Key Takeaways
Intel's fiscal year 2025 (calendar year ended December 31, 2025) was the most consequential year in the company's post-Gelsinger strategic reset — a year that tested whether the Intel Foundry Services (IFS) model could attract external customers at scale while the core Products business (Client and Data Center CPUs) continued losing market share to AMD and facing structural disruption from Arm-based chips in both PC and server markets. Total revenue reached approximately $52-54B, roughly flat to slightly down from FY2024's $53.1B, as modest Client CPU volume recovery was offset by ongoing Data Center share erosion and the continued ramp-up costs of the 18A and 20A process nodes in Foundry. GAAP net income remained deeply negative (approximately -$12B to -$15B) as restructuring charges, Foundry capital investment depreciation, and goodwill impairments dominated the P&L; non-GAAP EPS of approximately $1.00-1.30 significantly understated the cash drain. The FY2026 thesis is existential in nature: can Intel's 18A process node — its first internally-developed leading-edge node since the 10nm debacle of 2017-2021 — achieve competitive yields and attract anchor foundry customers (Microsoft, Nvidia, Qualcomm) who committed to 18A designs before seeing production silicon, or does the foundry pivot fail and Intel revert to a fabless-like model that destroys the logic behind decades of capital investment?
Intel was founded in 1968 by Gordon Moore and Robert Noyce in Mountain View, California, and built the semiconductor industry's most powerful vertical integration model: Intel designed its own chips, manufactured them on its own leading-edge process nodes, and sold them to PC and server OEMs under the Intel Inside brand. This model generated extraordinary returns for three decades — from the 8086 microprocessor through the Pentium era through Core i-series — and at its peak in 2021, Intel was the world's most valuable semiconductor company. The structural crisis began in 2012-2018 when Intel failed to shrink its process node from 14nm to 10nm on schedule, allowing TSMC to close and ultimately surpass Intel's manufacturing advantage. TSMC's superior 7nm, 5nm, and 3nm nodes enabled AMD's Zen CPUs (manufactured at TSMC) to close the performance gap with Intel and surpass it by FY2022-FY2023 in server CPU performance per dollar.
CEO Pat Gelsinger, who returned to Intel from VMware in 2021, launched the most ambitious strategic pivot in Intel's history: rebuilding Intel's manufacturing to process leadership with five new nodes in four years (Intel 7, Intel 4, Intel 3, 20A, 18A) while simultaneously opening Intel Foundry Services to external customers. The plan required $100B+ in capital investment over five years, an organization restructuring that eliminated 15,000+ jobs in 2024, and the patience of investors through multiple years of GAAP losses. Gelsinger's departure in December 2024 and the appointment of an interim CEO raised questions about whether the foundry strategy would be maintained or scaled back — a question that defined the FY2025 investment narrative.
Business Structure
Intel reports two primary divisions with multiple segments.
Intel Products (~$44B revenue, ~82% of total in FY2025): The traditional CPU and network chip business.
-
Client Computing Group (CCG) (~$27B, ~50%): PC CPUs for consumer and commercial markets, including the Core Ultra series with integrated NPUs for AI PC workloads. CCG benefits from the ongoing AI PC refresh cycle (Intel's NPU-equipped processors are required for Microsoft Copilot+ PC designation) but faces structural competition from Qualcomm's Snapdragon X Elite Arm chips and Apple's M-series in premium segments.
-
Data Center and AI Group (DCAI) (~$13B, ~24%): Server CPUs (Xeon Scalable / Granite Rapids) and AI accelerators (Gaudi 2/3). DCAI has lost significant market share to AMD's EPYC (from approximately 95% in 2019 to approximately 73-75% in 2025) and faces longer-term disruption from hyperscaler custom silicon (AWS Graviton, Google Axion, Microsoft Cobalt) that bypasses x86 entirely for cloud-native workloads. Gaudi 3 AI accelerator has not gained meaningful traction against NVIDIA H100/H200.
-
Network and Edge (NEX) (~$4B, ~7%): Ethernet switching, network adapters, and edge compute chips.
Intel Foundry (~$10B revenue, ~18%): External foundry revenue from customers using Intel's manufacturing capacity, plus internal "transfer revenue" from Products division. The external foundry revenue target is $15B+ by FY2027, but FY2025 external revenue remained limited (approximately $1-2B) as 18A qualification continued and customers awaited yield data on production silicon.
Key Core Metrics Performance
Revenue by Segment (FY2021–FY2025)
Intel's revenue decline from FY2021-FY2023 reflects share loss in both PC CPUs (Arm competition, Apple silicon in Mac) and server CPUs (EPYC gains), partially offset by data center demand recovery in FY2024-FY2025.
| Fiscal Year | Total Revenue | CCG (Client) | DCAI (Data Center) | Intel Foundry |
|---|---|---|---|---|
| FY2021 | $79.0B | ~$40.0B | ~$21.1B | N/A (pre-IFS) |
| FY2022 | $63.1B | ~$31.7B | ~$19.2B | ~$1.0B |
| FY2023 | $54.2B | ~$25.2B | ~$15.6B | ~$3.0B |
| FY2024 | $53.1B | ~$29.3B | ~$12.8B | ~$4.3B (reported) |
| FY2025 | ~$53.5B | ~$27.0B | ~$13.0B | ~$6.0B |
The revenue stabilization at approximately $53-54B conceals the underlying mix shift: legacy high-margin CPU products (Xeon server, premium Core desktop) declining as AMD and Arm take share, offset by volume recovery in mid-range PC CPUs.
Profitability and the Foundry Investment Burden (FY2021–FY2025)
Intel's profitability collapse reflects both share loss and the capital investment burden of the foundry build-out.
| Fiscal Year | Revenue | Non-GAAP Gross Margin | Non-GAAP Op. Margin | Non-GAAP EPS | GAAP Net Income |
|---|---|---|---|---|---|
| FY2021 | $79.0B | 57.4% | 33.5% | $5.47 | $19.9B |
| FY2022 | $63.1B | 49.4% | 25.0% | $4.29 | $8.0B |
| FY2023 | $54.2B | 42.5% | 8.2% | $1.05 | $1.7B |
| FY2024 | $53.1B | 41.0% | 3.0% | $0.89 | -$16.4B |
| FY2025 | ~$53.5B | ~40.5% | ~3.5% | ~$1.10 | ~-$10B |
The GAAP losses of FY2024-FY2025 reflect massive impairment charges on Foundry assets (goodwill and PP&E written down as external customer traction lagged), restructuring charges from the 15,000+ headcount reduction, and accelerated depreciation on fab equipment. The non-GAAP EPS of approximately $1.10 excludes these charges but is still far below the $5+ peak levels of FY2021-FY2022.
Intel Foundry 18A Node — the Pivotal Bet (FY2024–FY2025)
| Milestone | Timeline | Status |
|---|---|---|
| 18A process development complete | Q1 2025 | ✓ Confirmed |
| Microsoft (first external customer) volume production | H2 2025 | Qualification ongoing |
| Qualcomm tape-out (committed) | 2025-2026 | Pre-production |
| Nvidia internal discussions | Exploratory | Uncommitted |
| IFS revenue run-rate $15B | FY2027 target | Requires external ramp |
Market Evaluation
Intel trades at approximately 20-30x forward non-GAAP EPS — a multiple that is difficult to defend on near-term earnings power but reflects the optionality value of 18A succeeding. The bull case is that 18A achieves production-quality yields in H2 2025, Microsoft begins volume production in early 2026, and the demonstrated performance data triggers Qualcomm and Amazon to commit foundry capacity, putting Intel on a path to $15B+ foundry revenue by FY2028. Combined with a PC CPU cycle recovery and partial Data Center stabilization, this could support $3-4 non-GAAP EPS at normalized margins. The bear case is that 18A struggles with yield ramp (as Intel's 10nm did from 2015-2019), external customers defer or cancel commitments, and the foundry model is abandoned — forcing a wholesale strategy pivot under a new CEO that destroys years of capital investment and leaves Intel as a sub-scale fabless chip designer competing with AMD, Qualcomm, and Apple Silicon without manufacturing differentiation.
18A Process Node and the Foundry Customer Imperative
Intel's 18A process represents the company's attempt to reclaim process leadership after nearly a decade of manufacturing delays. The 18A node uses two breakthrough technologies: RibbonFET (Intel's gate-all-around transistor architecture, equivalent to TSMC's GAA in N2) and PowerVia (backside power delivery, which routes power through the back of the wafer to free up front-side routing for signal wires). Both technologies are industry firsts at production scale, and early benchmark data suggests 18A's performance and power efficiency are competitive with TSMC N2 — Intel's first credible claim to process leadership since 2018.
The foundry customer acquisition challenge is psychological as much as technical: hyperscalers and fabless chip designers have built 5-10 year supply chain relationships with TSMC and Samsung that are not easily disrupted. Intel must offer a combination of price competitiveness (matching or beating TSMC N2 pricing), capacity commitment (guaranteeing allocation for customer designs), and performance equivalence (demonstrated yield on real silicon) to win volume commitments. Microsoft's commitment to produce a chip on 18A — announced in 2024 — provides a credibility anchor, but a single customer does not make a foundry business at scale. The IFS $15B revenue target by FY2027 implies winning 5-7 additional significant customers beyond Microsoft, which requires delivering production silicon from 18A by H1 2026 at the latest.