12-03-2021 | | By Robin Mitchell
Intel is one of the last major giants in the semiconductor industry who remains stubborn in their production method; they must make their own devices. What challenges has Intel faced, is Intel about to undergo a major change, and what future possibilities could foundry specific companies help to usher in?
Time and time again, Intel's recent struggles have been in the news in one form or another. Intel, one of the worlds biggest designers and manufacturers of CPUs, used to be at the forefront of semiconductor manufacture. Intel's technology would leave other manufacturers in awe, and Intel's competitors would generally be behind Intel in performance.
But the age-old practice of Intel joining node to architecture has left Intel struggling for the past few years. In essence, the architecture that Intel design refers to the CPU capabilities and design, whereas the node is how small a transistor Intel can manufacture (such as 14nm). Since Intel has historically released new architectures with the release of nodes, if any one of the two suffers a delay, the entire next-generation product line suffers a delay.
Intel has never delayed on their architecture, but the past few years have seen Intel struggling to reduce their transistors' size. Intel has been unable to release their next-generation devices, and competitors such as AMD have taken the lead.
Semiconductor companies of the past all designed and manufactured their own semiconductor devices. During the early development of semiconductors, money was found in the manufacturing processes just as much as the design of new components using semiconductors. Furthermore, when semiconductor technology was young, any company that could reduce the size of their transistors, or improve characteristics could have a massive competitive advantage over other foundries.
However, as semiconductor research became increasingly expensive and specialised, designers of semiconductor devices began to shift their focus to the design side, and thus the fabless company was born. Like PCBs' manufacturing, designers are not interested in how the transistors or oxide layers are made. Instead, designers only care about how their design will function.
To date, many semiconductor companies are fabless including AMD, Acer, Broadcom, Xilinx, and even Apple who now design their own ICs. Companies such as Texas Instruments and NXP have their own foundries for creating specific logic and analogue semiconductors such as DC/DC controllers, op-amps, and logic gate. However, large complex semiconductors are increasingly being designed by fabless companies and outsourced to foundries such as TSMC.
Intel, however, is quickly becoming the exception to the norm; they both design and manufacture their own device. While this business model may have been practical historically, the increasing difficulty of reducing transistor size shows that companies need to pick their area of expertise and stick to it. In the case of Intel, it might not be a bad idea to sell off the foundry business and become fabless.
A recent report discusses the idea of Intel becoming fabless, and such a move could prove highly beneficiary in several regards. To start, the initial selling off of Intel foundries could see a net profit of $30 billion. The resulting sale would surge Intel stock's value, and Intel could then focus on getting designs out faster by no longer combining node with design. Intel would also be free to solely focus on processor design while the foundry business would be purchased by a company (such as TSMC), which already has research into the field with better results.
The improvements in semiconductor technology and improvements in technology in general could see some major changes to the semiconductor industry. While high-tech semiconductor foundries would be reserved for Intel, AMD, and Xilinx, foundries of more basic technology could begin to crop up.
Such foundries could offer services similar to PCB fabrication plants, allowing customers to send in designs to be manufactured. While such services exist for the semiconductor industry, they are costly and out of reach from most projects.
The use of devices such as FPGAs and CPLDs currently provides designers with all the custom silicon they require. Still, as more companies become fabless, there is a chance that demand for custom silicon will increase. If economies of scale also apply, individual designers could become customers to a very lucrative industry. In fact, the ability to create microchips at home has already been demonstrated.
Custom silicon provides designers with a lot of opportunity as individual transistors can be custom crafter for extremely specific purposes. While this may not be useful in the digital processing world, the analogue world is a very different game. Maybe one day, you will be able to order a custom semiconductor at the price of JLCPCB or PCBCART!