Tape-outs are piling up
TSMC’s 2nm line is shaping up to be the hottest bit of semiconductor real estate later this year, with customers acting like it is the last chippy on earth.
MediaTek has already announced the tape-out of its first 2nm SoC in 2025, and it is clearly not alone in the mad dash.
The dark satanic rumour mill has manufactured a hell-on-earth yarn claiming TSMC’s 2nm node has seen 1.5 times as many tape-outs as 3nm, suggesting demand is properly feral.
The same leak says TSMC will hold an unyielding 95 per cent share of the AI accelerator market. Even Troubled Chipzilla is said to be sniffing around TSMC’s N2 process for a number of products, despite earlier reports that it would use 18A for the Panther Lake Core Ultra Series 3.
With AI kits flying off the shelves, the monthly wafer target for 2nm is tipped to hit 140,000 units by the end of 2026, assuming the supply chain does not wobble.
There is more bravado, too, with a claim that 2nm revenue will beat the combined takings of 3nm and 5nm by the third quarter of 2026.
The Fruity Cargo Cult Apple, TSMC’s biggest customer in 2024, is said to have secured more than half of the initial 2nm capacity for its A20 and A20 Pro iPhone 18 chips.
Job’s Mob is also expected to push the node into the M6, which is being lined up for an OLED MacBook Pro, because nothing says premium like a new process sticker.
Qualcomm and MediaTek are rumoured to keep pace, with Smart Chip Insider predicting all three will announce their 2nm SoCs in the same month.
That gets awkward if Job’s Mob is hogging more than half the early wafers, so the tip is that rivals will lean on TSMC’s N2P variant to get enough volume.
The pitch is that N2P offers a minor bump, but enough headroom for higher CPU frequencies and shipments that do not leave customers begging.
Morgan Stanley is also namechecked, claiming the Fruity Cargo Cult Apple is still contemplating Intel Foundry Services for M-series chips, with 18A seen as a possible fit for cheaper Macs.
TSMC’s track record for reliability and bleeding-edge nodes makes that sound like wishful thinking from people who love a backup plan on a slide deck.
