Inside Apple's US Chip Supply Chain Buildout

What Apple Showed Inside America’s Chip Supply Chain

WSJ reporter Rolfe Winkler toured Apple’s US production facilities alongside its partners. His report covers a wafer plant in Sherman, Texas, a TSMC semiconductor fab in Arizona, and a Foxconn assembly facility in Houston. WSJ article

Reading this report, what interested me wasn’t the dollar figure but which stages of the process can actually be completed in the US right now. The $165 billion TSMC announced includes plans it has yet to execute. Breaking ground on a factory and being able to reliably supply the chips you need are two different things, and they need to be looked at separately.

The Silicon Wafer You Need Before You Can Make a Chip

To make a silicon wafer, you first melt silicon that’s been purified to remove almost all impurities, then grow it into a single large crystal. The cylindrical result is called an ingot¹, and slicing it thin produces a wafer². After the surface is polished smooth, cleaned, and inspected, it gets shipped to chip fabs. SUMCO’s manufacturing process explanation

The Sherman, Texas plant featured in the report is run by GlobalWafers America (GWA), the US subsidiary of Taiwan’s GlobalWafers. Construction began in 2022, and the plant opened in May 2025. At the opening, the company announced it would expand its US investment plan to $7.5 billion — adding $4 billion to the existing $3.5 billion commitment. That doesn’t mean the additional amount has already been spent. GlobalWafers announcement

The US government’s CHIPS program³ funding, confirmed at up to $406 million, covers not just the Sherman plant but also facilities like the MEMC site in St. Peters, Missouri. US NIST notice

GlobalWafers describes this plant as the first new advanced silicon wafer production facility of its kind built in the US in over 20 years. According to NIST, roughly 90% of the world’s silicon wafer supply is concentrated in East Asia. That means ramping up chip production in the US also requires securing the supply of these foundational materials.

TSMC Arizona’s Current Production and Expansion Plans

The company responsible for etching circuits onto wafers is called a foundry⁴. Through multiple rounds of processes like photolithography, etching, and deposition, transistors⁵ and wiring are formed. One of the key pieces of equipment used to create these fine circuit patterns is ASML’s EUV lithography equipment⁶.

ASML’s equipment fires a laser at a tin droplet to create plasma, then uses the extreme ultraviolet light emitted from that plasma. Mirrors collect light with a wavelength of 13.5nm and transfer the circuit pattern onto the wafer. See ASML’s technical explanation for more detail — by the way, ASML happens to be my favorite stock.

In March 2025, TSMC announced it was expanding its U.S. investment plan to a total of $165 billion. The plan includes 6 semiconductor fabs, 2 advanced packaging facilities, and an R&D center, which the company described as the largest single foreign direct investment⁷ in U.S. history. TSMC’s announcement

As of early 2026, the company described the progress of its Arizona fabs as follows:

• Fab 1: N4 process mass production began Q4 2024
• Fab 2: Building construction complete. Equipment move-in and installation in 2026, targeting mass production in H2 2027
• Fab 3: Construction underway. Plans to introduce N2 and A16 processes

You can check these details in TSMC’s Q4 2025 earnings materials and the Arizona facility overview. Note that TSMC’s A16 here refers to a manufacturing process name, distinct from Apple’s A16 chip.

The numbers attached to process names like N4, N3, and N2 distinguish technology generations. They don’t mean that a specific component’s actual length is literally 4nm, 3nm, or 2nm. Newer processes generally help pack more circuitry into the same area or improve power efficiency, but the performance of a finished chip also depends on its design.

The Apple A16 mentioned in the report is a chip that’s already in products currently on sale. It would be a mistake to take the news that Fab 1 in Arizona started production as meaning that all of Apple’s latest chips are now made in the U.S. Different fabs adopt different manufacturing processes and can produce different products.

Also, after a chip is fabricated, it still needs to go through packaging⁸ — connecting it to external circuitry and protecting it — as well as testing. Even wafers made in the U.S. are sometimes sent to Asia for these downstream processes. Amkor’s advanced packaging facility in Arizona, where Apple is participating as the first major customer, broke ground in October 2025 and is targeting production in early 2028. Amkor’s announcement

The start of chip manufacturing in the U.S. is clearly progress. Still, it hasn’t yet reached the stage where manufacturing, packaging, and testing for all chips happen entirely within the U.S.

Server Assembly in Houston and Mac mini Production Plans

The process of assembling chips and other components into a finished product like a server, then testing and packaging it, is called FATP⁹. Foxconn’s Houston facility handles this stage. Apple has said it has been producing AI servers there since 2025, supplying them to its own data centers in the US. Apple’s February 2026 announcement

Apple also announced it will begin producing Mac mini units in Houston sometime in 2026. The plan is to build up more experience manufacturing certain finished products domestically. That said, localizing production of servers and Mac minis doesn’t mean the whole product lineup, including the iPhone, is moving to the US.

I read the decision to expand Mac mini production as an attempt to gain operational experience running an assembly facility in the US. Still, with production volumes and costs for each product undisclosed, it’s hard to flatly call this the “lowest-risk option.” We’d need to confirm how much can actually be produced and supplied.

I think growing demand for running AI agents on local devices could also factor into decisions about Mac mini production. Though rising interest and confirmed long-term sales demand are two different things worth watching separately.

Oswarld's Lens

I think this supply-chain disclosure carries both real production progress and a political message. There’s genuine advancement here — wafers made in America, chips produced, servers assembled. At the same time, as a company, Apple also needs to show the government and consumers that it’s expanding manufacturing on US soil.

Especially with tariffs and pressure to relocate production looming, I think Apple’s emphasis on US manufacturing serves a persuasive purpose aimed at outside audiences too. That said, this doesn’t make the actual volume being produced in these factories meaningless. To judge the real performance, you need to separate out the amount announced, the amount actually spent, and the production capacity actually running.

The 2024 report from SIA and BCG notes that the US share of global semiconductor manufacturing capacity fell from 37% in 1990 to 10% in 2022. Factoring in the investment plans announced at the time, it projected US manufacturing capacity would grow to roughly 3 times its 2022 level by 2032. That 3x figure isn’t an achievement already realized — it’s a projection contingent on those investment plans actually proceeding. Introduction to the SIA/BCG report

What I see as the more important goal is making sure that if supply gets cut off in one region, some volume can still be secured elsewhere. That’s a different scale and approach from the goal of producing all needed semiconductors domestically. I think the current wave of US investment is best understood as this kind of supply diversification.

A new factory isn’t ready to use the moment the building is finished. Equipment has to be installed, processes stabilized, and quality verified to customers’ specifications. So it’s hard to expect today’s announced investment to resolve any immediate supply disruptions.

The observation from the on-site report that the automated production floor didn’t have many people was striking too. But the number of people visible inside a factory alone can’t tell you the full employment impact — construction, equipment maintenance, R&D, and partner companies all need workers too. I think we should look at the employment effects, but also evaluate how much this reduces the risk of supply disruption as a core measure of the policy’s success.

Closing

More facilities in the US are producing wafers, fabricating chips, and assembling some finished products. Other processes, like advanced packaging, are still at the stage of building out facilities. Understanding this distinction keeps you from confusing capacity that’s actually available now with plans for the future.

Going forward, when I look at these investments, I plan to focus less on the number of factories and more on which products can be reliably made through which stages of the process. The real payoff of supply-chain diversification will show up not in the announced dollar figures, but in the production options actually secured.

References & Further Reading

• Rolfe Winkler, What I Saw Inside Apple’s U.S. Chip Supply Chain, WSJ, 2026: the on-the-ground reporting that sparked this piece.
• GlobalWafers’ U.S. plant opening announcement and NIST’s program overview: lay out the investment plan and scope of government support.
• TSMC’s expanded U.S. investment announcement, Q4 2025 earnings materials: confirm the investment plan and plant-by-plant progress as of early 2026.
• Apple’s Houston production expansion announcement, Amkor’s Arizona facility groundbreaking announcement: detail plans for the assembly and packaging stages.
• SIA/BCG, Emerging Resilience in the Semiconductor Supply Chain, 2024: projections on manufacturing capacity by country and supply-chain diversification.
• SUMCO’s wafer manufacturing process, ASML’s explanation of light and lasers: further detail on the technical terms used in the body text.

The author is Oswarld (Kwangseob Ahn). Current roles: Adjunct Professor at Sejong University, Strategy Consultant at INLEVEL9. Career, research, books, and recent work are kept current on the About page. Latest · July 2026: HEMA-2: A Consolidation-Aware Tri-Memory Architecture with Multi-Channel Scheduling for Lifelong Conversational AI.

Footnotes

1. Ingot: the large silicon crystal block that becomes the raw material for wafers. ↩

2. Wafer: a thin substrate on which semiconductor circuits are built. Silicon wafer diameter varies by use, with the 300mm standard widely used for advanced chip manufacturing. ↩

3. CHIPS and Science Act: a U.S. law enacted in 2022. It includes programs for semiconductor manufacturing support, R&D, and workforce development. Subsidies are disbursed based on project-specific conditions and procedures. ↩

4. Foundry: a company or business that manufactures semiconductors designed by other companies on a contract basis. TSMC is a leading example. ↩

5. Transistor: a semiconductor device that controls electric current. In digital chips it mainly functions as a switch, and chip performance depends not just on transistor count but on design and process as well. ↩

6. EUV lithography: a technology that transfers circuit patterns onto a wafer’s photoresist using extreme ultraviolet light with a 13.5nm wavelength. Actual structures are then formed through subsequent processes such as etching. ↩

7. Foreign Direct Investment (FDI): investment in a company in another country made with the intent of sustained managerial involvement. This includes not just building new plants but also methods like acquiring companies. ↩

8. Packaging: the process of protecting a semiconductor chip and configuring it to connect with external circuitry. Advanced packaging technologies also exist that connect multiple chips in close proximity. ↩

9. FATP (Final Assembly, Test and Pack): the process of assembling components into a final product, inspecting it, and packing it for shipment. The subject of this work differs from the packaging of the chip itself. ↩