stockrowWhat it does
Advanced Micro Devices, Inc. (AMD) is headquartered in Santa Clara and is led by Chair, President and Chief Executive Officer Lisa Su. It develops computing chips and supporting software for data centers, PCs, gaming devices and embedded equipment.
For data centers, AMD sells server CPUs, AI accelerators and GPUs, networking products, DPUs, AI network interface cards, FPGAs and adaptive SoCs. These components are used in cloud infrastructure, enterprise systems, supercomputers and AI training and inference workloads. Its Helios platform combines server CPUs, MI450-series GPUs, Pensando networking and Rackham software into a rack-scale AI system.
For PCs, AMD sells Ryzen and Ryzen AI processors, APUs and chipsets to computer makers and distributors. Its graphics products serve PC gaming, creation and professional workstation uses. It also supplies semi-custom SoCs for Sony PlayStation 5 and Microsoft Xbox Series S and X consoles.
Its Embedded business sells CPUs, APUs, FPGAs, system-on-modules and adaptive SoCs for aerospace and defense, automotive, industrial, healthcare, communications, test equipment and other systems. These products can be tailored to a customer’s system requirements and often support applications requiring low power, small form factors, long operating life or real-time processing.
AMD also develops software tools, compilers, drivers and AI software stacks for its hardware platforms, including ROCm, Vitis and Vivado.
How it makes money
AMD earns revenue by selling processors, accelerators, graphics products, networking components, programmable chips, software-related tools and semi-custom designs. It reports three operating segments: Data Center, Client and Gaming, and Embedded. It can also sell or license portions of its intellectual-property portfolio.
Data Center was AMD’s largest segment in FY2025, representing 48.0% of revenue. The segment sells products used in servers and AI infrastructure, including EPYC server processors, Instinct accelerators, GPUs, networking products, FPGAs and adaptive SoCs. Revenue depends on product shipments to hyperscalers, OEMs, ODMs, system integrators and distributors.
Client and Gaming represented 42.0% of FY2025 revenue. The client business sells PC processors and chipsets principally through PC OEMs, distributors and selected direct customers. Gaming revenue includes discrete Radeon graphics products and semi-custom SoCs and development services. Certain customers pay non-recurring engineering fees as well as for resulting products.
Embedded represented 9.97% of FY2025 revenue. Its products are sold directly and through distributors and OEM partners into multiple specialized end markets. The company also provides development kits, hardware and software tools that support customers building systems around its programmable products.
Sales are generally based on customer forecasts, purchase orders and contractual arrangements rather than minimum-purchase commitments. AMD generally provides one-year product warranties, with up to three-year limited warranties for certain products. Distributor arrangements can include price protection, stock rotation or return rights, affecting net revenue.
FY2025 net revenue was $34.6 billion; Q2 2026 revenue was $11.5 billion.
| Segment | Revenue, FY2025 | Share | Change on the year |
|---|---|---|---|
| Datacenter | $16.6 billion | 48.0% | 32.2% |
| Client and Gaming | $14.6 billion | 42.0% | 50.8% |
| Embedded | $3.5 billion | 9.97% | −2.90% |
Customers and how it sells
AMD sells through a direct sales force, OEMs, independent distributors and sales representatives. Data Center customers include hyperscale data centers, OEMs, ODMs, system integrators and distributors. Large cloud providers and private hyperscale operators buy AMD products directly or indirectly through systems built by OEMs, ODMs or contract manufacturers.
Client and Gaming customers include PC OEMs, distributors, motherboard ODMs, add-in-board manufacturers, game-console makers, portable-gaming-device makers and contract manufacturers. Embedded products are sold directly and through distributors and OEM partners to customers in markets ranging from aerospace and defense to industrial, healthcare and communications infrastructure.
AMD works with customers on product features, performance and development timing, and uses application engineers to support customer system design, testing and qualification. Its product-purchase agreement with OpenAI provides for deployment of 6 gigawatts of AMD GPUs, beginning with capacity powered by MI450-series products.
The United States generated $11.4 billion, or 32.8%, of FY2025 revenue. China generated $7.8 billion, or 22.4%; Taiwan generated $5.2 billion, or 15.0%; Singapore generated $4.3 billion, or 12.4%; and Other Countries generated $6.1 billion, or 17.5%. International sales were 67% of FY2025 net revenue.
| Region | Revenue, FY2025 | Share | Change on the year |
|---|---|---|---|
| UNITED STATES | $11.4 billion | 32.8% | 30.7% |
| CHINA | $7.8 billion | 22.4% | 24.4% |
| Other Countries | $6.1 billion | 17.5% | 53.4% |
| TAIWAN | $5.2 billion | 15.0% | 57.1% |
| SINGAPORE | $4.3 billion | 12.4% | 18.5% |
Suppliers and costs
AMD is fabless: it designs products but relies on third parties for wafer fabrication, board manufacturing, assembly, testing, marking and packaging. It uses Taiwan Semiconductor Manufacturing Company Limited (TSMC) for wafers for high-performance computing, FPGA and adaptive-SoC products. It uses GLOBALFOUNDRIES for certain 12 nm and 14 nm HPC wafer purchases, and TSMC, United Microelectronics Corporation and Samsung Electronics for programmable-logic devices.
Assembly and test partners are located in the Asia-Pacific region. AMD’s ATMP joint ventures with Tongfu Microelectronics, along with Siliconware Precision Industries and King Yuan Electronics, provide assembly, test, mark and packaging services. It outsources board-level graphics manufacturing.
Costs include wafers, substrates, assembly and test, employee expenses, research and development, go-to-market activities, acquisition-related intangible amortization and stock-based compensation. FY2025 R&D expense was $8.1 billion, while marketing, general and administrative expense was $4.1 billion. AMD had approximately $12.2 billion of unconditional commitments at December 27, 2025, primarily for wafers, substrates, cloud-service arrangements, software and technology licenses.
- Revenue$34.6B
- Cost of revenue−$17.5B
- Research and development−$8.1B
- Selling, general and administrative−$4.1B
- Other operating costs−$1.2B
- Operating income$3.7B
Competitive position
AMD operates in semiconductor markets that it describes as highly competitive and rapidly evolving. In data-center CPUs, GPUs, DPUs and AI network-interface cards, it competes primarily with Intel Corporation and Nvidia Corporation. It competes with Altera in data-center FPGAs and adaptive SoCs, while smaller fabless companies and Arm-based CPU providers also target data-center workloads.
In client CPUs and APUs, Intel is AMD’s primary competitor, while Arm-based PC platforms may gain adoption. In discrete graphics, Nvidia is the market-share leader; Intel also sells integrated graphics processors and gaming-focused discrete GPUs. In embedded products, named competitors include Altera, Lattice Semiconductor Corporation, Microsemi Corporation, Broadcom Corporation, Marvell Technology Group, Analog Devices, Texas Instruments, NXP Semiconductors, Qualcomm and NVIDIA.
AMD’s competitive position depends on product performance, total cost of ownership, product timing, quality, energy efficiency, pricing, software and hardware compatibility, software tools, security, brand recognition and supply availability. It develops hardware and software together, including ROCm, and holds approximately 18,900 patent matters worldwide.
Our reading: AMD’s broadest moat source is its ability to combine CPUs, GPUs, networking, programmable silicon and software into one data-center system; Helios is evidence of that integration, but it is still being ramped. Our reading: its software ecosystem and customer co-design work can support adoption, while Intel and Nvidia remain direct competitors across the largest CPU and GPU markets.
| Competitor | Where they compete | Threat (our reading) |
|---|---|---|
| Intel Corporation | Data-center and client CPUs; graphics; embedded CPUs | High |
| Nvidia Corporation | Data-center GPUs and AI; discrete graphics | High |
| Altera | Data-center and embedded FPGAs and adaptive SoCs | Medium |
| Broadcom Corporation | Embedded ASSP and programmable-logic markets | Medium |
| Lattice Semiconductor Corporation | Embedded FPGA and programmable-logic markets | Medium |
| Microsemi Corporation | Embedded FPGA and programmable-logic markets | Low |
| Marvell Technology Group, Ltd. | Embedded ASSP markets | Medium |
| Qualcomm Incorporated | Embedded ASSP markets | Medium |
Growth opportunities
AMD’s principal growth vector is data-center computing, including server CPUs, AI accelerators, networking and rack-scale systems. Data Center revenue was $16.6 billion in FY2025, up 32.2% from FY2024. In Q2 2026, Data Center revenue was $6.7 billion, more than double year over year, and represented approximately 58% of total revenue.
Management said on the call that Helios production shipments were set to begin later in Q2 2026 and ramp through the fourth quarter and into 2027. Management also said that OpenAI and Meta have multi-generation gigawatt-scale deployments, Anthropic will deploy up to two gigawatts of MI450-series GPUs in Helios, and Microsoft will deploy Helios on Azure for frontier-model inference.
The company sizes the high-performance AI-computing market at $2.00 trillion by 2030. It expects the data-center AI-accelerator market to reach $1.40 trillion by 2030 and the server CPU market to reach $220.0 billion by 2030. Management said on the call that it expects Data Center segment revenue to more than double year over year in 2027, driven by the data-center AI ramp and server CPU growth.
Client growth is tied to Ryzen processors, commercial PC adoption and AI PCs, although management said on the call that it is planning for a softer PC market in the second half of 2026 because of higher memory and component costs. Embedded growth is supported by design wins across networking, data center, communications, test, aerospace and defense; management said on the call it was tracking more than $18.0 billion of new design wins in FY2026.
- $16.4BFY2021
- $23.6BFY2022
- $22.7BFY2023
- $25.8BFY2024
- $34.6BFY2025
Structural risks
-
Manufacturing dependence: AMD relies on third parties to manufacture, assemble and test products. A failure to supply products on time, in sufficient quantities or with competitive technologies could materially hurt the business.
-
Intense competition and rapid change: Its markets are highly competitive and rapidly evolving. Product performance, timing, cost, power use, software, supply availability and customer preferences can change competitive outcomes.
-
Export controls and trade actions: Export regulations, tariffs and trade protection measures may limit sales to some customers. New U.S. restrictions on certain semiconductors to China led to approximately $800 million of MI308 inventory and related charges during Q2 2025, partly reversed after licensed shipments began.
-
Supply-chain capacity: The company needs wafer, back-end, substrate, memory and system-component capacity. Management said on the call that server CPU supply was tight in 2026, though it expected the 2027 situation to improve.
-
Complex AI-system execution: Helios combines CPUs, GPUs, networking and software in a rack-scale system. Management said on the call that yields are expected to improve during the initial ramp quarters because the product is highly complex.
-
Global operating exposure: Worldwide operations are subject to political, legal, economic and natural-disaster risks. Acquisitions, investments and integration efforts may not deliver anticipated benefits and could disrupt operations.
Bottom line
Our reading
Our reading: AMD is a fabless computing-platform business whose revenue spans data-center chips and systems, PC processors and graphics, gaming silicon, and embedded products. Its current growth is increasingly tied to server CPUs, AI accelerators and the Helios rack-scale platform, while production depends on external fabrication, packaging and component supply. Competition from Intel, Nvidia and other semiconductor suppliers, export restrictions and the execution of complex system ramps are central constraints on that model.