Why the Intel Foundry Turnaround Thesis Is Still Wrong at $50

The bull case has been consistent: Intel is the only US-based pure-play leading-edge foundry. The CHIPS Act subsidies provide a capital tailwind that private competitors cannot match. The foundry business is launching customer wins (Microsoft, DoD, Pentagon contracts, multiple hyperscaler design services deals). And the transition to 18A node technology is tracking to schedule for volume production in 2026.

Each of those points has merit in isolation. The issue is the time-scale between narrative and cash flow. Foundry businesses require 4-5 years between committed customer wins and meaningful revenue contribution because of the chip design and qualification cycle. Intel's earliest external foundry customer revenue at scale is 2027-2028. Between now and then, the business absorbs substantial capex without earnings offset.

The market has accelerated the discounting. Intel now trades as if the foundry business is already generating the economics it is projected to generate in 2028. That is not how valuation works in cyclical capital-heavy businesses. It is how it works in speculative bubble markets.

Run the numbers. Intel's FY25 revenue was $52.9 billion. Gross profit was $18.4 billion, implying gross margin of 34.8%. That is a meaningful decline from the 53-55% gross margin Intel historically generated in its peak monopoly era.

Capital expenditure in FY25 was $14.6 billion. For context, the largest foundry operator in the world spent approximately $36 billion on capex in 2024. Intel's foundry capex at $14.6 billion is a fraction of that, and yet it represents nearly 28% of Intel's revenue. The capital intensity is not sustainable without a meaningful revenue inflection.

Free cash flow was negative $4.9 billion in FY25, after being negative $15.7 billion in FY24 and negative $14.3 billion in FY23. The three-year cumulative FCF is negative $35 billion. Intel has funded this through debt issuance, asset sales (Altera minority stake IPO), CHIPS Act grants, and draw-down of cash reserves. The funding gap is narrowing but has not closed.

The specific question for 2026: where does the cash come from if the foundry customer wins deliver revenue only in 2027-2028? The answer is more debt, more asset sales, and hoped-for CHIPS Act tranches. None of these are free, and all of them dilute shareholders in one form or another.

Historically, when capital-intensive technology businesses have rerated on turnaround stories before the financial statements confirmed the turnaround, the subsequent 18-month returns have been negative. Micron Technology 2015-2016 is the clearest analogue; the stock rallied on DRAM pricing thesis, then gave back 50% before the fundamental recovery materialised. The Intel setup carries a similar structural risk.

Revenue in FY21 was $79.0 billion. Revenue in FY25 was $52.9 billion. The business has shrunk by $26 billion over four years. That is not a turnaround; that is a still-shrinking business that has partially stabilised operating expenses.

Gross margin collapsed from 55.4% in FY21 to 34.8% in FY25. Every gross margin point lost on $53 billion of revenue is $530 million of gross profit that is structurally gone. Some of the compression is product mix; some is lower chip prices in a competitive PC and server market; some is the dilutive effect of foundry services revenue running at substantially lower gross margins than classic Intel product revenue.

Inventory levels tell a specific story. Intel has been working down inventory for four quarters; that helps cash flow temporarily but will reverse when the product demand picks up and restocking resumes. The FY25 capex of $14.6 billion was also a reduction from the FY24 $23.9 billion; some of that reduction reflects lower pace of tool purchases, not a lower required capex intensity. The steady-state capex for a leading-edge foundry is probably $20-25 billion annually, which means Intel's FY25 capex is temporarily below run-rate.

The market cap of $344 billion implies that this business is worth 6.5x FY21 peak operating income. That is a growth multiple applied to a company that has lost 33% of its revenue and 99% of its operating income over four years. The gap between the narrative and the data is the mispricing.