The Semiconductor Shortage Is Back And It Is Different This Time

Lucas (0:01)

Look at the SOX-X index right now.

It is up 1.4% over the last five days, which feels like stability on the surface. But if you drill into the components, there is a quiet fracture forming in the foundation of the industry. Everyone is talking about the race for 3nm processed nodes and the explosive demand from artificial intelligence. But the real story today is happening in the older, less glamorous factories.

Luna (0:27)

You mean the mature nodes? The ones that power everything from washing machines to electric vehicle inverters?

Lucas (0:34)

Exactly. We have spent the last year watching giants like Taiwan Semiconductor Manufacturing Company and Intel Poor Billions into cutting edge logic. They are chasing the high margins of AI accelerators and flagship mobile processors. But by shifting their best wafers toward these advanced processes, they are effectively starving the existing capacity for legacy chips.

The result is a classic supply-demand mismatch that nobody wants to admit yet.

Luna (1:03)

It reminds me of the panic we saw back in 2021, but this time it is not a pandemic lockdown causing it. It is pure capital allocation choice.

Lucas (1:13)

Precisely. In 2021, factories were closed. Today, they are running at maximum efficiency, but they are making the wrong product for the bulk of the market.

Take Qualcomm. Their stock has jumped 7.8% recently because their clients need those mid-range connectivity chips more than ever. Meanwhile, memory stocks like Micron are down 4%.

The market is signaling that bandwidth is not the bottleneck anymore. It is the sheer number of available transistors in mature packages.

Luna (1:46)

So when I buy a new laptop this fall, I might be paying a premium not for faster speed, but just to secure the basic processor inside.

Lucas (1:55)

That is a very likely scenario. Let us look at Intel. Their shares are up 7.5% this week, driven by hopes that their foundry business will finally turn a corner. But here is the rub.

Intel is trying to build fabs for 22nm and 12nm processes, while simultaneously competing with Apple and Nvidia in the sub 5nm space. You cannot do both efficiently without diluting your yield rates. They are stretching their operational bandwidth too thin.

Luna (2:27)

And that dilution hurts everyone who isn't building the absolute fastest supercomputer. It creates a vacuum in the middle market.

Lucas (2:35)

Right. The middle market is where the volume is.

When you look at the automotive sector, an average modern car uses nearly 2000 discrete chips.

Most of them are not high-performance computing units. They are sensors, power management ICs, and display controllers. These rely on 10-year-old or 15-year-old technology that works perfectly fine. But because it is not sexy, fab owners are not expanding capacity for it. They are retooling lines for something entirely different.

Luna (3:08)

It is fascinating how the definition of scarcity shifts.

A few years ago, scarcity was about having any chip at all. Now, it is about having the right chip, and the right chip is often the oldest one on the shelf.

Lucas (3:23)

And that is where the pricing power flips. Usually, you think of innovation driving price. Here, stagnation is driving price. Because no one is investing in new lines for these mature nodes, the existing capacity becomes incredibly valuable. We are seeing lead times for certain legacy parts stretch out again. If you are an automotive OEM or a medical device manufacturer, you are not just waiting for parts.

You are bidding against each other for access to wafer slots that haven't been touched in a decade.

Luna (3:56)

Does this mean companies like Broadcom and Analog Devices are sitting on a goldmine? Their focus is heavily on these specialized analog and mixed signal chips.

Lucas (4:06)

Broadcom is certainly benefitting from the AI boom, yes, but the analog players are seeing a secondary surge. Because their products don't require the most advanced lithography, they can use the capacity that is being freed up as fabs upgrade. However, even they are feeling the pinch because the equipment needed to maintain these older lines is becoming scarce. ASML, the Dutch lithography giant, is down slightly today, but the maintenance costs for EUV machines are skyrocketing.

Fab operators are hesitant to keep old tools running when they could be putting that floor space to work for higher margin advanced nodes.

Luna (4:44)

So it is a double bind. You need the old tech to run the world, but you can't afford to keep the old tech alive.

Lucas (4:51)

That is the core tension. And it is getting worse because the geopolitical landscape adds another layer of complexity.

With export restrictions on advanced technology to certain regions, manufacturers are trying to diversify their supply chains. They are building new fabs in places like Japan, Germany, and Arizona. But building a new fab takes three to five years. Those facilities will produce advanced chips, not legacy ones. So in the short term, the gap between supply and demand for mature chips only widens.