Apple is facing increasing challenges in securing production capacity at TSMC as Nvidia emerges as the dominant force in the semiconductor supply chain. When TSMC CEO CC Wei visited Apple in August, he delivered difficult news: the company would need to accept significant price increases for its chips. While Apple’s leadership absorbed the blow, the deeper issue was that the iPhone maker, once TSMC’s top client, now must compete fiercely for limited manufacturing space. According to supply chain sources and analysis by Culpium, Nvidia likely surpassed Apple as TSMC’s largest customer in one or more quarters of 2025. While TSMC has not officially confirmed the shift, CFO Wendell Huang declined to comment on client rankings, saying only, “We don’t discuss that.” Final data will come with TSMC’s annual report, but signs point to a major change in the hierarchy. If it didn’t happen in 2025, it’s expected to occur in 2026. The shift reflects a fundamental transformation in the semiconductor industry. TSMC’s revenue rose 36% to $122 billion in 2025, driven largely by demand from high-performance computing (HPC), particularly AI chips. Nvidia’s sales for the fiscal year ending January 2026 are projected to grow 62%, while Apple’s product revenue—excluding services—is expected to rise just 3.6% over the same period. Apple’s role as TSMC’s primary growth engine ended five years ago, and now AI-driven demand is reshaping the landscape. HPC revenue at TSMC jumped 48% in 2025, up from 58% the year before, while smartphone-related revenue grew only 11%. This trend is expected to continue, with TSMC projecting 30% revenue growth in 2026 and capital expenditure rising to between $52 billion and $56 billion—up 32%. Over the next five years, HPC chip demand is forecast to grow at an average of 55% annually, far outpacing overall industry growth. TSMC’s gross margin reached an impressive 62.3% in the December quarter, nearly matching software companies. This strength is partly due to high-margin AI and HPC chips, though overseas fabs in Arizona and Japan slightly dampen the figure. Still, the company’s profitability underscores its strategic advantage in advanced manufacturing. Apple remains a critical client, but its position is evolving. While its iPhone processors are still a major part of TSMC’s business, Apple also uses HPC nodes for Mac chips and custom silicon in accessories. However, Nvidia’s demand is more concentrated and massive—each of its AI GPUs consumes significantly more wafer space than a smartphone chip. This gives Nvidia disproportionate leverage. TSMC is responding by rapidly expanding capacity. It is already producing chips at its cutting-edge 2nm node (N2), with Apple as a key customer. In the second half of 2026, it will ramp up a new variant, N2P, and introduce the A16 node, designed for complex HPC workloads with advanced power delivery technology called Super Power Rail. Unlike traditional methods, TSMC builds new fabs for each new node, ensuring no disruption to older production lines. Looking ahead, the A14 node—expected to launch around 2028—is designed for both mobile and HPC from the start, signaling a potential return to balance. Apple’s strength lies in its broad and stable manufacturing footprint across a dozen TSMC fabs, giving it resilience even if AI demand cools. While Nvidia dominates today’s AI boom, its market is still niche compared to Apple’s diverse product ecosystem. TSMC’s leadership knows the cycle won’t last forever. As Wei warned at the investor conference, “I am also very nervous.” Overbuilding risks massive losses if demand falls, and TSMC’s high capital intensity—over 33%—means it bears the full burden of investment. In contrast, Nvidia and Google operate with much lower capital intensity and depreciation costs. They benefit from TSMC’s scale without the manufacturing risk. This dynamic explains why eight of the world’s ten largest companies rely on TSMC. For now, Nvidia is the king of demand, but Apple’s enduring presence ensures it remains a vital partner. The real test will come when the AI boom slows—then TSMC’s long-term strategy will be put to the ultimate test.