Micron Technology is the biggest US maker of memory chips, and it is sprinting to add capacity before the nastiest supply crunch the sector has seen in more than 40 years.
In its hometown, Micron is ploughing $50 billion into more than doubling its 450-acre campus, with two new fabs going up on the site.
The first plant should begin production of its inaugural silicon wafers in mid-2027, producing DRAM that feeds high-bandwidth memory (HBM), the material modern AI computing now demands.
Micron reckons both Boise plants will be in production by the end of 2028, assuming reality does not throw a spanner at the schedule.
Each fab comes in at 600,000 square feet, which is more than 10 football fields, and they are being billed as some of the biggest clean rooms ever built in America.
Engineers have already chewed through more than seven million pounds of dynamite, while a small city of trailers houses crews working round the clock.
Micron says each Boise fab will gulp 70,000 tons of steel, close to what went into the Golden Gate Bridge, plus 300,000 cubic yards of concrete, enough for four Empire State Buildings.
Boise is only one front; Micron is breaking ground near Syracuse on a $100 billion fab complex that New York State is calling its largest private investment ever.
Late last year, Micron also announced a $9.6 billion fab investment in Hiroshima, Japan, while SK Hynix said in January it would build a $13 billion fab in South Korea on top of a $4 billion complex in Indiana.
The frenzy is being driven by AI, as large language models balloon and outfits like OpenAI, Oracle, xAI and Anthropic talk up trillions of dollars worth of data centres.
Micron, vice president Scott Gatzemeier said: “I’ve been here for 28 years, and I’ve never seen anything so disruptive as AI. As we moved from training to inference, the amount of data required exploded, and we didn’t have enough clean-room capacity to meet demand. We realised we had a huge problem.”
That shortage has triggered a gold rush for Micron, SK Hynix, and Samsung, with Micron shares rising more than sixfold since April last year to about $414, valuing the company at nearly half a trillion dollars.
As Micron pivots from simpler mobile memory to fatter-margin data-centre HBM, gross margins have ripped from 18.5 per cent in early 2024 to 56 per cent in its most recent quarter.
Micron says it expects gross margin in the current quarter to hit 68 per cent, drifting towards the more than 73 per cent that Nvidia pulls on its flagship graphics processing units.
For decades, memory was treated as a commodity, cheaper and easier than the fancy AI chips, which is why this profit spike has caught plenty of people leaning the wrong way.
Micron chief financial officer Mark Murphy said: “Our business is on an extraordinary trajectory,” and he told an investor conference on 16 February 2026 that Micron can only meet about one-half to two-thirds of demand for some key customers.
“On the supply side, we are doing everything we can to add capacity,” and “But there is no easy or fast way to get that done,” Murphy said.
Micron’s chief business officer, Sumit Sadana said HBM demand went “through the roof” between August and October as developers, cloud firms, and hyperscalers continued to announce large-scale projects.
“Memory has gone from being a system component to being a strategic asset,” and “The promise of AI is all ahead of us,” Sadana said.
Micron was already building its first Boise fab, ID1, but it decided to accelerate ID2 as data centre construction picked up and the order book became more robust.