There was a time when Intel (INTC 1.77%) was dead set on broadening its reach in the data center. The company is the longtime leader when it comes to server CPUs, although rival AMD now has a stronger product lineup. But there's a lot more that goes into a data center than CPUs, and Intel saw that as an opportunity.
Intel acquired Barefoot Networks in 2019 in an attempt to expand its presence in the networking portion of the data center. Barefoot Networks designed Ethernet switch chips and software, with a particular focus on programmability and flexibility. "[T]he addition of Barefoot Networks will support our focus on end-to-end cloud networking and infrastructure leadership," read the Intel news release announcing the deal.
Intel's Ethernet switch chip efforts were eventually put into the company's Network and Edge Group (NEX), which generated $8.9 billion of revenue in 2022. This segment includes a wide range of products, and Intel still views it as an important piece of its transformation efforts.
Getting out of Ethernet switches
While NEX will remain a significant part of Intel, the company announced alongside its fourth-quarter report that it was ending all future investments in its network switching product line. This means that development of the Tofino series of chips acquired with Barefoot Networks is being halted. Intel will continue to support existing products and customers.
The shutdown of the switching business marks the seventh business that Intel has exited since CEO Pat Gelsinger took the helm in early 2021. Those exits have produced more than $1.5 billion in savings, which will prove critical as Intel navigates an unprecedented downturn in the PC market.
Other notable exits include essentially the entire memory business. Intel has sold off its flash memory and solid-state drive businesses to SK Hynix in a complicated deal that was announced in 2020. While that transaction won't be fully complete until 2025, Intel has already transferred its SSD business and its NAND memory manufacturing facility in China. Getting out of memory looks like a smart move. While demand for PC and server CPUs is sinking, the memory chip industry is suffering an even steeper decline.
The other piece of Intel's memory business was Optane, its brand name for 3D XPoint-based products. 3D XPoint was a new type of memory Intel co-developed with Micron, and it was certainly promising. It was non-volatile like NAND but significantly faster, capable of being used in ultra-fast SSDs or even replacing DRAM in some cases. But the technology never gained enough traction. Intel killed the business last year, writing off $559 million of inventory.
Focusing on the best opportunities
Under Gelsinger, Intel is focused on areas where it already has or can conceivably gain a durable competitive advantage. PC and server CPUs fit the bill. So does the company's efforts to break into the discrete graphics card market with its Arc graphics cards. And Intel's push into the foundry business makes sense as well, given that the number of companies capable of manufacturing chips using bleeding-edge technology is tiny, limited by the intense capital requirements and technical know-how.
Memory never made a ton of sense. For the most part, memory chips are commodity products, with pricing a function of supply and demand. 3D XPoint made sense in theory, but it just didn't work as a business in practice. Getting into the Ethernet switching chip business didn't seem like a bad idea in 2019, but Intel has decided that further pursuing that market isn't worth the investment.
Intel is aiming to slash costs by $3 billion this year, with total cost cuts reaching as much as $10 billion by the end of 2025. Getting out of non-core businesses is one way to drive cost savings, and the Ethernet switch chip business appears to be a casualty of Intel's aggressive efforts to bring its cost structure under control.
