Taiwan Semiconductor Holds the Cards As NVIDIA, Apple and AMD Battle for Limited AI Chip Capacity - Benzinga

Skyler Monroe (0:10)

Hey, everyone, welcome back to The AI Hardware Show. I'm your host Skyler Monroe, computer engineer turned silicon obsessive, and this is the show where we break down the chips, the data centers, and the hardware making AI actually happen. Big thanks to our sponsors today. Ada, helping businesses integrate AI into their workflows. Ago Consulting, that's Ago, your go-to for silicon development from AI to So. And Zen Semiconductor, the powerhouse behind modern data center silicon, including their Sierra RISC vCPUs and Loom AI fabric, plus Ada for deploying AI at scale. Today, we're diving deep into TSMC record profits, capacity wars, and a possible Texas Mega Fab, and then AMD is throwing some serious punches at the data center GPU market. Let's get into it. All right, story one. The headline is Taiwan Semiconductor holds the cards Nvidia, Apple, and AMD battle for limited AI chip capacity. And honestly, that headline kind of says it all, but let me unpack why this matters so much. So here's the thing about TSMC, Taiwan Semiconductor Manufacturing Company. They are the world's most important chip factory, and I don't mean that loosely. When we talk about the most advanced chips on the planet, the ones powering your iPhone, your RTX GPU, your AI accelerators, the overwhelming majority of them are made at TSMC. There is no real substitute at the bleeding edge right now.

Think of TSMC like the only master chef in town who knows how to cook a dish that everyone desperately wants. You've got Nvidia coming in saying, I need a table. Apple walks in saying, no, I need a table. AMD shows up, Qualcomm's right behind them, and the chef just looks out at the packed restaurant and goes, I'll get to you when I get to you. That's basically the capacity situation at TSMC's most advanced nodes right now. We're talking three nanometer and two nanometer class manufacturing. These are the process nodes where the transistors are so small, you're packing billions of them into a space the size of your thumbnail, and there just isn't enough fab capacity to go around. Now the reason this is hitting a fever pitch right now is AI.

The demand for high-end AI accelerators has gone absolutely vertical.

Nvidia's Blackwell GPS, AMD's MI series, custom silicon from hyperscalers like Google's TPS and Amazon's Tranium. They all want TSMC's most advanced manufacturing, and TSMC can only build so many wafers per month. A wafer, for context, is basically a big silicon disk, usually 300 millimeters across, and you slice it up into individual chips. The more advanced the process node, the more chips you can fit per wafer, but also the harder and more expensive the process is to run. TSMC has a finite number of these advanced machines, and it takes years to expand capacity. So what happens when demand outstrips supply? Prioritization, and that's where the politics get really interesting. TSMC reportedly prioritizes based on volume commitments and the length of the relationship. Nvidia is one of their biggest and oldest customers. Apple is essentially their anchor tenant. Apple alone accounts for a massive chunk of TSMC's revenue. AMD is a huge customer too. When all three of those companies are simultaneously screaming for more capacity, someone has to wait. And right now, smaller or newer customers, or even big customers who want to ramp faster than expected are finding that TSMC just can't accommodate them on their preferred timeline. The strategic implication here is enormous.

It means your chip road map isn't just about how good your chip design is. It's about how much fab capacity you've secured. In the AI era, access to advanced manufacturing is itself a competitive moat. If you can't get your chips made at scale, your brilliant design just sits on the drawing board.

This is also why you're seeing so many companies trying to diversify looking at Samsung's Foundry, or Intel Foundry services trying not to have all their eggs in one basket.

But the reality is, neither Samsung nor Intel Foundry is quite at TSMC's level on yield and reliability at the cutting edge. So everyone ends up back at TSMC's door anyway. And TSMC knows this. They're in an extraordinary position of power. The entire global AI buildout, hundreds of billions of dollars in investment is flowing through this one company in Taiwan. That's a geopolitical story, a business story, and a technology story all wrapped up in one. We'll keep coming back to this one.

Okay, story two.

TSMC stock drops despite record profit amid extremely robust demand for AI chips. And I love this one because it's such a perfect illustration of how weird markets can be.

So TSMC just reported record profits, record, as in the best financial performance in their history.