Inside America's $165 Billion Bet to Rebuild Chipmaking · Issue #55

Opening

Dear reader, a Wall Street Journal reporter recently accepted an invitation from Apple to tour America’s semiconductor supply chain facilities firsthand — from the silicon wafer plant in Sherman, Texas, to TSMC’s fab in the Arizona desert, to Foxconn’s assembly line in Houston. This wasn’t just a factory tour. It was the first time the process of rebuilding America’s 30 lost years of manufacturing capacity was opened up to outside eyes.

But as I read this report, I found myself more curious about the structural question hiding behind the flashy figures. Even after pouring in $165 billion, can America really replace Taiwan?

The Start of the Supply Chain — From Sand to Wafer

The first step in the semiconductor supply chain is surprisingly primitive. High-purity silicon extracted from sand in North Carolina is melted at 2,500°F (about 1,370°C) and grown into a cylindrical ingot¹. This is sliced with a wire saw into 12-inch wafers², then polished, inspected, and packaged before being sent on to the next stage.

This work is handled by GlobalWafers America (GWA) in Sherman, Texas. A subsidiary of Taiwan’s GlobalWafers, it broke ground in 2022 and officially opened in May 2025. The investment grew from an initial $3.5 billion to a total of $7.5 billion with an additional $4 billion, and the facility also received $406 million in CHIPS Act³ funding.

There’s one point worth noting here. This is the first time in 20 years that a new 300mm advanced silicon wafer fab has opened in the United States. About 90% of the world’s silicon wafers were being produced in East Asia. The fact that even the most basic material in the supply chain was import-dependent shows just how much this reshoring[4] effort has to start from scratch.

The Heart of the Fab — What TSMC Arizona Can and Cannot Make

The heart of the supply chain is the foundry⁴ — the process of etching trillions of transistors⁵ onto a wafer. The machine that makes this possible is ASML’s EUV (extreme ultraviolet) lithography⁶ equipment, which costs between $220 million and $380 million per unit. It fires a laser at droplets of tin to create extreme ultraviolet light that doesn’t exist in nature, then bounces it off mirrors polished to atomic-level smoothness to etch chip circuitry onto the wafer. Calling this an exercise in testing the limits of physics isn’t an exaggeration. (For what it’s worth, ASML is my favorite stock.)

TSMC plans to invest a total of $165 billion in the Arizona desert to build six fabs, two advanced packaging facilities, and one R&D center. It’s the largest foreign direct investment (FDI)⁷ in U.S. history. Here’s where things currently stand:

Fab 1: Began mass production of the 4nm (N4) process in Q4 2024 — currently producing Apple’s A16 chip and Nvidia’s Blackwell GPU
Fab 2: Construction complete, equipment installation in H2 2026 → targeting 3nm (N3) mass production in 2027 (moved up from the original 2028 target)
Fab 3: Broke ground in April 2025, targeting 2nm (N2) and A16 processes → production target of 2029 *Note: The nm process (nanometer process) is a unit describing how fine the circuitry etched onto a chip is. The smaller the number, the finer the process and the better the performance. Technology progresses from 4nm → 3nm → 2nm, and 1nm is about 1/100,000th the width of a human hair. That said, it doesn’t correspond exactly to physical size anymore — it’s closer to a marketing label distinguishing generations.

But this is where a structural limitation shows up. What TSMC Arizona currently makes is the A16 chip — the chip that goes into the iPhone 15 and the entry-level iPad. The latest iPhone 17’s A19 chip uses a more advanced process that can only be produced in Taiwan. And wafers that come out of the fab still have to be sent back to Asia to be cut into individual chips and packaged⁸, because there’s no advanced packaging facility in the U.S. yet.

Put simply: America has started “baking” chips, but making the latest chips or assembling them into finished products still depends on Taiwan.

The Last Piece of the Puzzle — The Reality of the Assembly Line

The final stage of the supply chain is FATP (Final Assembly, Test and Pack)⁹. At a Foxconn facility north of Houston, Apple is assembling roughly 10 AI servers per hour. The plan is to add a Mac Mini production line as well — by converting an empty 220,000-square-foot (about 20,000 m²) warehouse.

CEO Tim Cook has said Apple is “deeply committed to the future of American manufacturing,” and the company is pursuing this expansion as part of its $600 billion, four-year U.S. investment plan. But a few numbers are worth examining:

Mac Mini accounts for less than 5% of total Mac sales — and about 1% of Apple’s total revenue**.**
Assembly facilities in Asia employ hundreds of thousands of workers, while Houston’s operation is measured in the hundreds
iPhone assembly still takes place in China and India The symbolism here is significant, but the scale tells a different story. For Apple, the Mac Mini is the lowest-risk test case. Production volume is small, so even if something goes wrong, the impact on revenue is minimal. It’s a strategy that delivers the political message of “Made in America” without touching the core supply chain. Of course, this is a strategy that could change dramatically if the agent boom sparked by OpenClaw and the runaway success of the MacBook Neo push Mac Mini demand much higher.

Oz’s Lens

Honestly, I think the picture this report paints is half genuine, half politics.

Companies usually open up their supply chains to the press for one of two reasons. First, when they genuinely have something to brag about. Second, when they want to signal to stakeholders that “we’re working hard on this.” Apple’s tour this time looks closer to the second. Amid the Trump administration’s tariff pressure and reshoring demands, Apple is putting TSMC and Foxconn front and center to build the narrative that “a domestic supply chain exists.”

But the data is unsentimental. America’s share of global semiconductor manufacturing fell from 37% in 1990 to about 10-12% in 2022. Private investment triggered by the CHIPS Act has topped $540 billion, with a plan to triple U.S. chip manufacturing capacity by 2032. Yet it could take more than 10 years for TSMC Arizona’s monthly output to reach Taiwan’s level. And TSMC’s own analysis suggests that building a fab in the U.S. costs 4 to 5 times more than in Taiwan.

The core question, as I see it, is this: Does America want semiconductor “self-sufficiency,” or does it want “insurance”? If it’s the former, the current pace falls hopelessly short. If it’s the latter, the direction is right, but expectations need managing. If a Taiwan Strait crisis or a major earthquake happens tomorrow, a fab that just broke ground today won’t help at all. Semiconductor reshoring is a decade-scale infrastructure project, not a short-term crisis response.

One more thing — the WSJ’s observation that these plants had almost no people in them is telling. Semiconductor manufacturing is highly automated. America isn’t trying to reclaim this industry for mass employment; it’s trying to close a strategic vulnerability. This is security policy, not jobs policy. Being clear about that distinction is what will narrow the gap between expectations and reality.

Closing

Rebuilding has begun across every stage of America’s semiconductor supply chain — wafer, foundry, and assembly — but a clear gap with Asia persists at each step
Apple and TSMC’s investments are real, but the most advanced processes and packaging still depend on Taiwan — what America can currently make is still a generation behind
Understanding this as a 10-to-20-year infrastructure project rather than a short-term response is what lets you tell real progress apart from political packaging If you’d like to dig deeper into this topic, the Carnegie Endowment report below does a good job summarizing the structural limits of the CHIPS Act. It should help you understand not just the “why” of reshoring, but also “what’s still missing.”

References & Further Reading

Christopher Mims, “What I Saw Inside Apple’s U.S. Chip Supply Chain”, The Wall Street Journal, 2026. : This is the report article that sparked today’s newsletter. It’s a rare on-the-ground account of each stage of the supply chain.
Semiconductor Industry Association, “2025 State of the U.S. Semiconductor Industry”, 2025. : An annual report that lays out the current state of the U.S. semiconductor industry in data. It shows the decline in manufacturing share and the outlook for recovery.
TSMC Arizona official page. https://www.tsmc.com/static/abouttsmcaz/index.htm : You can check the six-fab plan, investment scale, and process roadmap directly.
Carnegie Endowment for International Peace, “After the CHIPS Act: The Limits of Reshoring and Next Steps for U.S. Semiconductor Policy”, 2022.
Deloitte, “2026 Global Semiconductor Industry Outlook”, 2026.

The author, Kwangseob Ahn, is a professor of business administration at Sejong University and lead consultant at OBF (Oswarld Boutique Consulting Firm). He teaches statistics and data analysis — business data management and business analytics — while leading GTM and AI strategy consulting in the field, designing the seam between technology and business. He has published academic research on a memory architecture for AI dialogue systems (HEMA) and runs Daily Arxiv, a daily curation of global AI papers. He holds a master’s from Korea University’s Graduate School of Technology Management and a KMBA. He is the author of Homo Brainless: The People Who Outsource Their Thinking.

Footnotes

Ingot: A cylindrical mass grown from melted silicon. It’s sliced thin to make wafers. You could call it the starting point of a semiconductor. ↩
Wafer: A round silicon plate sliced thin from an ingot. It serves as the “canvas” for semiconductor chips. 12 inches (300mm) is the current cutting-edge standard, and hundreds of chips are made from a single wafer. ↩
CHIPS Act (CHIPS and Science Act): A semiconductor promotion law passed in the U.S. in 2022. Worth about $52.7 billion, it invests in semiconductor manufacturing subsidies, R&D, and workforce training. Simply put, it’s a law that says, “If you build a chip-making factory in the U.S., the government will help fund it.” ↩
Foundry: A specialized manufacturer that doesn’t design chips itself but produces chips designed by other companies. TSMC is the prime example. Think of it like a restaurant: the chef (Apple, Nvidia) creates the recipe, while a specialized kitchen (TSMC) does the cooking. ↩
Transistor: A tiny switch that turns the flow of electric current on and off. It’s the most basic building block of a semiconductor chip, and the more of them packed in, the better the chip’s performance. The latest chips contain anywhere from billions to trillions of them. ↩
EUV Lithography (Extreme Ultraviolet Lithography): Equipment that uses extreme ultraviolet light to etch circuit patterns onto wafers. Only the Dutch company ASML can make it, and a single unit costs hundreds of billions of won. It’s the most critical and most expensive piece of equipment in semiconductor manufacturing. ↩
FDI (Foreign Direct Investment): Investment in which a foreign company directly establishes a factory or business in another country. Unlike portfolio investment, which simply involves buying stock, this refers to building and operating physical facilities. ↩
Packaging: The process of cutting individual chips from a wafer, mounting them onto a substrate so they can connect to external circuitry, and covering them with protective material. Advanced packaging has evolved into a technology that bundles multiple chips together to boost performance, and in recent years it has become just as important a competitive edge as chip design itself. ↩
FATP (Final Assembly, Test and Pack): The final stage of the semiconductor supply chain. It’s the process of assembling chips and components into products, testing whether they work properly, then packaging and shipping them. It’s a more labor-intensive process than semiconductor manufacturing itself. ↩