A semiconductor fab (short for fabrication plant) is a factory that manufactures computer chips. These facilities take thin discs of silicon and, through hundreds of precise chemical and physical steps, transform them into the processors, memory chips, and sensors that power everything from smartphones to data centers. Fabs are among the most expensive and technically demanding factories ever built, with leading-edge facilities now costing tens of billions of dollars.
From Silicon Wafer to Finished Chip
The basic input of a fab is a silicon wafer, a flat, circular disc that looks a bit like a shiny dinner plate. The current industry standard is 300 mm (about 12 inches) in diameter, which became dominant in the early 2000s for high-volume, advanced chip production. Older 200 mm wafers are still widely used for less complex chips like automotive sensors and power management components. A proposed jump to 450 mm wafers has stalled due to extreme costs and a lack of standardized equipment, so 300 mm remains the workhorse.
Inside the fab, each wafer passes through more than 1,000 individual processing steps over a period of weeks. The core cycle repeats in layers: a light-sensitive coating is applied to the wafer, a pattern is projected onto it using ultraviolet light (a process called lithography), unwanted material is etched away, and new materials are deposited. This cycle builds up the billions of tiny transistors that make a chip work. A single 300 mm wafer can yield hundreds of individual chips, which are later cut apart, tested, and packaged.
Why Cleanrooms Matter
Chip features are now measured in single-digit nanometers. For perspective, a human hair is roughly 70,000 nanometers wide. At that scale, a single speck of dust landing on a wafer can destroy an entire chip. That’s why the heart of every fab is its cleanroom, a sealed environment where the air is filtered to remove nearly all airborne particles.
Cleanrooms are rated on an ISO scale from 1 (cleanest) to 9. The most sensitive areas of a fab, where lithography and material deposition happen, typically require ISO Class 3 or 4, meaning fewer than a handful of particles larger than a fraction of a micrometer per cubic meter of air. General manufacturing areas operate at ISO Class 5 or above. Workers inside these zones wear full-body suits (sometimes called “bunny suits”) that cover skin and hair completely, and they pass through airlocks to enter.
The Equipment Inside
A modern fab contains more than 1,200 specialized tools, but the most critical and expensive is the lithography machine. The latest generation uses extreme ultraviolet (EUV) light with a wavelength of just 13.5 nanometers to project circuit patterns onto wafers. Only one company in the world, ASML in the Netherlands, makes these machines. A single EUV system is estimated to cost upward of $150 million, and the most advanced models cost significantly more.
EUV lithography is what makes today’s smallest chips possible. Chipmakers use these systems to print the most intricate layers of their 7 nm, 5 nm, and 3 nm chips, while older deep ultraviolet (DUV) tools handle the less complex layers. ASML’s next-generation platform, called High-NA EUV, is designed to push production to 2 nm and beyond starting in 2025 and 2026.
Beyond lithography, fabs rely on etching tools that carve patterns into silicon, deposition systems that lay down ultra-thin films of various materials, ion implanters that alter the electrical properties of specific regions, and inspection systems that scan every layer for defects. Each of these tools costs millions of dollars and requires constant calibration.
Engineering Against Vibration
When you’re printing features a few nanometers wide, even the faintest vibration from a passing truck or a nearby HVAC system can ruin a chip. Fabs use extraordinary structural engineering to prevent this. Site selection itself is the first line of defense: engineers look for locations with stable bedrock, low seismic risk, and distance from highways or rail lines.
The buildings themselves often use a “building within a building” approach. Critical equipment sits on massive concrete inertia blocks that can weigh hundreds of tons, providing a platform so heavy it naturally resists motion. These blocks are isolated from the surrounding structure using elastic or pneumatic mounts that absorb vibrations before they reach the equipment. Some fabs use floating floors resting on springs or rubber pads, and the most sensitive equipment spaces may be enclosed in fully independent “box-in-box” rooms that are structurally disconnected from the rest of the building.
For low-frequency vibrations that passive methods can’t handle, active feedback systems use sensors to detect motion in real time and actuators to produce equal and opposite forces, effectively canceling out the vibration before it affects anything.
Cost and Construction Timeline
Building a leading-edge fab is one of the most capital-intensive undertakings in any industry. TSMC’s expanding operations in Phoenix, Arizona, illustrate the scale: the company’s total planned investment in the U.S. has reached $165 billion, covering multiple fabrication plants, advanced packaging facilities, and a research center. That figure represents the largest single foreign direct investment in U.S. history.
Construction timelines are measured in years, not months. Between 1990 and 2020, roughly 635 new fabs were built worldwide, with an average construction time of about 682 days (just under two years) from breaking ground to the start of production. That figure doesn’t include the permitting and planning phase that precedes construction. In the United States specifically, build times have gotten longer over the decades, rising from an average of 1.8 years in the 1990s to 2.5 years during the 2010s, a 38 percent increase driven largely by regulatory complexity.
Who Operates the World’s Fabs
The fab landscape is dominated by a small number of companies, and the concentration at the top is striking. TSMC (Taiwan Semiconductor Manufacturing Company) held a 67% revenue share of the global foundry market in the fourth quarter of 2024, driven by massive demand for AI chips. Samsung Foundry was a distant second at 11%, followed by China’s SMIC at 5%.
These companies operate as “foundries,” meaning they manufacture chips designed by other companies. Apple, Nvidia, AMD, and Qualcomm all design their own processors but rely on foundries (primarily TSMC) to actually produce them. A smaller number of companies, most notably Intel and Samsung, both design and manufacture their own chips in-house while also offering foundry services to outside customers. This model is called an integrated device manufacturer.
The geographic concentration of advanced fabs, with the vast majority located in Taiwan and South Korea, has become a major geopolitical concern. That’s a key reason governments in the United States, Europe, and Japan are now offering billions in subsidies to attract new fab construction on their own soil.
Why Fabs Take So Long to Build
The sheer complexity of a fab explains why you can’t simply throw money at the problem and speed things up. The cleanroom infrastructure alone requires specialized construction. Thousands of tools must be installed, connected to ultra-pure water systems, chemical delivery networks, and exhaust handling systems, then individually qualified. The supply chain for fab equipment is itself constrained: lead times for EUV machines and other critical tools can stretch well over a year.
Once tools are installed, the fab enters a “ramp” phase where engineers run test wafers to calibrate every process step and push yields (the percentage of working chips per wafer) high enough for commercial production. This ramp can take an additional six to twelve months. From the first permit application to the first chip shipping to a customer, five years or more is common for a brand-new facility.