Advanced Micro Devices Inc.
Advanced Micro Devices, Inc. (AMD)
Overview
Advanced Micro Devices, Inc. (AMD) is a global semiconductor company headquartered in Santa Clara, California, that designs high-performance computing and graphics processors. It operates in the technology hardware/semiconductors sector, competing at the center of the AI, PC, gaming, and data-center infrastructure buildouts. AMD is fabless — it designs chips but outsources manufacturing (primarily to TSMC) — and reported full-year 2025 revenue of $34.6 billion (up 34% year-over-year) and net income of $4.3 billion, with roughly 28,000–32,000 employees worldwide as of 2025–2026. The company has grown rapidly on the back of AI accelerator demand, server CPU share gains from Intel, and strong gaming and client PC sales.
What They Do & How They Make Money
AMD makes its money by designing and selling semiconductor chips — CPUs, GPUs, and specialized accelerators — that it does not manufacture itself. It licenses or develops chip architectures (most notably its "Zen" CPU core and "RDNA"/"CDNA" GPU architectures), has them fabricated by third-party foundries such as TSMC, and then sells the finished processors to computer makers (Dell, HP, Lenovo), cloud and hyperscale data-center operators (Microsoft, Meta, Oracle, and others), game console makers (Sony PlayStation, Microsoft Xbox use AMD custom silicon), and directly to consumers and system builders through retail channels. Revenue is transactional (per-chip sales) rather than subscription-based, so it is heavily influenced by product cycles, unit volumes, and average selling prices. AMD's fastest-growing revenue stream is data-center AI accelerators (the Instinct MI-series GPUs), which it sells to cloud providers and AI labs building out large training and inference clusters — sometimes bundled with software (its ROCm stack) and multi-year supply agreements, as seen in its widely reported large-scale deals with OpenAI and Meta.
Business Segments
AMD reports four operating segments:
- Data Center ($16.6 billion in FY2025 revenue, up 32% YoY; $3.6 billion operating income): Sells EPYC server CPUs and Instinct AI accelerator GPUs to cloud providers, hyperscalers, and enterprises. This is AMD's largest segment and the primary growth driver, fueled by continued EPYC share gains against Intel and the ramp of Instinct GPU shipments for AI training/inference.
- Client and Gaming (combined $14.6 billion, up 51% YoY): Reported as two sub-lines:
- Client ($10.6 billion, up 51% YoY): Ryzen CPUs and related chipsets for desktop and laptop PCs.
- Gaming ($3.9 billion, up 51% YoY): Radeon discrete GPUs for PC gaming, plus semi-custom SoCs supplied to Sony and Microsoft for the PlayStation and Xbox consoles.
- Combined segment operating income was $2.9 billion.
- Embedded ($3.5 billion, down 3% YoY; $1.2 billion operating income): Embedded processors and FPGAs (from the 2022 Xilinx acquisition) sold into industrial, automotive, aerospace/defense, communications, and networking equipment. The FY2025 decline reflected customer inventory adjustments earlier in the year.
Competitors
- CPUs (data center and client): Intel is AMD's longtime primary rival in x86 server and PC processors; Arm-based competitors (including custom silicon from Amazon, Google, and Microsoft, plus Qualcomm and Ampere) increasingly compete in servers and laptops.
- GPUs / AI accelerators: Nvidia is the dominant competitor and market leader in AI training/inference GPUs (roughly 80% market share), with AMD a distant but growing number two (~5–7% share). Custom AI silicon from hyperscalers — Google TPUs, Amazon Trainium/Inferentia, and Broadcom-designed ASICs for Google and Meta — represents a fast-growing alternative to merchant GPUs from either AMD or Nvidia.
- Embedded/FPGA: Intel (Altera), Lattice Semiconductor, and Microchip Technology compete in embedded and programmable logic markets.
- Gaming/console silicon: Nvidia (discrete GPUs) and, to a lesser extent, Apple (integrated graphics in its own silicon) compete for gaming and creator GPU sales.
Competitive Position
AMD's core competitive advantage is architectural: its Zen CPU core family let it leapfrog Intel on performance-per-watt and manufacturing node (via TSMC's leading-edge process access) starting in the late 2010s, which drove sustained server (EPYC) and desktop (Ryzen) market-share gains against a stumbling Intel. In gaming, AMD's semi-custom silicon business gives it an entrenched, multi-year revenue base as the chip supplier for both major game consoles. AMD is a fabless designer, which gives it flexibility to use the best available foundry process without carrying the capital burden of owning fabs — the same model Nvidia uses, and a structural disadvantage relative to Intel's integrated device manufacturer model turning into an advantage as TSMC has pulled ahead of Intel's own fabs in recent years.
The company's biggest opportunity and biggest challenge is the same market: AI data-center accelerators. AMD's Instinct MI-series GPUs offer a real technical edge in memory capacity (288GB HBM3E on the MI350X versus competing Nvidia parts) and are priced 30–50% below equivalent Nvidia GPUs, which has won AMD large, headline supply commitments from OpenAI and Meta and made dual-sourcing (buying from both AMD and Nvidia) increasingly standard practice among hyperscalers. However, Nvidia's CUDA software ecosystem remains far more mature than AMD's ROCm stack, and independent benchmarks suggest AMD's silicon still runs meaningfully below its theoretical peak performance in practice, with weaker multi-node training scaling. AMD has not yet demonstrated full-scale training runs of the largest frontier models on its hardware. The net effect is that AMD has become the credible number-two AI chip supplier — a dramatic improvement from a near-zero position a few years ago — but Nvidia's absolute revenue lead is still widening in dollar terms even as AMD's percentage growth is strong, and custom ASICs from hyperscalers are growing faster than either merchant GPU vendor, posing a longer-term threat to AMD's data-center ambitions.