2019 was a stellar year for TSMC and its partners. Computer chips based on the foundry’s flagship 7nm process dominated markets across various segments. Apple, AMD, Huawei, Mediatek, and Qualcomm all had their latest chips fabbed by TSMC, putting the company’s N7 process in the limelight. Not only does TSMC have the most clients, but it also has a process advantage over competitors Samsung, Global Foundries and Intel, with the latter struggling to ramp up 10nm production. GF, on the other hand, is still stuck on 12nm with no apparent plans to progress to smaller nodes.
Samsung has a brighter portfolio, but it still pales in comparison to TSMC. While the latter has already shipped several hundred thousand chips based on the N7 process, Samsung is yet to start volume production based on its 7nm EUV process. The company intends to introduce an 8nm LPP process to bridge the gap between the upcoming 7nm and present 10nm nodes, but as of now, NVIDIA appears to be the only major client.
In terms of the Foundry revenue per wafer, TSMC once again has a major advantage. The deltas between the three main competitors are rather glaring here. Where TSMC makes nearly $1,550 per wafer, Samsung makes less than $650. That’s less than half as much. Global Foundries is better off with a fairly consistent revenue of $1,000 per wafer.
Over the last decade, TSMC is the only foundry whose fortunes have improved. Global Foundries’ profits have remained largely the same while Samsung has mainly competed with UMC, with 2019 being the company’s worst year in over a decade.
Intel and Samsung’s fabrication troubles are part of a bigger problem. IC miniaturization has started to hit its limits. No matter how much Bob Swan and Co try to deny it, Moore’s Law is dying if not already dead. The Foundry with the process advantage is still leading, but how long that lasts remains to be seen.
PS: Special thanks to Takeshi Hatori for his remarkable graphs and industry coverage.