SMIC N+1: A 7nm-Class Node Behind a Closed Door

SMIC N+1 is a 7nm-class node, and SMIC is building it behind a door the United States slammed shut four months ago. In December, Washington put SMIC on its Entity List, restricting the company’s access to American technology and, more pointedly, helping ensure SMIC will never get the one machine the whole industry uses to build chips this small: an EUV lithography system. SMIC N+1 sits roughly at the 7-nanometer class, the node above the 14nm FinFET it reached in 2019, and SMIC has to build it without the tool everyone else uses to build it. That is the entire story of SMIC from here forward.

To understand how hard this is, you have to understand what SMIC has been denied. Extreme ultraviolet lithography is how the leaders print the tiniest features at 7nm and below, in a single clean exposure, and there is exactly one company on earth that makes EUV machines, in the Netherlands, and it has been blocked from selling them to China since before the Entity List made it formal. SMIC ordered one years ago. It will never arrive. So SMIC has to reach 7nm with the older deep-ultraviolet light, the DUV the industry used before EUV, which means printing each tiny layer in multiple painstaking exposures stacked on top of each other, slower, costlier, lower-yielding, the hard way nobody chooses if they have a choice. SMIC does not have a choice.

The one piece of good news is that it is not actually impossible, and the proof is the competition. TSMC built its first 7nm node entirely on DUV, before it moved to EUV, multipatterning its way to the node and shipping it in enormous volume. So 7nm-class on DUV can be done; TSMC did it. The difference is that TSMC then moved to EUV for the refined version and left DUV behind, because DUV at 7nm is a brutal way to live, low yields and high costs and punishing complexity. SMIC has to set up permanent residence in exactly the place TSMC passed through and fled. It can reach the node. It just has to stay there, on the hard path, with no exit to the easy one.

No commercial logic explains this. Building 7nm on DUV is more expensive per chip than buying 7nm from TSMC ever was, the yields are worse, the economics are upside down, and a company optimizing for profit would never do it. Profit is not what SMIC is optimizing for. The goal is a Chinese leading-edge foundry that no foreign government can switch off, and against that the terrible economics are simply the price. China has decided it would rather have expensive, low-yield, sovereign 7nm than no sovereign 7nm at all, because the entire point of SMIC was always to make chips when the outside world would not. The Entity List did not kill that logic. It activated it.

Everything about SMIC is different now. For twenty years it bought its tools and talent on an open global market, and that era ended in December. The new condition is permanent restriction: the best machines withheld, the technology partners scared off, the roadmap now a function of what can be built or improvised inside a sanctions perimeter rather than what can be bought. SMIC’s competitors race each other on yield and cost and speed, while TSMC and Samsung are already shipping 5nm on EUV in volume and SMIC’s own advanced-node output is a rounding error next to either. SMIC races the embargo instead, trying to advance faster than the controls can tighten, building capability the country can keep no matter what Washington does next. It is a different kind of foundry from anything that has existed before.

Whether SMIC can actually ship a real 7nm-class node on DUV in volume is, in 2021, deeply uncertain, and plenty of serious people think the answer is no. Perso, I lean skeptical on the timeline: I think SMIC gets something working in a lab and a low-volume line, but that DUV 7nm at any meaningful scale, with yields that are not catastrophic, stays out of reach for years, maybe longer, because the costs are just punishing and the embargo only gets tighter. I would not bet on a Chinese 7nm phone chip in volume anytime soon. The thing I am confident about is the will: SMIC will keep pushing on this regardless of economics, because the economics were never the point, and a national project does not stop because the math is ugly. Whether the will is enough to beat the physics and the embargo at the same time is the question. In 2021 the smart money says no.