Most people following the artificial intelligence (AI) story remember the GPU crunch that hit the industry a few years ago.
When ChatGPT first launched, big tech companies rushed to buy super-fast GPUs—the raw computing power that allows AI systems to “think.”
These specialized chips process billions of calculations every second so chatbots can answer your questions in real time. As the leading supplier of those chips, Nvidia (NVDA) became the most valuable company in the world.
Investors who spotted that bottleneck early made a fortune.
Since then, it’s been a race to identify the next bottleneck forming in the AI stack. Find it early… identify the companies solving it… and you could have another Nvidia on your hands.
Today’s essay is about where that bottleneck is shifting next.
- We don’t just need more chips… but the machines that make them.
For the past few years, chipmakers have been squeezing every ounce of output from their factories. Now, they’ve hit a physical wall.
Take chip giant Taiwan Semiconductor (TSM), for example. It used the spare manufacturing capacity it built during the COVID chip boom to meet the initial surge in AI demand.
That spare capacity is now gone.
Today, its fabs are running at 100%. So it plans to spend $54 billion this year alone to expand production—up from $40.9 billion in 2025 and $29.8 billion in 2024.
And Taiwan Semi isn’t alone. Every major chipmaker in the world is racing to build more capacity. We’ve reached the point where the physical ability to manufacture advanced chips is tapped out.
But here’s the part most investors miss: You can’t just “decide” to make more chips.
Building a fab isn’t like adding another warehouse. You can’t pour concrete, hire workers, and flip a switch.
You need the highly specialized equipment that goes inside the fab—the tools that actually manufacture the chips.
And those tools are among the most complex machines humanity has ever created.
- Why chipmaking equipment is the real story…
Chipmaking equipment accounts for roughly 70% to 80% of the total cost of building a new fab.
In other words, when a chipmaker announces a multibillion-dollar expansion, most of that money isn’t going toward the building itself.
It’s going toward the tools that make the chips.
The newest 2-nanometer fabs cost roughly $28 billion to build, compared with about $20 billion for the previous generation. Nearly all of that increase comes from more advanced equipment.
As chips shrink, the manufacturing process becomes exponentially more complex.
Consider ASML Holding NV (ASML), the Dutch company that makes extreme ultraviolet (EUV) lithography machines—the tools that use light to “print” circuits onto silicon wafers.
The wavelength of light involved is so small it gets absorbed by air. So the entire process must happen in a vacuum.
Each machine costs more than a Boeing 737 and takes up an entire room. Yet it’s precise enough to etch patterns smaller than a virus.
And here’s the kicker: As chips get smaller, the number of manufacturing steps roughly doubles. More steps mean more tools are required to produce the same number of chips.
A modern 2-nanometer chip can pass through more than 1,000 individual process steps before it’s complete. Each step must be nearly perfect. A single microscopic defect can ruin the entire chip.
So when global demand for AI chips explodes, it doesn’t just require more factories. It requires more of the equipment inside those factories.
That’s the new bottleneck.
- A map of where the money will flow…
Chip manufacturing is extraordinarily complex.
But at a high level, it breaks down into five core stages. And each stage has companies that dominate it.
- Lithography
This is where the process begins.
Circuit patterns are projected onto silicon wafers using extraordinarily precise light. ASML has a near-monopoly on the most advanced lithography systems. Without its tools, no one makes cutting-edge chips.
- Deposition and Etching
Once the circuit pattern is created, ultra-thin metal layers—sometimes just a few atoms thick—are laid down and then carved into circuits.
Modern chips can contain more than 100 layers, meaning these tools run 24 hours a day to keep production moving.
Applied Materials (AMAT), Lam Research (LRCX), and Tokyo Electron are major players here.
- Inspection and Metrology
At this scale, precision is everything.
Metrology tools measure chip dimensions down to fractions of a nanometer. Inspection tools detect defects—sometimes as small as a single dust particle.
Companies like KLA Corp. (KLA) and Nova Ltd. (NVMI) dominate this space, alongside specialists such as Onto Innovation (ONTO) and Camtek Ltd. (CAMT).
- Advanced Packaging
Once chips are manufactured, they must be assembled and packaged.
As AI systems grow more powerful, this stage has become increasingly complex. Chips must be stacked, connected, and assembled with extreme precision.
Companies like Amkor Technology (AMKR), Kulicke and Soffa Industries (KLIC), and BE Semiconductor Industries NV (BESIY) are critical players here.
- Materials and Contamination Control
Chip fabs are roughly 10,000X cleaner than a hospital operating room.
Companies like Entegris (ENTG), Ultra Clean Holdings (UCTT), and Ichor Holdings Ltd. (ICHR) provide ultra-pure materials, filtration systems, and gas delivery equipment needed to make defect-free chips.
Without them, even a $400 million lithography machine is useless.
- 2026: A record year in the making
With every major chipmaker expanding capacity at the same time, demand for chipmaking equipment is set to surge.
Wall Street calls these companies semiconductor capital equipment, or “semi-cap,” stocks.
They don’t get the same headlines as Nvidia. And they aren’t very flashy.
But when the world needs more AI chips—and right now, it does—these are the companies that make that possible.
The AI boom didn’t create this opportunity overnight. It has been building for years, one bottleneck at a time.
And the next bottleneck is pointing directly at the companies supplying the tools to make the chips.
If you liked this, consider signing up for my free Jolt investing letter, where I write about disruptive technologies and how to invest in them. Go here.
Stephen McBride
Chief Analyst, RiskHedge
