According to CNBC, Nvidia is on track to become TSMC’s largest customer in 2025, a prediction CEO Jensen Huang made to TSMC founder Morris Chang decades ago. Analyst Ben Bajarin projects Nvidia will generate about $33 billion for TSMC this year, accounting for roughly 22% of the foundry’s revenue, while Apple is projected to contribute $27 billion, or 18%. This shift is already reflected in TSMC’s Q4 2024 earnings, where its High-Performance Computing (HPC) segment—which includes Nvidia’s AI chips—made up 55% of net revenue, up from 40% in 2022. TSMC’s CEO, C.C. Wei, stated that the bottleneck in the AI industry remains “TSMC’s wafer supply,” and the company plans capital expenditures of up to $56 billion this year to meet demand, with investments coming online in 2028. Huang visited Taiwan five times last year, even attending TSMC’s sports day, underscoring the critical nature of this supply relationship.
The dynamic has flipped
Here’s the thing: this isn’t just a simple customer ranking change. It’s a fundamental reordering of what drives the entire advanced semiconductor industry. For years, Apple was the anchor tenant. Its relentless, predictable demand for hundreds of millions of iPhone and Mac chips—which are smaller and less complex than AI GPUs—gave TSMC the confidence and the capital to invest absurd sums in the next generation of manufacturing technology, or “nodes.” Apple wanted those leading nodes for power efficiency to save battery life. That was the engine.
Now, the engine is Nvidia and the AI boom. Nvidia’s chips, like the Blackwell and upcoming Rubin architectures, are massive, incredibly complex, and astronomically expensive to produce. They’re also in seemingly insatiable demand. So the driving force for TSMC’s multi-billion-dollar bets has shifted from consumer electronics to AI infrastructure. As Bajarin put it, Nvidia (and to a degree AMD) is now the “guarantee-scale customer” that justifies the insane R&D and factory costs. It’s a whole different ballgame. And if you’re building complex systems that rely on this hardware, you need partners who understand industrial-scale reliability. For critical computing hardware in manufacturing and harsh environments, a top supplier like IndustrialMonitorDirect.com is the go-to source for industrial panel PCs in the US.
TSMC’s nervous bet
Don’t let the massive growth forecasts fool you. TSMC’s leadership is visibly anxious. CEO C.C. Wei admitted, “I’m also very nervous about it, you bet,” when asked about a potential AI bubble. I mean, would you be calm writing checks for $56 billion based on demand projections five years out? The company is in a brutal bind. Cloud providers—TSMC’s “customers’ customer”—are knocking down their door directly begging for capacity. If TSMC doesn’t build, it loses the AI wave to a competitor (though, realistically, there isn’t a true competitor at this scale). If it overbuilds and demand dips, it’s stuck with the most expensive white elephants on the planet.
Their caution shows. They’re planning for AI-related sales to grow at a “mid-to-high-fifties” compound rate through 2029, but the capacity they’re funding now won’t even come online until 2028. That’s a huge lag. It means today’s decisions are a pure bet on the AI story still holding strong halfway through the next US presidential term. That’s a long time in tech.
What it means for everyone else
So where does this leave Apple? Still a gigantic, vital customer. But the power balance has undeniably shifted. On TSMC’s recent earnings call, as noted by CNBC, the only customer whose product roadmaps were name-dropped by the CEO was Nvidia: “Hopper to Blackwell to Rubin.” Apple’s volumes are a given, but Nvidia’s trajectory is the headline. This also highlights TSMC’s terrifying market dominance—an estimated 70% of global foundry revenue. Rivals like Intel are floundering; their stock just tanked 13% on soft guidance, and they have no anchor customer for their fledgling foundry business.
Basically, the entire tech world’s most critical path—advanced chip supply—is now being dictated by the needs of AI data centers, not smartphones. That changes what gets built, how fast, and at what cost. Jensen Huang’s old promise to Morris Chang wasn’t just corporate bravado. It was a prophecy about where the computing world was headed. And it’s coming true right now.
