Inside the new $50 billion chip factory in Texas
Step by step, what it takes to build the most advanced chip factory on Earth. And compete with TSMC.
Something colossal is rising from the Texas plains: a $50 billion microchip factory. One of the most expensive and complex projects in U.S. history.
Today we’ll break down:
what it really takes to build the most advanced chip factory on Earth;
why even billions of dollars can’t guarantee success;
and the key takeaways from this story.
What’s happening with Samsung?
On paper, this new chip factory is a marvel, built to rival TSMC’s new mega fab in Arizona, just two states away. But before it even came to life, it is already failing. To understand why, we have to go back to the company behind it - Samsung - and the moment it almost conquered the chip world. A decade ago, it was unstoppable.
Samsung was, and still is, the second-largest chipmaker on the planet, and the only one still keeping pace with Taiwan Semiconductor Manufacturing Company, or TSMC.
Samsung’s dominance in memory is nothing short of legendary - from DRAM in every server and smartphone to NAND that stores the world’s data.
And memory wasn’t enough. They also built logic chips - the brains behind Apple’s first iPhones, NVIDIA’s early GPUs, and Tesla’s first Autopilot computer. By 2015, Samsung was on top, mastering 14nm, landing Apple, and standing just behind TSMC. For a moment, it looked like they might even overtake TSMC. But then everything turned.
Apple became a direct rival, and trusting Samsung to build iPhone chips was like asking your enemy to sharpen your sword. So Apple left, moving all chip production to TSMC, a foundry that built for everyone but competed with no one. Then came the cracks from within. Unlike TSMC, Samsung was pulled in every direction, building chips, but at the same time displays and smartphones.
So their 10-nanometer process slipped behind schedule. Yields crashed. By the time TSMC ramped 5- and 3-nanometer production, Samsung was still fixing 7. Yields fell to around 40%. NVIDIA walked away, then Qualcomm, then Tesla. Each loss drained billions, and confidence.
It wasn’t just the logic side that struggled - Samsung’s memory business was shrinking too. Micron and SK Hynix kept taking a bigger piece of the pie.
And that’s when Samsung decided to do something no one expected - rebuild its empire on the other side of the world.
In 2021, they placed a $17 billion bet to rebuild their legacy, but not in Korea, but in the heart of Texas. A factory so ambitious, it was announced as “the most advanced semiconductor factory ever built on U.S. soil.”
The Fall
The company chose Taylor, a quiet town northeast of Austin. Far from Pyeongtaek, but close to its biggest customers: NVIDIA and Tesla. Taylor was the perfect choice - close to Tesla, Qualcomm, and Google, and surrounded by a booming tech scene hungry for AI silicon.
Even the land itself seemed made for it: vast, flat, and geologically stable, far from earthquakes or floods, with almost zero risk of natural disasters. Perfect for massive cleanrooms and billions of dollars in equipment.
On paper, the plan looked simple. Break ground in 2021. Start 4-nanometer microchip production by 2024, and push toward 2-nanometer soon after.
How to Build a Fab
And here’s where the story gets truly insane, because building a chip fab isn’t just engineering, it’s choreography at the atomic level.
Imagine you are about to build the most advanced semiconductor factory on Earth. You don’t start with concrete, not with steel, not even with machines. You start by locking direction.
Step one: lock the node, the transistor generation your fab will produce. And lock the customer, the company whose chips you’ll build. That first customer’s design is how the fab calibrates and optimizes every step - it’s the key to yield and performance.
That first step defines everything that follows: the tools, geography, recipes, floorplan, power grid, the entire playbook.
First, Samsung decided for 4-nanometer, a safe, mature node. But just as construction began, the world hit fast-forward. AI workloads exploded, and every major chipmaker started chasing 3nm and below. TSMC was already ramping Apple and NVIDIA chips on its newest process. Samsung didn’t have an anchor customer of that scale.
So Samsung made a fateful decision: pivot Taylor to 2 nanometers. It sounded bold. In reality, it was catastrophic. That single change turned a controlled build into chaos.
It meant new tools, new recipes, and a steep new learning curve. Even in Korea, Samsung was still struggling with the 2-nanometer node. The math blew up overnight. Because moving from 4 to 2 nanometers isn’t just swapping a few machines, it’s reinventing the entire factory.
The original $17 billion plan ballooned to $50 billion. Every change pushed the schedule out - and the budget through the roof.
Ground
And that was before they even broke ground. Because before you build the walls of a chip factory, you have to conquer the earth beneath it.
The soil under Taylor looks calm from above, but it’s not. It’s caliche - hard, dry, and uneven. Strong enough to hold a highway. But not stable enough to hold a semiconductor fab still.
And when you’re printing transistors smaller than a human eye can see, even the tiniest vibration can destroy everything. That’s why absolute stillness is fundamental.
Inside an EUV lithography machine, the tool that prints those transistors, every mirror must stay perfectly still, within just a few nanometers. If the floor vibrates or moves, even slightly, the laser misses its mark - and an entire batch of wafers, worth millions, is gone.
To fight the unstable Texas clay, they built one of the most extreme foundation systems ever made for a semiconductor fab. Every part was designed to keep the entire building perfectly still, down to the scale of atoms. They drilled more than twenty thousand shafts, each about 110 feet (34 meters) deep, and filled them with over half a million cubic yards of concrete.
Five concrete plants were built right on site just to feed that demand. That’s enough material to build several skyscrapers, but this time, they poured it downward instead of upward. Because every inch mattered - one wrong vibration could wipe out months of work.
Standard fabs use reinforced slabs with local vibration damping under specific tools. Samsung took a different path, it turned the entire foundation into a massive floating platform. Those deep piers anchor directly into bedrock, isolating the cleanroom from the shifting Texas soil above.
The result is a “building within a building” - a floating floor that absorbs every shock before it reaches the chipmaking tools. Here they achieved not just stability, but absolute stillness.
What 2nm Really Means
It’s almost ironic, this colossal project, billions of dollars, thousands of workers, mountains of concrete, an entire city of machines, all built to create the smallest devices on Earth.
These tiny new transistors are a huge deal. Every major customer wants them - Tesla, AMD, Qualcomm - because they’re not just smaller, they redefine how computation works at the atomic level.
For seventy years, the transistor has been the beating heart of technology. But now, at 2 nanometers, that transistor itself is changing shape, and it’s only the second time in history that’s ever happened.
The last time was over a decade ago, when the world moved from flat, two-dimensional transistors to FinFETs. But at two nanometers, even FinFETs can’t keep up. At this scale, electrons stop acting like particles, they ripple and interfere like waves. So engineers had to reinvent the transistor itself. They called it Gate-All-Around, a radical new design.
Instead of one tall fin, it uses multiple horizontal nanosheets - tiny ribbons of silicon, each just a few atoms thick, wrapped completely by the gate that controls them. And that’s what makes GAA so powerful.
But making these transistors isn’t easy — it’s atomic surgery. Each nanosheet has to be formed, stacked, and aligned with angstrom-level precision - that’s one-tenth of a nanometer.
And the only tool capable of printing them is EUV lithography - Extreme Ultraviolet Light - fired by lasers bouncing off mirrors polished smoother than any surface on Earth. If the mirror shifts by a few atoms, if the surface vibrates, the pattern fails. The wafer is gone. And that’s just the beginning.
This is what Samsung is trying to pull off inside the Taylor fab. That’s why so much engineering goes into building the fab - and that’s why they need this floating foundation to kill every possible vibration.
Below we’ll break down five more pillars of building a semiconductor fab. Who pulled Samsung’s project back from the brink. And the key takeaways that matter.
