TSMC raised its 2026 growth target to 44% on the Q3 2026 earnings call. AI demand, in management's telling, sits "far above" what the foundry can ship. Two-nanometer wafers cleared 8% of revenue for the first time.

The Hsinchu-headquartered company, the world's largest contract chipmaker, lifted full-year guidance well above prior estimates. The 44% number would make 2026 TSMC's strongest growth year of the decade, and it crystallizes a supply story executives have been telling for quarters: leading-edge demand outruns capacity, and the gap will not close inside 2027.

What does 8% of revenue from 2nm actually mean?

The N2 share figure is the headline inside the headline. Reaching 8% of revenue inside roughly four quarters of high-volume production is a faster curve than the 7nm and 5nm ramps at the same stage, both of which took more quarters to hit similar share milestones after launch.

N2 is doing the heavy lifting behind the 44% target. The node is pulling in customers at the high end of the mix — high-performance compute for AI training and inference, plus premium mobile SoCs — while ASPs sit well above legacy wafer prices. Legacy nodes, by implication, are growing slower than the corporate average, even if they remain profitable.

The yield curve has tracked ahead of the N3 ramp at the same stage, based on management's broader 2026 commentary. TSMC has not published a hard yield number for N2. The 8% revenue share, though, is itself a proxy. Bookings at that scale only happen once defect density reaches commercial-grade levels.

How big is the supply gap?

Management did not name a dollar figure for unmet AI demand. The "far above" framing, though, matches language used on prior 2026 calls. The practical implication: N2 and N3P capacity through 2027 is essentially committed.

Pricing on those nodes, on that read, holds the upper hand. Customers building frontier AI systems have limited alternatives. Samsung Foundry's 2nm GAA has not reached a comparable 8% revenue share on a similar ramp timeline. Intel's 18A continues to court external customers without yet matching TSMC's leading-edge footprint. The competitive gap at the frontier, in short, is widening, not narrowing.

Why does the Taiwan concentration still matter?

Geopolitics has not changed the commercial picture. The bulk of N2 output through 2027 still comes from Taiwan. New fabs in Arizona, Kumamoto, and Dresden are adding capacity, but they are not replacing Taiwan's role as the leading-edge hub. Customers building AI systems on N2 are, knowingly or not, accepting that geographic concentration.

That concentration is the one variable that could break the 44% trajectory. A disruption event, a tariff, or a customer diversification push would each carry weight. None of those show up in the current numbers.

What to watch next

The Q4 2026 call, due in January 2027, will test two things. First, whether 2nm crosses double-digit revenue share. Second, whether management sticks with "far above" or softens the framing. Either move would reset the 2027 setup and pull forward the conversation about when, not whether, supply finally catches up.