Amkor's $7 Billion Arizona Bet Is a Timing Game, Not a Demand Play

Amkor is spending $7 billion to build America's first high-volume advanced packaging facility in Arizona. The story the headlines tell is about AI demand pulling the semiconductor industry toward reshoring. The more useful way to look at it is as a supply-side timing bet. Amkor is leveraging nearly $5 billion in new debt to position itself for a packaging capacity squeeze that may ease by the time its doors open.

The numbers behind that bet are the story.

The constraint that drives the bet

Advanced packaging — the process that connects multiple chips together into higher-performance modules — has become the binding constraint in the AI supply chain. TSMC's CoWoS packaging capacity, the dominant source, was sold out through 2026. TSMC scaled CoWoS production from roughly 35,000 wafers per month in late 2024 to a projected 130,000 by late 2026, but allocation remains a three-dimensional problem: teams must lock wafer starts, packaging slots, and HBM supply simultaneously, and packaging is the tightest link.

That constraint is what Amkor's Arizona campus is designed to address. The $7 billion, two-phase project in Peoria will deliver over 750,000 square feet of cleanroom space. The first manufacturing facility is expected to complete in mid-2027, with production beginning in early 2028. It will be the first U.S.-based high-volume advanced packaging facility.

The question isn't whether advanced packaging capacity will be needed. It's whether the specific constraint Amkor is building for will still exist in its current form 18 months after production starts.

The revenue that gets you here

Amkor's recent financials show the demand side is real. Second quarter 2026 revenue hit a record $1.90 billion, up 26% from the year-ago quarter. Advanced products — flip chip, memory, and wafer-level processing — generated $1.557 billion, or roughly 82% of total sales. Mainstream products contributed the remaining $341 million.

Customer concentration is narrowing, at least slightly. The top 10 customers accounted for 66% of net sales in Q2 2026, down from 72% a year earlier. Automotive and industrial hit record revenue, up approximately 38% year over year. Computing also hit record quarterly levels, rising approximately 26%.

The profit numbers are less convincing. Gross profit rose 75% year over year to $318.6 million, pushing gross margin to 16.8%. But the operating margin for the trailing twelve months sits at roughly 7.6%. This is a capital-intensive business running on thin margins, and those margins are about to come under pressure from the very expansion that is supposed to solve the capacity problem.

The capital expenditure that changes the picture

Here is where the structural story matters more than the demand story. Amkor's capital spending has accelerated sharply. Trailing twelve-month capex was roughly $744 million in fiscal 2024, rose to $905 million in fiscal 2025, and reached approximately $1.05 billion through the first half of 2026. Full-year 2026 guidance puts capex between $2.5 billion and $3.0 billion — roughly four times the 2024 run rate.

That acceleration has drained free cash flow. Free cash flow for the trailing twelve months fell to $167 million in early 2026, down from $345 million in fiscal 2024 and $191 million in fiscal 2025. The free cash flow margin has compressed from roughly 5.5% of revenue to 2.4%.

To fund the gap between operating cash flow and capex, Amkor borrowed heavily. Total debt rose from $2.76 billion at fiscal year-end 2024 to $5.03 billion as of June 2026 — a nearly 82% increase in roughly 18 months. Proceeds from long-term debt in the first six months of 2026 alone totaled $1.15 billion against just $80.3 million in repayments. The debt-to-equity ratio nearly doubled from 0.28 to 0.53 over the same period.

The CHIPS Act helps but doesn't transform the picture. Amkor was awarded up to $407 million in direct funding, with total federal support potentially reaching $600 million when tax credits are included. That is less than 9% of the $7 billion total project cost. The company is self-funding the overwhelming majority of this expansion.

The timing problem

Amkor's production doesn't begin until early 2028. That timing sits inside a window where the packaging constraint may be evolving in ways that don't favor a new entrant.

TSMC is scaling CoWoS aggressively. The projected move to 130,000 wafers per month would represent a nearly 4x increase from late 2024 levels. If that ramp stays on track, the packaging shortage that exists today may be substantially eased by 2028.

More importantly, the constraint is migrating. The semiconductor supply chain doesn't have one bottleneck — it has competing bottlenecks that shift over time. The constraint moved from wafer fabrication to advanced packaging and is now moving toward power delivery and high-bandwidth memory. Where the constraint sits determines who has pricing power. A facility that opens when the constraint has moved elsewhere enters the market with high fixed costs and no scarcity premium.

Amkor's financial structure compounds this risk. Operating cash flow of roughly $1.2 billion in the trailing twelve months is being absorbed by $1.4 billion in capex. At the guided 2026 run rate of $2.5-3.0 billion in annual capex, free cash flow turns deeply negative and stays there through the construction phase. Debt service on $5 billion in obligations requires stable margins that the company has not yet demonstrated.

The peer comparison

The stock is trading at a forward P/E ratio near 23x, with an EV/EBITDA of roughly 9.5x. Those are not cheap multiples for a business with a 14% gross margin, a 2.4% free cash flow margin, and debt that doubled in 18 months. The stock is up approximately 96% over the past year and 28% year-to-date.

What the market is pricing in is not today's business. It's the assumption that Amkor will capture meaningful share of advanced packaging demand when the Arizona facility opens, that margins will expand as it transitions from mainstream to advanced mix, and that the debt load will be serviced by a revenue base that continues growing at mid-to-high-teens rates.

Each of those assumptions requires the packaging constraint to hold — or to shift to a segment where Amkor competes. The company's recent revenue mix shows advanced products at 82% of sales, which is the right trajectory. But revenue mix today doesn't guarantee pricing power tomorrow, particularly in a market where TSMC's capacity is expanding from the same base that created today's shortage.

The implication

This is not a demand story. Amkor's customers are pulling because TSMC can't deliver, not because the underlying chip orders suddenly multiplied beyond reason. The demand exists. The scarcity is the price premium.

What the investor needs to track is whether scarcity persists. The key issue isn't whether Amkor will deliver on its construction timeline — that's an execution question with visible milestones. The more important question is whether the packaging capacity that constrains the industry today will still be the constraining packaging capacity when the Peoria facility ramps in 2028, and whether TSMC's scaling will have compressed the premium that justifies the capex Amkor is undertaking.

The stock's performance this year prices in a favorable outcome. That pricing is not unreasonable — it's just conditional. The condition is that the constraint doesn't migrate faster than Amkor can build.