on the same track. Artificial intelligence is pushing more chip demand toward Taiwan Semiconductor Manufacturing Company, while also being used by Nvidia to cut the time needed to prepare designs for the next generation of factories. (tsmc.com, tomshardware.com)
