HPC has become the volume and mix anchor, so TSMC is increasingly tied to cloud capex rather than consumer replacement cycles

HPC 61% of 1Q26 sales versus 55% in 4Q25; smartphone fell to 26% from 32%

Platform mix moved further toward HPC even in a quarter when handsets should have dominated. That makes the revenue base more sensitive to…

Leading-edge concentration stayed exceptionally high, which keeps the revenue mix structurally richer than a typical foundry upcycle

3nm 25%, 5nm 36%, 7nm 13%; advanced nodes still 74% of wafer revenue

TSMC did not need a broad mature-node recovery to grow. It relied on concentrated demand in the nodes with the highest pricing power and th…

Management is no longer describing AI as optional upside; it is describing a multi-year compute build that is resetting the top-line base

AI demand remains 'extremely robust'; FY26 revenue growth now above 30% in USD

The transcript tied demand to the shift from generative AI toward agentic AI, which raises token consumption and compute intensity. Managem…

This was an execution-led beat, not a sudden pricing squeeze on customers

Gross margin beat the guided range by 120 bps

Management attributed the upside to higher-than-expected utilization and better cost improvement efforts, and later reiterated that pricing…

That is why investors should not treat 66% as a clean new floor, because the real pressure starts when N2 and overseas fabs scale in 2H26

N2 startup dilutes FY26 gross margin by 2%-3%; overseas fabs dilute 2%-3% early and 3%-4% later

TSMC can offset dilution for a while with productivity and mix, but once N2 ramps and the global footprint gets larger, the fixed-cost base…

N2 is already in high-volume manufacturing with good yield, which lowers the risk that the next node slips just as AI customers need more compute

N2 HVM started in 4Q25; demand is coming from both smartphone and HPC/AI

The node is no longer just a roadmap milestone. It is in the factory and ramping across Hsinchu and Kaohsiung, which matters because custom…

TSMC is making N3 a global platform because the demand pipeline is too large to serve from Taiwan alone

New Tainan N3 fab in 1H27, Arizona N3 in 2H27, Japan second fab now planned for 3nm in 2028

Management explicitly said this breaks with the normal pattern of stopping capacity additions once a node reaches target scale. The change…

Packaging is still tight, which means the bottleneck has moved beyond wafer starts and into advanced integration

Advanced packaging capacity remains very tight; large-reticle CoWoS is still the main supply base

TSMC said it is working with OSAT partners, building a pilot line for the next-generation packaging route and still treating capacity as co…

Validate the bull case if leading-edge utilization stays high enough to keep revenue and margins above the guided ranges even as N2 ramps

2Q26 revenue $39.0B-$40.2B; gross margin 65.5%-67.5%

Another guided-quarter beat while absorbing early N2 and overseas dilution would show that the earnings base has genuinely reset upward rat…

Watch whether AI accelerator demand still tracks toward the higher end of the 50s CAGR and whether cloud customers keep leaning in

Management still sees AI accelerator growth toward the higher end of the 50s CAGR range

The long-duration bull case depends on AI spending remaining a multiyear infrastructure build rather than normalizing after the first deplo…

The near-term break risks are macro and input costs before they are physical supply disruption

Middle East tension can lift chemical and gas costs, but TSMC does not expect near-term materials or energy disruption

TSMC said it has multisource supply, safety stock and government-backed LNG coverage, so the first hit would likely be profitability noise…