SMIC 45nm went into production this year, and for the first time the company reached a node only about two years after the leaders did rather than three or four. SMIC 45nm, and the 40nm half-step beside it, is the most current process the Shanghai foundry has ever run, a real recovery from the wreckage of the 2009 lawsuit that cost it a founder and a chunk of equity. The gap to TSMC is still there. It is just, for once, getting smaller instead of larger.
The bigger news under the node is that SMIC turned a profit last year, its first since 2004. A small one, around fourteen million dollars, but real, after most of a decade of losses absorbed by a patient state. It matters less for the cash than for what it proves: SMIC can run as a real foundry, not just a subsidized science project, taking volume orders and filling fabs and earning on the wafers. Whether the black ink holds through this year is another question, because the foundry market has cooled and the margins are thin, but the point stands. The state money built the company. The company is starting, slowly, to stand on the business.
The shape of SMIC’s real business is now clear, and it is not the leading edge. It is the enormous, unglamorous mature-node market, the 45nm and 40nm and older processes that build the chips most electronics actually need, the power management and the display drivers and the microcontrollers and the cheap application processors. China assembles most of the world’s electronics, and most of those chips do not need the bleeding edge. The natural customer base is China’s own fast-growing fabless industry, HiSilicon inside Huawei, Spreadtrum turning out phone chips, a widening crowd of design houses. The humbling catch is that even they do not default to SMIC: surveys of Chinese fabless companies still rank TSMC first by a wide margin and SMIC a distant second for their most demanding work. SMIC sitting on a deep mature-node base, inside China, serving Chinese assemblers, is a strong and durable position. It is also a floor, not a summit.
Above that floor, the climb gets steep fast. SMIC holds only a few percent of global foundry revenue, well behind TSMC’s roughly half and behind GlobalFoundries and UMC in the tier between, and the leaders are pulling away again. TSMC is at 28nm and moving toward the first FinFET nodes, the three-dimensional transistors that will define the next decade, and SMIC has no FinFET, no clear path to it yet, and a widening technical chasm between mature-node competence and leading-edge capability. The two are almost different industries. Running 45nm well does not teach you how to run 14nm; the physics changes, the tools change, the difficulty compounds. SMIC has proven it can be a good mature-node foundry. Whether it can ever be a leading-edge one is a completely separate question, and 45nm does not answer it.
What comes next, if SMIC is lucky, looks like a stretch of steady, quiet progress, the company consolidating its mature-node business, inching toward 28nm, building the engineering depth that the lawsuit-era shortcut never gave it. This is the part of the national project that does not make headlines, the slow accumulation of capability that a patient state funds and waits out. There is no drama in 2011, and that is the point. After the chaos of 2009, boring is exactly what SMIC needs, a few years of just running the business and climbing one careful node at a time.
SMIC in 2011 is a healthier, humbler company than the one that came out of the courtroom, focused and backed by the same bottomless patience. Perso, I think the next decade is steady mature-node growth with a slow grind toward the leading edge, the gap to TSMC narrowing gradually as Chinese engineering matures, nothing dramatic either way. The thing that could change that calculus, for better or worse, is politics, and in 2011 politics is not really in the picture; chips are a global business and SMIC buys its tools on the open market like everyone else. As long as that stays true, the only question is how fast SMIC climbs, not whether it can. The tools are available and the money is patient. From here it just looks like a matter of time.