AMD
The only company globally that competes head-on with Intel and NVIDIA in both CPU and GPU markets. Provides data-center-grade MI series AI accelerator cards and EPYC server CPUs.
Possesses extremely strong chip design capabilities (CPU+GPU integration), making it the only effective choice for cloud providers to counterbalance NVIDIA. The ROCm software ecosystem is rapidly catching up with CUDA, offering significant cost-performance advantages.
Company Overview
AMD (Advanced Micro Devices) was founded in 1969 by Jerry Sanders, with its headquarters in Santa Clara, California, USA. The company initially started as a producer of x86 CPUs, long operating in the shadow of Intel.
In 2014, Lisa Su, a female engineer of Chinese descent, became CEO and led AMD through a revolutionary transformation. The Zen architecture CPU launched in 2017 enabled AMD to re-emerge in the server market; the MI300X AI accelerator card released in 2023 made AMD the only competitor posing a substantial threat to NVIDIA in the AI computing chip market. As a result, Lisa Su has been hailed as the "Chip Queen" in the industry.
Core Business
Data Center AI Accelerators | Core Growth Driver (Revenue Share ~40%, Fastest-Growing)
- MI300X: Flagship AI accelerator released in late 2023, featuring a Chiplet design that integrates 12 chiplets with a total of 153 billion transistors. In LLM inference scenarios, the MI300X's performance and cost-effectiveness approach or even partially surpass Nvidia's H100.
- MI325X / MI350 Series: Successive iterations aimed at further narrowing the gap with Nvidia's Blackwell in training performance.
- ROCm Software Ecosystem: AMD's open-source software platform competing with Nvidia's CUDA. Although its maturity and developer base still lag far behind CUDA, major customers such as Meta and Microsoft are actively driving the maturation of the ROCm ecosystem.
Server CPUs (EPYC Series) | Stable Revenue Source
AMD's EPYC server processors comprehensively surpass Intel's Xeon processors in core count, power efficiency, and price-performance ratio. In AI servers, EPYC CPUs handle data preprocessing, model distribution, and system scheduling, making them an indispensable companion to GPU clusters.
PC CPUs and GPUs
Ryzen series CPUs and Radeon series GPUs target the consumer market, providing AMD with stable cash flow and brand influence.
Technology Moat
Chiplet Design Capability: AMD is an industry pioneer in chiplet architecture. By decomposing a large chip into multiple smaller chiplets (CCD + IOD), AMD can significantly reduce manufacturing costs and improve yield. The MI300X uses a complex combination of 12 chiplets, demonstrating AMD's deep expertise in advanced packaging and system integration.
CPU+GPU Dual-Line Integration: AMD is the only company in the world that simultaneously offers high-performance x86 CPUs and AI GPU products. Under heterogeneous computing architectures (such as AMD's Infinity Architecture), the efficiency of data transfer between CPU and GPU is the key bottleneck for system performance.
High Cost-Performance Competitive Strategy: Against the backdrop of NVIDIA GPU supply shortages and surging prices, AMD has leveraged an aggressive pricing strategy (typically 30%-50% lower than comparable NVIDIA products) to successfully win over major customers such as Microsoft and Meta.
Market Landscape
| Dimension | Data |
|---|---|
| AI accelerator card market share | ~10% (growing rapidly) |
| Server CPU share | ~30% (continuously taking share from Intel) |
| Main competitors | NVIDIA (GPU), Intel (CPU) |
| Key customers | Microsoft, Meta, Google, Oracle |
Financials & Growth
| Metric | Data |
|---|---|
| Revenue (2024) | ~USD 25 billion |
| Data Center Revenue | Over 70% of total (exceeding PC business for the first time) |
| Gross Margin | ~50% |
| Growth Drivers | MI series AI chip volume ramp-up + EPYC CPU market share gains |
Risks and Summary
Key Risks:
- The ROCm ecosystem lags far behind CUDA in maturity, requiring major customers to still invest substantial engineering resources for adaptation.
- NVIDIA locks in customers through system-level solutions (GPU + NVLink + networking), while AMD still lags in system integration.
- Supply chain constraints: advanced packaging (CoWoS) capacity is also limited by TSMC quota allocations.
Core Industry Value: AMD is the "only balancer" in the AI compute chip market. Without AMD, NVIDIA would completely monopolize and control the pricing of AI compute capacity. AMD's presence gives cloud providers critical bargaining leverage and a second leg to stand on.