Within the semiconductor industry, manufacturing memory chips may be the toughest task. Not technically tough -- manufacturing an advanced CPU or GPU is a lot more complicated than churning out DRAM or NAND chips. Tough because memory chips are largely a commodity. Supply and demand drives pricing, and bad downturns can push the bottom line of memory chip manufacturers deep into the red.
Getting out of memory chips for good
Intel (INTC -0.25%) has been working its way out of the memory chip business over the past few years. The company announced the sale of its NAND memory operations to SK Hynix in late 2020, exchanging its NAND component and wafer business, its SSD business, and a NAND manufacturing facility in China for $9 billion. The sale process is complicated and won't be fully completed until 2025.
In an interview earlier this year with Ben Thompson of Stratechery, Intel CEO Pat Gelsinger made it very clear how much he dislikes the memory business. "I never want to be in memory, you see I'm doing everything I can to exit our memory businesses in that regard," Gelsinger said.
NAND was a big part of Intel's memory business, but it wasn't the only part. Intel launched its Optane product family in 2017 built around a new type of memory technology, 3D XPoint, co-developed with Micron. 3D XPoint is non-volatile like NAND, meaning it retains data when power is cut off, but it's much faster.
Intel's Optane memory targeted multiple use cases. In PCs, Optane-based SSDs could be used in place of standard SSDs, providing improved read and write performance. And in servers, expensive DRAM could be replaced with cheaper Optane memory. Optane isn't as fast as DRAM, but for applications where that performance difference isn't very important, Optane offered a strong value proposition.
Optane seemed to have a lot of potential back in 2017 when it first launched, but Intel hasn't been able to turn it into a business that really works. Along with its second-quarter report, Intel announced that it was winding down the Optane business. That effectively kills 3D XPoint as a technology, since Micron had abandoned it in 2021.
Winding down the Optane business involves a substantial write-down of inventory. It makes little sense for any customer to buy more Optane memory when the underlying technology is being abandoned, so Intel is effectively stuck with a bunch of chips it can't sell. Intel wrote down its Optane inventory by $559 million in the second quarter, knocking down the bottom line.
Optane is out, but Intel has plenty of new products
Intel's core products are CPUs aimed at PCs and servers, but the company has been working on building out a host of new businesses over the past few years. Intel will soon launch graphics cards aimed at PC gamers and creators, it's already shipping its Blockscale ASIC for mining , and the company is investing billions in an effort to become a major player in the foundry market. Intel has announced 10 new revenue-producing product lines this year, and it has plans for more.
We can be sure that none of those new products will be memory. Instead, Intel will focus on products where it can differentiate itself. The company's entry in the graphics card market is particularly notable. While it will take time to build that business up, graphics cards are a multibillion-dollar opportunity for Intel in the long run.
It's important for Intel to expand beyond its core CPU businesses, but it's also important to recognize when a product simply isn't working. Optane never found enough traction, and Intel is right to cut the cord as it goes after better opportunities.