Why Chip Stocks Are Plunging Despite AI Boom

Lucas (0:01)

Look at the board right now, if you're tracking the SOXX, that semiconductor ETF, and it's down nearly 9.4% over the last five days. It feels jarring because on one hand, every major cloud provider is screaming for more AI chips.

Yet on the other, stocks like Intel have dropped 11.3% and Applied Materials is sliding 17.7%.

Luna (0:24)

It really does feel disconnected from the headline narrative we've been hearing for two years straight about insatiable AI demand. Is this just a technical correction or are we seeing a fundamental shift in how the market values hardware?

Lucas (0:39)

I think it's both, but let's focus on the hardware side first because the disparity is so wide. You have NVIDIA holding relatively steady despite its own 5.4% dip, which suggests the market still believes in the top tier proprietary silicon model. But when you look at companies like Advanced Micro Devices dropping 10% or Micron tumbling 11%, we're seeing a clear rejection of the broader commodity and legacy logic.

Luna (1:05)

Right. So the premium is staying with the designers who control the architecture, while the manufacturers and memory makers get punished for the cyclical downer. Does this mean the margin expansion we expected for foundries is actually stalling?

Lucas (1:20)

That's the core tension. Look at Taiwan Semiconductor Manufacturing Company. They're down 7.1%.

The issue isn't lack of orders, it's the cost of those orders. Building these advanced nodes requires massive capital expenditure, and investors are starting to discount the timeline for return on invested capital. When you spend billions on a fab that takes three years to ramp, any sign of slowing downstream demand gets priced in aggressively.

Luna (1:50)

So we're seeing a flight to quality, where only the absolute best-in-class fabrication technology commands a multiple. What about the equipment makers like LAM Research or KLA Technology?

Lucas (2:02)

Exactly. LAM Research is down almost 15% and KLA is down 14.7%.

These companies sell the tools that fabs buy to expand capacity. If the fabs decide to pause expansion because their customers aren't buying enough high-end servers, the tool makers get hit twice. First by the order cancellation, then by the long-sale cycle lag. It's a classic leverage effect in manufacturing.

Luna (2:28)

But haven't we heard repeatedly that AI training clusters require more layers of complexity, meaning more etching and deposition steps per chip? That should theoretically boost equipment spending, not reduce it.

Lucas (2:41)

It does, but there's a timing mismatch. The spending happens upfront when you build the fab, but the utilization comes later.

Right now, October 19, 2026, we're in a lull where early buyers are digesting their inventory. I'm looking at reports that data center operators are holding back on new rack deployments until the software stack matures further. Until that software efficiency improves, the hardware ROI looks thin.

Luna (3:09)

So the bottleneck has shifted from silicon availability to software optimization? That changes the whole investment thesis for the equipment vendors.

Lucas (3:19)

It absolutely does.

And consider Intel specifically. They are down 11.3%, which might seem small compared to some peers, but their structural issues are different. They are trying to become a foundry for others while competing against TSMC and Samsung. The market is skeptical that they can execute the turnaround fast enough to justify the current valuation, especially with cash burn rates rising.

Luna (3:45)

Intel's foundry ambitions have always been ambitious, but do you think the recent drop reflects a loss of confidence in their customer base, who would actually use their process nodes anyway?

Lucas (3:56)

It's less about the customer base right now and more about execution credibility.

Every time they miss a node milestone or delay a product launch, the risk premium increases.

With Broadcom also down 7%, we see that even the custom ASIC designers are feeling pressure. If the hyperscalers start questioning the economics of building custom silicon versus buying off the shelf from NVIDIA, the entire ecosystem contracts.

Luna (4:24)

That's a scary thought for the mid-tier chip designers. If the big guys pull back on custom designs, doesn't that leave everyone exposed?

Lucas (4:33)

It leaves them exposed to commoditization. But let's zoom out for a second. We've got Qualcomm down only 1.3% and Nexus Instruments down less than 2%.

Those are the analog and connectivity players. Why are they holding up while the digital giants get crushed?

Luna (4:52)

Because their markets are diversified across automotive, industrial and consumer electronics. They aren't tied exclusively to the hypervolatile AI server buildout. Their cycles are slower and steadier.

Lucas (5:05)

Precisely. So we're seeing a bifurcation in the semiconductor sector.

On one side, you have the AI infrastructure play which is currently overheating and now correcting. On the other, you have the industrial and automotive side which is stable but boring. Investors are rotating out of the former into the latter, or perhaps just into cash entirely.