15-01-2021 | | By Robin Mitchell
Recently, a hedge fund company who has purchased a large stake in Intel has urged the company to go the way of other semiconductor developers and go fabless. Why has the hedge fund made this recommendation, what do competitors of Intel do, and should Intel go fabless?
Recently, the hedge fund company Third Point LLC has purchased a substantial stake in Intel, and almost immediately after called on Intel to become a fabless company. Such a move would see Intel sell its foundries as separate businesses, and fully turn its attention to designing CPUs and other associated products. Intel has responded saying that it respects all its shareholders' wishes and looks forward to discussing with Third Point LLC on how its future could be shaped.
According to Third Point LLC, Intel has seen a continuous decrease in market share and has lost more than $60 billion of market capitalization over the past years. Furthermore, Third Point LLC also mentions “failed acquisitions” made by Intel which suggests that some companies Intel has been purchasing have not performed as expected. With AMD competitors now taking the lead in CPU products, Intel may now have to look towards going fabless if it is to remain competitive.
The CEO of Intel, Bob Swan, made a statement back in October 2020 mentioning that Intel would decide to either stick with 7nm technology or drop it by January 2021. Now that January 2021 is here, the question remains; will Intel go fabless?
Intel is a company that designs and manufactures semiconductor products, while many of its competitors only design semiconductor products. Under normal circumstances, this helps Intel not only maximise profits but also allows Intel products to utilise their own process fully. This has kept Intel at the forefront of CPU technology and has historically been renowned for their quality and capabilities.
However, using in-house manufacturing created a culture whereby Intel engineers would release new CPU architectures with reductions in silicon technology (referred to as a node). Therefore, when Intel reduced the node from 100nm to 50nm, it also updated its product line. However, as silicon technology approached the 10nm mark, Intel faced issues with its process (due to the difficulty in feature size reduction), which stalled the release of newer products.
During this time, competitors such as AMD could catch up and surpass Intel releasing CPU products that contained newer features. Intel has faced multiple node reduction issues to make matters worse and has now had to turn to other silicon foundries (such as TSMC) to get its latest products fabricated.
While there are many semiconductor manufacturers on the market, it appears that becoming fabless is the way to go. Not only has Intel suffered problems with its own manufacturing methods, but some of its customers are also turning to the development of their own chips.
For example, Apple has recently launched the M1, a totally custom device designed by Apple with their needs. The use of ARM cores allow Apple to use pre-designed components, and the ability to order their manufacturer from any foundry provides greater flexibility in production.
While AMD used to manufacture silicon devices, they became a fabless company more than a decade ago. Many other popular names in the semiconductor industry such as Xilinx and Nvidia are also fabless companies. If Intel were to become fabless, it could remove its dependency on node improvements, focus on CPU design, and catch up with the competition.
However, Intel becoming fabless would be no small decision; the company has historically led semiconductor manufacture. Intel selling off its fabrication business may send a message to the industry that it cannot keep up with other manufacturers, and this could also cast doubt over their CPU products.
The world is moving towards custom silicon devices, and the development of FPGAs and ASICs allows smaller companies to design their own processors and SoCs. Furthermore, the increasing use of ARM and RISC in desktop environments may see CISC processors slowly fade away as specialised functions (such as AI and cryptography) are moved to dedicated hardware circuitry.