Last week Light Reading reported that Intel is considering selling 5G-related company assets to rid itself of a troublesome business and raise cash as it seeks to recover from years of stumbling.
Two-Phase Havoc
Intel’s 5G entry wreaked havoc in two phases. In the honeymoon phase a decade ago, customers could see Intel’s processors getting faster and process technology improving every year. Rivals focused on embedded processors weren’t as consistent and relied on lesser foundry technology. They certainly didn’t have the resources to keep up with Intel, which could fund R&D from its lucrative PC and server businesses. Appearing to be a stronger supplier, Intel landed major networking and 5G wins at Cisco, Ericsson, Nokia, and others. With the notable exception of Cavium (later acquired by Marvell), rivals either folded their tents or faded away.
In the second phase, Intel’s consistent execution ended. The company failed to deliver key products on time to Ericsson, Nokia, and other 5G equipment makers. In response, Intel focused on its biggest customer, Ericsson, delivering the Snow Ridge CPU-NPU chip (where NPU means network processor) and custom silicon. Other OEMs had to fend for themselves as the 5G market window narrowed and chip-supply options dwindled. But now Ericsson is at risk, too, as Intel has continued to falter.
Come Back Down from This Cloud
Meanwhile, Intel had helped fuel interest in software-defined systems (SDx), backing initiatives such as network-function virtualization (NFV), software-defined networking (SDN), and 5G virtual radio-access networks (vRAN). Such systems supposedly would be flexible and cost less because they’re based on white-box hardware (generic Xeon servers) instead of bespoke networking equipment. Intel’s ever-faster microprocessors would enable them to meet performance requirements that heretofore demanded specialized silicon.
In the SDx honeymoon phases, the many drawbacks to SDx were ignored. Although NFV and SDN have achieved a degree of success, vRAN continues to struggle. Its long-term viability is questionable, and Intel’s situation exacerbates vRAN’s uncertainty. Here, too, Ericsson is exposed, having developed Intel-based vRAN alternatives to its conventional designs.
The OEM’s influence on Intel can be seen in recent Xeon designs, which speed up 5G and network processing by offloading forward error correction (FEC) and dynamic load balancing (DLB) from the CPUs. (It turns out that generic servers can’t do everything in software.) It’s unlikely that Intel has the resources or motivation to make similar design changes to gain share in smaller markets like networking and telecom when its core business must be fixed and expenses trimmed.
Technology Nihilism?
As the Light Reading article notes, Intel’s poor health and market control are a toxic combination. The result will be most acutely felt in wireless telecom. Innovation will slow without a rich semiconductor supplier base. But, with no application driving faster wireless networks, a more devastating thought for chipmakers and OEMs is that innovation and upgrades are unneeded.