TSMC’s 3nm era is officially in full swing. In the third quarter of its 2025 revenue period, the 3nm family accounted for 23 percent of total revenue, surpassing the 5nm node and firmly taking over as the foundry’s workhorse. While 2nm is gathering momentum with initial capacity for 2026 reportedly sold out at two local plants, industry chatter points to 3nm entering a golden period of mass production right now.
According to figures cited by Commercial Times, TSMC’s 3nm output has climbed from roughly 100,000 wafers per month at the end of last year to about 100,000–110,000 today, with forecasts calling for a ramp to around 160,000 wafers per month by the end of 2025. The biggest catalyst behind that surge isn’t Apple this time—it’s NVIDIA. The GPU leader is said to be adding approximately 35,000 3nm wafers per month to TSMC’s tally, reshaping the foundry’s mix toward high-performance computing and AI-centric workloads.
Apple still looms large in the 3nm picture. Earlier expectations tied a sizeable share of TSMC’s 3nm utilization to increased iPhone 17 shipments, and Apple accounted for about 24 percent of the foundry’s revenue in 2024. However, NVIDIA could steal the show over the next two years as its next-generation Vera Rubin and Rubin Ultra platforms reportedly move to TSMC’s N3P process, positioning 3nm at the heart of the company’s data center and HPC roadmap.
Looking ahead, analysts estimate that TSMC’s 3nm process could command 30 percent or more market share next year. Even so, Apple’s 2nm plans add a compelling twist. The company is rumored to have four 2nm chipsets in development for next year and has allegedly locked up more than half of TSMC’s initial 2nm capacity. If that pace holds, Apple could maintain its status as the foundry’s top revenue contributor in 2026, even as 3nm continues to scale for AI, data center, and premium mobile.
Bottom line: 3nm is the engine driving TSMC’s near-term growth, with NVIDIA’s AI and HPC demand accelerating the ramp, while 2nm is shaping up to be the next battleground where Apple’s early capacity reservations could pay dividends. For the semiconductor ecosystem, that means tighter capacity, faster node transitions, and a sustained focus on performance-per-watt at the cutting edge.
