TSMC’s $100 billion U.S. commitment could calm Taiwan tensions
Welcome to AI Decoded, Fast Company’s weekly newsletter that breaks down the most important news in the world of AI. I’m Mark Sullivan, a senior writer at Fast Company, covering emerging tech, AI, and tech policy.
This week, I’m focusing on the implications of TSMC’s surprising $100 billion commitment to building advanced chip fabs in Arizona. I also look at a report card on the Pentagon’s tech-buying habits, and at a new training-focused AI startup.
Sign up to receive this newsletter every week via email here. And if you have comments on this issue and/or ideas for future ones, drop me a line at [email protected], and follow me on X (formerly Twitter) @thesullivan.?
TSMC's new $100 billion U.S. chip investment could ease Taiwan tensions
TSMC (Taiwan Semiconductor Manufacturing Company) said it would invest another $100 billion in U.S.operations over the next ten years. This will go toward building three new fabrication plants, two advanced packaging facilities and an R&D center, the company said.?
TSMC had already committed $65 billion to building three new chip fabrication facilities near Phoenix, Arizona. One of these fabs, which makes 4-nanometer process chips for smartphones, went into production last year; the other two, which will fab 2-nanometer process chips used in AI acceleration, remain under construction.?
“This move underscores TSMC’s dedication to supporting its customers, including America’s leading AI and technology innovation companies such as Apple, NVIDIA, AMD, Broadcom, and Qualcomm,” the company said in a press release. The U.S. government currently collects no tariffs on imports from Taiwan, but the Trump administration is said to be considering 100% tariffs on Taiwanese semiconductors and electronic devices containing the chips. TSMC’s investment promise could go a long way toward forestalling any such plans.
TSMC isn’t exactly a household name in the U.S., but it plays a huge and growing role in the global economy. As AI moves further into business operations and consumer products, the demand for the powerful and sophisticated graphics processing units (GPUs) that power AI models will continue to grow. U.S.-based Nvidia, which now supplies almost all of the GPUs used for generative AI models, relies heavily on TSMC to fabricate its most powerful chips.?
Click here to read more about TSMC’s new chip investment.
Reagan Foundation gives Pentagon a “D” for defense modernization???
Every year the Ronald Reagan Foundation issues a report card evaluating how well the Pentagon sources the nation’s best technologies. This year’s report finds that while the U.S. remains a global leader in technological innovation—particularly in artificial intelligence, which is playing a growing role in warfare—the Department of Defense (DOD) continues to struggle with modernization.
“The U.S. remains a global leader in innovation, setting technological standards worldwide and excelling in research, particularly in artificial intelligence,” the report states. But the report gives the Pentagon low marks (a “D”) for modernizing defense systems, with the authors citing concerns about the DOD’s inability when it comes to integrating new capabilities into production. And while commercial technology adoption has increased in select areas, such as space communications, progress in many other sectors remains stagnant.??
Click here to read more about the Reagan Foundation’s Pentagon assessment.
Ex-Google engineering VP Anna Patterson unveils her new AI training company
A wave of AI infrastructure companies has sprung up to help enterprises (especially ones without teams of PhDs) more easily build and deploy AI models. Anna Patterson, an ex-Google VP of Engineering and founder of Gradient Ventures, is now bringing her new AI training-focused infrastructure company out of stealth. The company, Ceramic.ai, is made up of nine engineers and has so far raised $12 million in seed funding from New Enterprise Associates and others.?
Enterprises that decide to build AI infrastructure from scratch often run into problems and delays related to technical complexity, Patterson says. “With AI infrastructure there’s a real dichotomy between what is available to most enterprises, and what the biggest AI labs are using,” Patterson tells Fast Company. This can be especially taxing with training and fine-tuning models, which involves both science and art.
Ceramic’s training methods let models get the most out of the available training data and computing power (GPU time). The company organizes training data by topic before introducing it to the model. Ceramic can then help the enterprise customer train its model with its own proprietary domain knowledge.?
Click here to read more about Ceramic.
More AI coverage from Fast Company:?
Want exclusive reporting and trend analysis on technology, business innovation, future of work, and design? Sign up for Fast Company Premium.