Intel IDM 2.0 – The name game.

The end of silicon is now on the distant horizon.

• As part of its IDM2.0 strategy, Intel is changing its manufacturing node naming system which will bring it more into line with TSMC and Samsung but at the same time, it also underlines that the age of silicon is beginning to draw to a close.
• Intel’s old nomenclature, while simple, is also very misleading because the same numbers at TSMC and Samsung refer to something completely different.
• Historically, progress in semiconductors has been all about making the transistors smaller than previous generations so that more of them would fit into the same space giving better performance and lower energy consumption.
• This is why everyone has used microns or nanometres to name their process nodes as this referred to the width of the line being drawn on the silicon and therefore the size of the transistor.
• Unfortunately, there are different ways to measure this which has led to huge confusion, as in real terms, Intel’s 7nm process is equivalent to TSMC’s or Samsung’s 5nm process as all of them get around 90m transistors onto each square millimetre of silicon.
• Intel’s new naming process aims to get rid of this discrepancy and moves it more into line with TSMC and Samsung making comparison easier to achieve.
• As a result, Intel’s 10nm SuperFin keeps its name but the old 10nm Enhanced SuperFin will be renamed Intel 7 and the old 7nm will be renamed to Intel 4.
• Future iterations of Intel will be called Intel 3 with the next major node being named Intel 20A.
• This is effectively 2nm but is being referred to as 20A to reflect the move into the Angstrom era of measurement where 1nm is the same as 10 Angstroms.
• Intel currently expects to put Intel 20A into volume production in 2024 assuming all goes well with the new fabs and the IDM2.0 strategy.
• What is also clear from the nomenclature is that other new technologies such as packaging, power delivery, and transistor structure will have an increasing role in delivering the performance and power improvements that are demanded from successive chip generations.
• It is these sorts of improvements that will underpin Intel 3 as well as Intel 18A and I think it is clear that these will increasingly become what determines the performance difference in semiconductor chips going forward.
• This highlights that silicon is beginning to run out of road in terms of die shrinks and that other tricks and refinements will be needed to deliver the improvements that the industry has depended on for growth since the 1970s.
• There are physical limits to how small transistors can be made and we are already quite far beyond what many thought was possible when I started looking at the semiconductor industry in 1996.
• Hence, I think that thoughts and investments will soon turn to what lies beyond silicon, and here the lead candidate is quantum computing.
• It is still early days, and it could be 15 to 20 years before this gets a foothold in the mainstream of computing, but anyone who runs a discounted cash flow analysis on Intel, TSMC, or ASML will soon need to start thinking about this.
• Silicon still has many years to run and a giant industry will take a long time to migrate but the signs are here that the end for silicon is within the foreseeable future.