It has been confirmed that Intel will use TSMC’s 3nm process, an anomaly – considering that Apple’s iPhones had typically been the first customer of TSMC’s advanced manufacturing nodes. Now, Intel will become the first user of TSMC’s 3nm process. Even though TSMC has, as usual, declined to comment on its customers, sources from supply chain indicated that the 3nm process would be used for one type of Intel’s GPU product as well as three types of server processors. All of them would be Intel’s core products, and volume production would begin in July next year.
It marks the first time for Intel to outsource its chip production, as the company has been struggling with delays of its 7nm process and facing encroachments in its PC and server markets from challengers like AMD and Nvidia. Ironically, Intel has also rebooted itself to prepare for an all-out competition against TSMC. Under the new leadership of Pat Gelsinger, Intel has come up with the grandiose strategy of IDM 2.0 that sees Intel entering the foundry business and even renaming its process nodes.
As one of the most outspoken proponents of the United States’ chip autonomy project, Intel has also played a geopolitical role by raising the possibility of setting up new foundries in Europe and boosting its existent capacity there, as governments worldwide have become increasingly unnerved by the global supply chain’s over-dependence on TSMC.
With Samsung’s 3nm process due next year, Intel’s preference over TSMC also added more credibility to the Taiwanese foundry’s progress.
Apart from Intel, the British startup Graphcore will also adopt TSMC’s 3nm process. Founded in 2016 and backed by Microsoft, Graphcore has positioned itself as a main rival of Nvidia, and has been vocal in its opposition of Nvidia’s acquisition of ARM. Last year, Graphcore unveiled its Colossus MK2 IPU (Intelligence Processing Unit) built on TSMC’s 7nm node, claiming that it was 16 times faster than those offered by Nvidia. So far, not much is revealed regarding Graphcore’s latest 3nm chip.