Intel's 18,000+ job cuts may leave ailing telco unit high and dry

　　The smallish telecom sector was never the main concern for Intel, the giant US chipmaker that made most of its $54.2 billion in revenues last year by selling chips for PCs and data center servers. But it barely got a mention on this week's earnings call with analysts. Intel, it seems, would rather forget about a market that is shriveling with age and remains heavily resistant to Intel's mainstream products.

　　David Zinsner, Intel's chief financial officer, had to insert an "outside of the telco space" caveat when he said the network and edge group (NEX), responsible for just 11% of Intel's revenues last year, was "starting to recover." That came after a comment about "excluding the previously discussed inventory digestion impacting the telco market." Photoshop telco out of the picture and NEX revenues grew 10% year-over-year for the first half. Leave it in and sales dropped 7%, to $2.7 billion.

　　Intel, of course, has much bigger worries today. For some time, it has been losing market share to AMD, a rival that bases its chip designs on the same x86 architecture used by Intel. A more recent and arguably more troubling threat comes from licensees of Arm, a UK-based company boasting an alternative, more power-efficient architecture. They range from Amazon, the hyperscaler, to Qualcomm, a designer of chips for mobile phones. Regardless, most data center investments are currently going into chips for artificial intelligence. And that still means Nvidia, not Intel.

　　Meanwhile, a strategic pivot under CEO Pat Gelsinger to set up a foundry, manufacturing chips designed by others, is proving costly. Combined with investments aimed at reestablishing Intel as a technology leader, it showed in the latest second-quarter results. Intel posted a net loss of $1.6 billion, compared with a profit of $1.5 billion a year earlier, even though headline sales were down just 1%, to around $12.8 billion.

　　How many job cuts?

　　The response is a massive cost-cutting exercise that will scrap 15% of roles, said Intel in its latest earnings release. Press reports generally assume Intel employs around 110,000 people, but the figure shown in its filing with the Securities and Exchange Commission (SEC) was 124,800 in December, down from 131,900 a year before. Unless Intel spent the first six months of this year cutting jobs, then, a 15% reduction could translate to the loss of 18,720 employees.

　　The goal is to slash various operational costs (research and development plus marketing, general and administrative expenses) to about $20 billion this year and $17.5 billion in 2025, "with further reductions planned in 2026," said Intel. That $17.5 billion would be about $4.2 billion less than Intel booked for these expenses in 2023, according to its last annual SEC filing, and a reduction of $7 billion compared with the figure for 2022.

　　Gross capital expenditure is also expected to fall by a huge 20% compared with earlier projections. That would lower investments this year to a range of between $25 billion and $27 billion. Next year, Intel plans to invest between $20 billion and $23 billion. The immediate issue for some analysts is whether this imperils the foundry strategy. It's unchanged, answered Gelsinger, before saying that "until we have committed orders, we're going to be modest on how much equipment we put against the shells and the sites that we have in place."

　　More broadly, it is hard to shake off the feeling that Gelsinger has simply given Intel too much to do in a short space of time – mount an AI challenge to Nvidia, defend the PC and server businesses from Arm rivals, instil confidence Intel can meet its ambitious product roadmaps (something it has failed to do in the past) and, to top it all off, set up a whole new foundry.

　　Telco trauma

　　Cuts that affect research and development and take such a huge number of people out of the business could obviously hinder Intel in any of these fields. But they could also have ramifications for Intel's activities in markets that are less critical, as it focuses on the strategically important stuff. And there have already been rumors that Intel's commitment to telecom is wavering.

　　It would hardly be a surprise. The latest gloomy news from the world of telecom was this week supplied by Samsung, a maker of 5G networks and Intel customer, which reported a 21% year-over-year drop in second-quarter sales at its networks business, to 740 billion South Korean won (US$541 million). That decline followed continued sales disappointments for Ericsson and Nokia.

　　Intel wants 5G operators to switch from custom silicon designed by telco vendors to its general-purpose processors (GPPs). But there has been very limited adoption of them in the radio access network (RAN) so far. Many telco executives remain unconvinced GPPs, especially based on x86, can measure up. Arguments about using the same platforms for multiple needs look spurious when most RAN compute is at the mast site, where it cannot realistically be shared with anything else.

　　Of the big three 5G kit vendors, only Ericsson says Intel is a good option for Layer 1, the category of most demanding RAN software. Huawei and Nokia remain vehemently opposed to it in this area. And most of Ericsson's 5G products today are based on its own custom silicon, not Intel's GPPs. With all its other distractions, the surprise is that Intel still bothers.