Inside Intel's comeback: government equity stake, 20,000 layoffs, and a bet on 14A to rival TSMC

Intel's 10% Government Stake and the Race to Build a Foundry Rival to TSMC

The U.S. government now owns 10% of Intel—an unprecedented equity intervention in a major American company that fundamentally shifts how the semiconductor industry operates domestically. The Treasury converted $3.2 billion in Defense Department contracts and billions in manufacturing grants into stock, according to an internal memo disclosed in shareholder litigation. The structure was non-negotiable. An administration official told Intel's CFO David Zinsner bluntly: this is how it happens, or it doesn't.

The equity deal closes a chapter on Intel's five-year decline. In 2024, the company posted $18.8 billion in losses. Its AI chip strategy had stalled. A critical partnership with ARM—which would have used Intel's foundry to manufacture ARM's data center processors—collapsed because Intel's processes weren't competitive enough. Meanwhile, customers including Apple, Nvidia, AMD, and Qualcomm had defected to TSMC, which now manufactures more than 90% of the world's most advanced chips.

Intel's new CEO Lip Bu Tan, who arrived in March 2025, inherited a company with a culture problem. For decades, Intel dominated through vertical integration so complete that customers took what Intel offered on Intel's terms. That model doesn't work in foundry. TSMC's flexibility and customer-first engineering are now the baseline. Tan's diagnosis was immediate: radical restructuring or slow death.

The 20,000-person reckoning

Tan cut more than 20,000 jobs—roughly one-fifth of the workforce—in his first six months. He pared capital spending and sold stakes in Altera and Mobileye for $5.2 billion combined. He kept only two members of the inherited leadership team: CFO David Zinsner and Naga Chandrasekhar, who runs the foundry operation. Everyone else was replaced. Two former Cadence colleagues joined to lead engineering and government technology groups. A senior ARM executive was appointed to lead the data center business.

The government equity stake arrived in August 2025, after Trump initially called for Tan to resign over his prior Chinese chip investments. Tan spent a weekend mapping outcomes, then met Trump, Commerce Secretary Howard Lutnick, and Treasury Secretary Scott Besson. The Malaysia-born executive's task was to convince the administration he was both patriotic and capable. The equity deal sent a two-year signal to prospective customers: Washington has Intel's back.

The 14A bet

Intel's future hinges on whether its 14A manufacturing process can match TSMC's capability and cost. This is not speculation. Tan has committed the company fully to 14A development after previously warning it might abandon the node. The October release of 14A's design kit will determine whether major customers actually commit to building chips at Intel rather than continuing to rely on TSMC.

The obstacle is neither technical ambition nor capital. It is customer risk. For Broadcom, MediaTek, Apple, or any major fabless designer, moving wafer volume to Intel means fighting for TSMC's remaining capacity in a period of chronic shortage. TSMC has leverage, and customers know it. As one industry source puts it: people don't want to upset TSMC because of the capacity crunch. Everyone is fighting for wafers.

Intel and Apple have found a workaround. Trump posted on Truth Social in June that Apple had agreed to help design and build chips in America—and neither company confirmed it. The silence is strategic. Apple is testing Intel's processes on older M-series laptop chips, a credibility-building exercise that avoids the appearance of abandoning TSMC for newer, higher-margin designs. Similarly, MediaTek was the first to announce use of both Intel and TSMC packaging technology, which builds trust without full dependence.

The foundry as secondary play

Intel has also begun leveraging its packaging business—the low-margin but trust-building work of encasing wafer dies into finished products. It had ignored this advantage for years to focus on high-risk, high-reward fabrication. Now it matters. The New Mexico packaging facility accelerated investment from the start.

Alongside foundry, Tan is rationalizing Intel's AI chip division, which has consistently disappointed against Nvidia. He has hired Broadcom and Marvell, which rose on AI custom-chip wins with Microsoft, Amazon, and Google. Intel signed its own custom-chip deal with Google in April. The company has also benefited from CPU demand—the Clearwater Forest chip launched in June serves AI workload management in data centers, a category now critical as agentic AI agents demand constant compute.

What comes next

Since the government stake arrived, Intel has raised capital spending to $20 billion this year and opened talks with multiple customers about foundry use. The company began manufacturing some of its own advanced PC and server chips at its Arizona facility, a sign it now trusts its own processes.

The core tension remains: Tan has not yet provided public evidence that 14A matches TSMC's technology in performance and power efficiency. One industry source said 14A is improving but not yet equal. Tan talks about ramp rate but avoids discussing competitiveness versus TSMC's equivalent node. Intel never discloses yield or other technical specifics that would allow independent verification.

The federal stake also carries punitive terms designed to deter Intel from abandoning foundry. That structure solves a political problem but creates a business one: committing $20 billion to a node whose competitive position remains ambiguous, while customers tread carefully to avoid upsetting the supplier they still need most.

For now, Tan is moving the needle on foundry rather than chasing points on the board through small partnership announcements. The needle is the fab business itself—years of investment in advanced fabs, inching forward against TSMC's decade-long lead. October's 14A design kit release is the next proof point. Without it, the government stake becomes a subsidy for managed decline, not a bet on American semiconductor manufacturing.