Intel's 18A Gambit Pays Off: Process Leadership Returns, But the Foundry P&L Bleeds On

Intel's 18A node is in high-volume manufacturing for the first time since 2016, reclaiming process leadership over TSMC's N2. Under CEO Lip-Bu Tan, the stock is up 83% YoY—but the Foundry segment still lost $10.3 billion in FY2025, and the profitability mission has barely begun.

INTC · Information Technology · March 27, 2026

S&P 500 Position

Intel sits in the Information Technology sector's semiconductor sub-industry, dwarfed by NVIDIA (~$4T market cap) and TSMC (~$1.8T) but well ahead of Texas Instruments (~$169B). AMD trades at a significantly higher valuation on a revenue basis. Intel's $220B market cap reflects its turnaround-stage status—the market prices it as a 'show me' story despite being the only U.S.-based company with both leading-edge logic manufacturing and advanced packaging capabilities. The U.S. government's 8.4% beneficial ownership stake (via CHIPS Act equity) makes Intel unique among S&P 500 semiconductor companies as a de facto strategic national asset.

Index Weight: ~0.40% | Rank: Approximately #55–70 in the S&P 500 by market cap

Company Overview

Intel is executing the most consequential semiconductor turnaround in a generation. The company has officially entered high-volume manufacturing on its 18A node—a 1.8nm-class process featuring RibbonFET gate-all-around transistors and PowerVia backside power delivery—putting it ahead of TSMC's competing N2 node, which is not expected until late 2026. Panther Lake client processors and Clearwater Forest server chips are the first products shipping on 18A out of Fab 52 in Arizona. Anchor foundry customers include Microsoft (for Maia 3 AI accelerators) and AWS (for custom silicon), along with the U.S. Department of Defense. The '5 nodes in 4 years' roadmap—Intel 7, 4, 3, 20A, 18A—is complete. The strategic picture has transformed under CEO Lip-Bu Tan, who took the helm in March 2025 after a year-long CEO vacuum following Pat Gelsinger's departure. Tan has imposed engineering-first discipline: slashing management layers by half, cutting headcount to roughly 75,000–85,000, reducing OpEx to $16.5 billion, and killing the standalone Falcon Shores AI training accelerator in favor of a custom ASIC and inference-centric strategy. The $5 billion NVIDIA investment closed in Q4 2025, anchoring a co-development deal for NVLink-integrated custom Xeon CPUs and x86 SoCs with NVIDIA RTX GPU chiplets. Intel is no longer trying to compete head-on with NVIDIA in AI training; it is positioning itself as the manufacturing and CPU platform layer beneath the AI ecosystem. The tension at the center of Intel's story is simple: technical execution is succeeding while financial returns lag. The Foundry segment lost $10.3 billion in FY2025, 18A yields are around 60% (improving at 7–8% per month), and Q1 2026 gross margins are guided to 34.5%—well below historical norms. Intel's thesis is that yield normalization, external foundry customer ramps in H2 2026, and the 14A node (risk production 2027, first to use High-NA EUV) will drive a path to Foundry breakeven by 2027. That timeline has zero margin for error.

Products & Revenue

Intel reports three primary segments: Client Computing Group (CCG), Data Center and AI (DCAI), and Intel Foundry. CCG and DCAI collectively form 'Intel Products,' which generated $49.1 billion in FY2025 revenue—the core profit engine. Intel Foundry recorded $17.8 billion, mostly from internal wafer sales to Intel Products, and ran a $10.3 billion operating loss. The company divested majority control of Altera (FPGAs) to Silver Lake in 2025 and retains minority stakes in Mobileye and Altera under the 'All Other' category.

Client Computing Group (CCG) (~48%): PC processor platforms for notebooks, desktops, and tablets. Panther Lake (Core Ultra Series 3) on 18A launched at CES 2026. Nova Lake follows in late 2026.

Data Center and AI (DCAI) (~25%): Xeon server CPUs (Granite Rapids shipping, Diamond Rapids next), AI accelerators (Gaudi 3), and networking silicon. 15% revenue growth in Q4 2025.

Intel Foundry (~27%): Semiconductor manufacturing, technology development, and foundry services. Includes internal wafer sales and nascent external customer revenue. Currently running at a $10.3B annual operating loss.

All Other (Mobileye, IMS, Altera residual) (~1%): Post-Altera deconsolidation, primarily Mobileye ADAS silicon and Intel Manufacturing Services. Q4 2025 revenue was $574M.

Based on Intel FY2025 10-K (fiscal year ended December 27, 2025) and Q4 2025 earnings release (January 22, 2026). NEX was folded into CCG and DCAI in Q1 2025.

Leadership

Lip-Bu Tan

CEO since 2025. Appointed CEO in March 2025 after serving on Intel's board from 2022–2024. Previously Executive Chairman of Cadence Design Systems, where he led a 3,200% stock price appreciation and more than doubled revenue. Deep venture capital background through Walden International, with board-level experience across dozens of semiconductor companies.

Naga Chandrasekaran, EVP & CTO/COO, Intel Foundry: Oversees the unified foundry organization spanning technology development, manufacturing, and go-to-market. Joined from Micron's technology development unit. The single most critical execution leader for the 18A yield ramp and 14A node development.

Kevork Kechichian, EVP & GM, Data Center Group: Joined from Arm, where he led engineering and the company's transformation from IP licensing to full-stack solutions. Owns the Xeon roadmap including Diamond Rapids and the custom NVIDIA NVLink Xeon program.

Jim Johnson, SVP & GM, Client Computing Group: 40-year Intel veteran leading Panther Lake ramp and Nova Lake development. Manages the largest revenue segment and the AI PC product strategy across the global OEM ecosystem.

David Zinsner, EVP & CFO: Joined from Micron in 2022. Served as interim co-CEO during the leadership vacuum. Architecting the balance sheet recovery including $7.86B in CHIPS Act funding, NVIDIA's $5B equity investment, and the Altera divestiture.

Dr. Craig H. Barratt, Incoming Independent Board Chair (effective May 2026): Former CEO of Atheros Communications (acquired by Qualcomm for $3.1B). Held senior roles at Intel, Google, and Qualcomm. Stanford EE PhD. His appointment signals deep semiconductor governance at the board level.

The AI Angle

Manufacturing the silicon beneath every AI stack

Intel's AI strategy has undergone a dramatic narrowing under Lip-Bu Tan. The company abandoned its direct assault on NVIDIA's AI training accelerator market—canceling Falcon Shores, the hybrid GPU that was supposed to succeed Gaudi—and pivoted toward three interconnected plays: AI-optimized x86 CPUs, custom ASIC design services, and inference-focused solutions. Gaudi 3 remains in market (shipping via Dell, IBM Cloud, and others on standard Ethernet fabric), but Intel's leadership has acknowledged it failed to achieve broad adoption due to an immature software ecosystem and CUDA lock-in. The company set a $500M Gaudi sales target for 2024 and missed it. Future AI accelerator efforts center on Jaguar Shores, a rack-scale solution rather than a standalone chip, expected in 2026. The more consequential AI play is Intel's positioning as the manufacturing backbone for AI silicon. Microsoft's Maia 3 AI accelerator, Amazon's AWS custom chips, and potentially NVIDIA components will be fabricated on Intel 18A. The NVIDIA partnership is the centerpiece: Intel is co-developing custom Xeon CPUs with NVLink Fusion integration, allowing x86 processors to serve as host nodes in NVIDIA's rack-scale AI systems (GB300 NVL72 and Vera Rubin architectures). NVIDIA invested $5 billion in Intel equity to cement this relationship. On the consumer side, Intel x86 SoCs will integrate NVIDIA RTX GPU chiplets via NVLink for next-generation AI PCs. Intel also announced a multi-year collaboration with SambaNova to deliver AI inference solutions combining SambaNova's Reconfigurable Data Units with Xeon processors and Intel networking. On the product side, every new Intel processor now ships with dedicated AI hardware. Panther Lake features a next-generation NPU and Xe3 graphics for local AI inference, targeting the 'AI PC' category that Intel expects to reach a $25 billion TAM by end of 2026. Core Ultra Series 3 is positioned as the most broadly adopted AI PC platform Intel has delivered. In the data center, Granite Rapids Xeon claims up to 68% TCO savings and 80% less power versus average installed servers—a pitch squarely aimed at inference-heavy workloads where Intel's CPU advantage still holds. Intel has also established a dedicated custom ASIC organization reporting into central engineering, building on its networking and infrastructure ASIC footprint. The competitive risk is clear: NVIDIA dominates AI data center revenue with an estimated 86% share, AMD's Instinct MI series is gaining traction, and hyperscalers are investing heavily in their own custom ASICs (growing 44.6% in 2026 versus 16.1% for GPUs). Intel's bet is that the AI era requires heterogeneous silicon—CPUs, NPUs, GPUs, ASICs, and advanced packaging—and that no one else can offer leading-edge domestic manufacturing, advanced packaging (Foveros), and x86 ecosystem integration in a single stack. The risk is that external foundry customers treat Intel 18A as a hedge rather than a primary node, and that the Foundry never reaches the volume needed to break even.

Financial Snapshot

Revenue (TTM): $52.9B — FY2025 | Net Income: Non-GAAP EPS $0.42 (GAAP net income significantly negative due to foundry losses and restructuring)

Margins: Gross 36.7% (non-GAAP), operating varies sharply by segment (CCG ~29%, DCAI ~20%, Foundry deeply negative)

Intel generated $9.7 billion in operating cash flow in FY2025 and $2.2 billion in adjusted free cash flow in Q4 alone. The balance sheet has been stabilized through $5.7B in accelerated CHIPS Act funding, NVIDIA's $5B equity investment, SoftBank's $2B investment, and $4.3B from the Altera sale to Silver Lake. Intel plans to repay all $2.5B of 2026 debt maturities and expects positive adjusted free cash flow for the full year 2026. The core tension: Intel Products is profitable and generating cash, but Intel Foundry is burning through billions annually during the 18A ramp—a structural gap that management targets to close by 2027.

1-Year Performance

$43.13 as of March 27, 2026. Up 82.6% year-over-year from approximately $23.60.

The 52-week low of $17.67 in April 2025 marked peak despair during the post-Gelsinger leadership vacuum and fears of a failed turnaround. The subsequent rally to $54.60 was driven by Lip-Bu Tan's appointment, the NVIDIA $5B investment, 18A high-volume manufacturing confirmation, and successive earnings beats (five consecutive quarters above guidance). The recent pullback from $54.60 to $43.13 reflects Q1 2026 supply constraints (depleted buffer inventory, tight DRAM/substrate availability), Middle East geopolitical tensions hitting chip stocks broadly, and margin pressure from early 18A ramp costs. Intel and AMD reportedly announced CPU price increases of up to 15% on March 26, which briefly rallied the stock before the broader selloff.

Recent News

Fun Fact: Intel's PowerVia backside power delivery technology—the key innovation in 18A—was originally demonstrated on a test chip codenamed 'Blue Sky Creek' in 2023, fabricated on the otherwise-canceled Intel 20A node. The 20A process was scrapped as a commercial product, but its sole purpose became proving that relocating power rails to the back of the silicon wafer could reduce voltage droop by up to 30%. Intel effectively burned an entire process node generation as a technology testbed—a $2B+ science experiment that made 18A possible. TSMC's competing backside power approach (Super Power Rail) won't ship at scale until its A16 node in late 2026, giving Intel a roughly 12-month architectural lead that represents its first genuine manufacturing advantage over TSMC since the 14nm era in 2014.