Nvidia Overtakes Apple at TSMC

Nvidia's revenue contribution to TSMC in 2025 reached approximately $23.2 billion, a figure more than double its 2024 total. This surge elevated Nvidia's share of TSMC's total revenue from 12% in 2024 to 19% in 2025. In contrast, Apple's revenue contribution stood at $20.5 billion, with its share of TSMC's revenue declining from 22% to 17% in the same period. The demand is largely for Nvidia's "Blackwell" B100 and GB100 GPUs, which are being produced on TSMC's 3nm process. This has prompted Nvidia to request a 50% increase in TSMC's 3nm production capacity to meet the high demand for its AI chips. The high-performance computing (HPC) segment, which includes AI chips, has become a major revenue driver for TSMC. This increased demand is creating a significant strain on TSMC's advanced packaging capacity, particularly its CoWoS (Chip-on-Wafer-on-Substrate) technology. Nvidia has reportedly booked more than half of TSMC's CoWoS capacity. The bottleneck in advanced packaging is a critical chokepoint in the AI hardware supply chain. For over a decade, Apple was TSMC's primary customer, often gaining exclusive initial access to new manufacturing processes. This symbiotic relationship was crucial for both companies, with Apple's high-volume orders helping to fund TSMC's research and development into next-generation nodes. The intense competition for advanced manufacturing is not just about volume but also the physical size of the chips. A single high-end Nvidia GPU can consume six to eight times more silicon wafer space than an Apple A-series chip, creating a direct trade-off for every wafer allocated. This dynamic has reportedly led to Apple facing significant price increases for future 2nm wafers. The surge in AI-related demand is causing ripple effects throughout the semiconductor supply chain, leading to shortages and longer lead times for other components like high-bandwidth memory (HBM). This has prompted some chip designers to explore alternative packaging solutions from other foundries to mitigate the bottleneck at TSMC. Nvidia's CEO, Jensen Huang, has stated that the scale of expansion required to meet future AI demand will be unprecedented. He anticipates that Nvidia's own demand for wafers will more than double in the coming years, potentially requiring TSMC to increase its overall capacity by over 100% within the next decade. In response to these capacity constraints and shifting dynamics, there are reports that both Apple and Nvidia are diversifying some of their production. Apple is reportedly looking to Intel for the production of some entry-level Mac and iPad processors, while Nvidia may use Intel for entry-level gaming GPUs and future-generation AI accelerators.