Skip to content

🐂 AMD — Company Profile (Teaching)

Teaching site, not investment advice. Specific market cap / stock price / financial figures change frequently and are intentionally omitted; use the SEC + Yahoo Finance links below to pull real-time data yourself.

📌 Basics

Dimension Content Source
Core business EPYC datacenter CPU + MI300/MI350/MI400 AI GPU + Ryzen desktop CPU + Radeon gaming GPU + Embedded (Xilinx FPGA) 10-K FY25
Founded / HQ 1969 / Santa Clara, California, USA wiki
CEO Lisa Su (since 2014/10; widely credited with turnaround) DEF 14A
Primary listing NASDAQ: AMD —
Fiscal year Ends Dec 31 (calendar year) SEC EDGAR

💡 Real-time market cap / share price / MI series revenue / DC GM: see SEC primary + IR links below.

🏭 Supply Chain Coordinates (Part 1-C6 framework)

Upstream (Who AMD Depends On)

Supplier What flows Key dependency
TSM Foundry for EPYC / MI / Ryzen / FPGA Irreplaceable — AMD is a top-3 TSM customer
SKH / MU / SSNLF MI300X uses HBM3, MI350 series uses HBM3E Multi-source (competing with NVDA for HBM3E)
SNPS + CDNS EDA tools Duopoly
ASE / SPIL Back-end packaging (advanced packaging like CoWoS still at TSM) Multi-source
COHR + LITE Optical modules (AI cluster fabric) Dual-source

Downstream (Who Depends on AMD)

Customer Relationship Concentration
MSFT Azure EPYC + MI300/MI350 GPU secondary + Cobalt co-design Hyperscaler mega-customer
META MI300/MI350 internal deployments (Llama inference) Hyperscaler mega-customer
OPENAI AMD-OpenAI 6GW deal (LEDGER F-025) — strategic Strategic
AMZN EPYC AWS mega-customer + MI300 secondary Hyperscaler
GOOGL EPYC GCP secondary (TPU primarily self-supplied) Hyperscaler
OEMs (DELL / HPE / SMCI) EPYC + Ryzen sold to enterprise + consumer via OEM Highly diversified

Competitors

  • AI GPU: NVDA (90%+ dominant share; AMD MI series catching up at 5-10%)
  • Datacenter CPU: INTC Xeon (historical leader, losing share to AMD EPYC); ARM server CPUs (AMZN Graviton / MSFT Cobalt / GOOGL Axion)
  • Desktop CPU: INTC Core (AMD Ryzen grabbing gaming + high-end desktop share)
  • Gaming GPU: NVDA GeForce / INTC Arc (negligible share)
  • FPGA / Embedded: LSCC Lattice (low-end); AMD is FPGA leader post-Xilinx acquisition

💼 Business Model + Value Capture (Part 1-C7 view)

  • Revenue mix: Data Center (EPYC + MI GPU) ~50%+ / Client (Ryzen + Radeon consumer) ~25% / Gaming (console custom silicon) ~15% / Embedded (Xilinx) ~10%
  • Gross margin: ~50% overall; MI GPU GM lower (vs. NVDA 70%+, reflecting weaker ROCm ecosystem and lower pricing power)
  • Pricing power: Medium — MI300 must be ~25% cheaper than H100 to dislodge customers; EPYC pursues a price/performance strategy
  • Substitution difficulty: CPU high (EPYC vs. Xeon locked in by hyperscalers) / AI GPU medium (ROCm still weak; customers use AMD mainly as NVDA secondary to avoid single-vendor risk)

(For specific quarterly revenue / MI revenue / DC GM, see the latest SEC 10-Q.)

🏰 Moat Assessment (Part 3-C2 Buffett 5-step lens)

Type Strength (0-1) Source
Brand 0.6 Engineers respect EPYC, but for AI GPUs the default is NVDA
Switching cost 0.5 EPYC and Xeon are both x86; switching is relatively easy (1-2 quarters qualification); ROCm-to-CUDA switching is extremely hard
Network effects 0.3 Far fewer ROCm developers than CUDA; some via FPGA design tools (Vivado)
Scale + cost 0.5 Same TSM node, but R&D spend is multiples below NVDA

Composite: 1.9 / 4 → CPU business has moat; AI GPU business moat is weak (against NVDA's dominance)

⚠️ Key Reverse-Trigger Conditions (Part 3-C5 anti-thesis view)

  • Trigger 1 (quarterly): MI300/MI350 shipments + revenue growth < 50% YoY — customers pull back on NVDA-secondary diversification
  • Trigger 2 (quarterly): EPYC datacenter share stops growing / reverses — INTC counterattack or ARM server CPU substitution
  • Trigger 3 (annual): ROCm software ecosystem fails to close CUDA gap — long-term AI GPU ceiling
  • Trigger 4 (event): China export controls expand — parts of AMD MI series restricted (like NVDA H20)
  • Trigger 5 (low probability): Hyperscaler in-house ASIC (Trainium / Maia / TPU) fully squeezes AMD's AI GPU share

🔗 Teaching Cross-References

📚 SEC Primary Filings (look it up yourself)

Real-time quote: Yahoo Finance AMD

Teaching use · numbers change — learn the method, not the specific figures · investment decisions are based on your own research + risk tolerance