TAIPEI, TAIWAN — In a quarterly earnings report that has left analysts simultaneously exhilarated and quietly concerned about the structural foundations of modern civilization, Taiwan Semiconductor Manufacturing Company reported a 58% surge in first-quarter profits this week, driven almost entirely by the world’s apparently inexhaustible hunger for artificial intelligence chips.
TSMC, which manufactures the tiny silicon wafers that power everything from your phone to the AI assistant currently writing your work emails, said demand is so far ahead of supply that the company has essentially run out of ways to describe how busy it is. The firm now forecasts full-year revenue growth of more than 30% — a number that has caused several TSMC shareholders to cry tears they later had an AI model analyze for emotional patterns.
“We are operating at full capacity,” said a TSMC spokesperson, who then paused, looked into the distance, and added, “all of our capacity. All of it. Constantly.” The company’s manufacturing plants in Taiwan, Arizona, and Japan are running around the clock, churning out chips at a rate that still cannot meet demand from major AI developers who keep ordering more as if they are selecting items from a restaurant menu that has no prices listed.
“We used to get two or three major chip orders per quarter,” said TSMC capacity planning analyst Kevin Chu, reached by phone. “Now I get three before breakfast. I’ve stopped answering the phone during meals. I’ve stopped eating meals.”
The news comes alongside a separate but equally alarming statistic: AI systems now consume more than 10% of all electricity generated in the United States. American utilities are planning to spend approximately $1.4 trillion over the next five years to keep up with power demand — a figure that represents either an extraordinary investment in the future or, depending on your perspective, a fairly aggressive commitment to making sure the AI that recommends movies has enough electricity to do so.
OpenAI, one of TSMC’s largest customers, separately reported surpassing $25 billion in annualized revenue this week and is exploring a potential public listing. Analysts expect the IPO to be the largest in history, described by Wall Street insiders as “too big to properly understand, but definitely investing in it anyway.”
“The chip shortage is real, it is serious, and it shows no signs of easing,” said tech analyst Priya Nair of Redwood Capital. “The good news is, if things get bad enough, we’ll probably have an AI figure out how to fix it. The bad news is that AI will need chips to do so.”
Anthropic’s Model Context Protocol, a technical standard for connecting AI systems to external tools, crossed 97 million installs in March — a figure that means something to approximately 12% of people who read it and everything to the remaining 88% who will simply nod and move on.
In related news, scientists at a research institution whose name will be redacted to protect its funding have announced a breakthrough computing approach that could reduce AI energy consumption by up to 100 times. These scientists have been asked, politely but firmly, to please keep it down — the data centers are trying to concentrate.
Globe News Daily editorial note: We asked our AI editorial assistant to weigh in on this story. It said “More chips, please.” We are not sure if it was speaking metaphorically.
Leave a Reply