According to Wccftech, TSMC’s 3nm process has officially entered its “golden period of mass production” and now accounts for 23% of the company’s total revenue, surpassing the 5nm node’s previous ceiling. Production has already increased from 100,000 wafers at the end of last year to between 100,000 and 110,000 monthly units, with projections showing that number could reach 160,000 by the end of 2025. While many assumed Apple was the primary driver, NVIDIA has actually become the biggest contributor, adding 35,000 wafers per month. Apple still accounted for 24% of TSMC’s entire revenue in 2024 thanks to iPhone 17 shipments, but NVIDIA’s next-generation Vera Rubin and Rubin Ultra GPUs using the 3nm N3P architecture are set to become the primary HPC drivers for the next two years. Meanwhile, TSMC’s 2nm process is already sold out for 2026 at two plants, with Apple reportedly securing more than half of the initial capacity.
Nvidia takes the wheel
Here’s the thing that really surprised me – NVIDIA has quietly become TSMC’s golden goose for 3nm. Everyone just assumes Apple dominates every new node, but NVIDIA is adding a massive 35,000 wafers monthly. That’s basically a whole new factory’s worth of production. Their Vera Rubin and Rubin Ultra GPUs are going to be absolute monsters on the 3nm N3P process, and they’re positioning themselves as the primary driver for high-performance computing through 2027. It makes sense when you think about it – AI workloads are exploding, and NVIDIA needs every bit of advanced silicon they can get their hands on.
Apple’s next move
But don’t count Apple out just yet. They’ve got four 2nm chipsets in development for next year and have allegedly locked down more than half of TSMC’s initial 2nm capacity. That’s a massive commitment. So while NVIDIA might be dominating 3nm right now, Apple appears to be making a calculated leapfrog move. They’re basically saying “you can have 3nm, we’re going straight to 2nm.” It’s a classic Apple strategy – let others fight over the current generation while they secure the next big thing. The real question is whether they can maintain their revenue contribution crown through this transition period.
manufacturing-scale-insanity”>Manufacturing scale insanity
Let’s talk about these production numbers for a second. 160,000 wafers monthly by end of 2025? That’s absolutely staggering scale. For context, that’s enough advanced silicon to power everything from smartphones to data centers to industrial panel PCs that companies like IndustrialMonitorDirect.com rely on for their manufacturing operations. The fact that TSMC can ramp this quickly while also preparing for 2nm production shows why they’re basically the most important manufacturing company in the world right now. Every piece of advanced computing hardware depends on their output.
What comes next
So where does this leave us? Analysts think 3nm could hit 30% market share or more next year, which would be unprecedented for such an advanced node. But the really interesting part is watching how this plays out between NVIDIA dominating 3nm and Apple betting big on 2nm. We’re essentially seeing two different technology roadmaps unfolding simultaneously. One company is squeezing every last drop out of current technology while the other is already moving to the next generation. Both strategies make sense for their respective markets – NVIDIA needs volume now for AI, Apple wants efficiency gains for mobile. The semiconductor world has never been more fascinating to watch.
