Earlier this year, Intel had a new CEO and launched a new business plan that would open its foundries to other chip design companies, just like the operation of TSMC and Samsung Semiconductor. At its “Intel Accelerated” event today, the company presented a roadmap for its future as a for-hire foundry. Along with the future of ever-smaller process nodes, the company also announced that it has named one of the world’s largest chip designers, Qualcomm, as a future customer of the foundry.
As part of its entry into the foundry market, Intel will begin to name its process nodes more like its rivals. The numbers of process nodes used for chips such as “5nm” started out as a measure of the size of transistors, but eventually marketers grabbed them and companies started cheating their numbers to appear more advanced. Intel says its new naming scheme will better align with how TSMC and Samsung talk about their foundry technologies. Gone are the days of “Intel 10nm Enhanced Super Fine” – instead the node is called “Intel 7”. It is expected to have a density comparable to TSMC and Samsung 7nm nodes and will be ready for production in the first quarter of 2022 (TSMC and Samsung are currently releasing “5nm” products). “Intel 4,” which Intel previously called “7nm,” would now be equivalent to TSMC and Samsung’s 4nm node, and it will begin manufacturing products in 2023.
If you’re wondering what happens when we run out of “nm” numbers, Intel’s selling point for this is the “Angstrom” era, a unit of measurement that’s a tenth of a nanometer. In 2024, the company wants to develop the “Intel 20A” process node (hence a “2nm” equivalent, but Intel called this node “5nm” before, but remember these are marketing numbers and not really units of measure). In early 2025, the company will be working on “Intel 18A”.
Part of the name change to “Intel 20A” instead of “2nm” appears to be due to the fact that this process node will include major architectural changes for Intel’s chips. For years, the company has used FinFET transistors, but for the Intel 20A, the company will switch to a gate-all-around (GAA) design that it will call “RibbonFET”. FinFETs would increase the current capacity of the channel by adding more fins and therefore more horizontal space. But GAA designs allow chipmakers to stack multiple channels on top of each other, making current capacity a vertical issue and increasing chip density. Intel 20A will also introduce “PowerVias”, a new chip design method that will place the power supply at the back of the chip. This design would put the power layer at the bottom of the chip, then the transistors, then the communication wires. Traditional chip design places the transistors at the bottom, and the upper signal and power layers have to intertwine to reach the transistor layer.
If Intel does manage to stick to its roadmap, it should be able to count Qualcomm as an interested customer. President and CEO Cristiano Amon expressed interest in the ’20A’ node saying, “Qualcomm is excited about the breakthrough RibbonFET and PowerVia technologies coming in Intel 20A. We are also delighted to have another cutting edge foundry partner activated by IFS [Intel Foundry Services] it will help the US factory-less industry get their products to a manufacturing site ashore. “
Today Qualcomm manufactures a lot of chips and is a customer of both TSMC and Samsung. The two companies compete regularly for each new design in the Qualcomm lineup, with industry reports often describing a realistic horse race where one beats the other. Whether or not Intel is in the mix for most of these foundry battles depends on its ability to catch up with TSMC and Samsung. At least now Qualcomm is giving Intel a place in the race, instead of the relevance of smartphones.
List image by Intel