Nvidia Corp. supplier Micron Technology Inc. said an ongoing memory chip shortage has accelerated over the past quarter and reiterated that the crunch will last beyond this year due to a surge in demand for high-end semiconductors required for AI infrastructure.

“The shortage we are seeing is really unprecedented,” Micron Executive Vice President of Operations Manish Bhatia said in an interview shortly after the chipmaker held a groundbreaking ceremony for a $100 billion production site outside Syracuse, New York, on Friday, amplifying a similar forecast the company provided in December.

High-bandwidth memory required to make artificial intelligence accelerators is “consuming so much of the available capacity across the industry that it’s leaving a tremendous shortage for the conventional side of the industry, for phones or PCs,” Bhatia said.

He added that PC and smartphone makers have joined the queue to try to lock up memory chips after 2026, while autonomous vehicles and humanoid robots will further drive demand for those components.

On Friday, Chinese media outlet Jiemian reported that major Chinese smartphone makers including Xiaomi Corp., Oppo and Shenzhen Transsion Holdings Co. are trimming their shipment targets for 2026 due to rising memory costs, with Oppo cutting its forecast by as much as 20%. All three did not respond to requests for comment.

Global smartphone shipments may decline 2.1% this year as a shortage of memory chips drives up costs and squeezes production, industry tracker Counterpoint Research estimated in December. PC makers including Dell Technologies Inc. have also warned they are likely to be affected by the ongoing shortage.

The big three of the global memory chip industry — Micron, SK Hynix Inc. and Samsung Electronics Co. — saw their share prices surge in 2025 thanks to the AI boom. SK Hynix said it has sold out its entire chip slate in 2026, while Micron has said its AI memory semiconductors are also fully booked this year.

To prioritize supplying strategic enterprise customers including Nvidia, Micron said in December it will end its popular Crucial-branded consumer memory business. The AI industry’s insatiable appetite for memory chips is also adding urgency to Micron’s manufacturing expansion in both the US and Asia.

On Saturday, Micron announced plans to pay $1.8 billion for a site with an existing plant in Taiwan, which serves as a key production hub for the Boise, Idaho-based chipmaker.

That move significantly shortens the time for Micron to bring a new factory online. The company says it will start meaningful output of DRAM wafers in the second half of 2027.

DRAM provides the operating environment for complex processors from Nvidia and Intel Corp. to make calculations and is at the heart of high-bandwidth memory required for AI accelerators to work optimally.

“What we’ll be doing at our Asian sites is continuing to transition to the next generation of technology,” Bhatia said during the Friday interview.

New wafer capacity, on the other hand, will take place almost entirely in the US, he added.

Micron’s $100 billion project near Syracuse is slated to host four DRAM fabs — each about the size of 10 football fields. First wafers are coming out of the production lines by 2030.

The US chipmaker is adding two fabs’ worth of capacity in Boise alongside existing research and development facilities there. The first Idaho fab is scheduled to begin production in 2027, and a second factory is being planned. It is also modernizing and expanding an existing manufacturing facility in Virginia.

The plans are all part of the company’s commitment to bring 40% of its DRAM manufacturing onto US soil, a goal enabled by a $6.2 billion Chips Act award the company clinched in 2024, and the ability to tap into a now-35% tax credit while construction is ongoing.