Data Center ($5.8B, +57%): EPYC server CPUs >50% YoY. Instinct GPUs (MI300X/MI325X). Meta deal (up to 6GW, custom MI450). OpenAI deal (6GW, MI450 H2 2026). DC margin 28%.
MI450 execution risk (High): Not yet shipped. Technical delays or yield issues would be catastrophic.
Customer concentration (Medium-High): Meta, OpenAI, Microsoft = bulk of AI GPU demand.
ROCm weakness (Medium): AMD software lags CUDA significantly.
TSMC dependency (Medium): Same Taiwan risk as NVIDIA.
Warning Signs Check
Warning Sign
Status
Assessment
Balance Sheet
✓ Strong
$12.4B cash
Insider Selling
✓ Normal
—
P/E (Trailing)
⚠️ 140x
But fwd 50-60x
CUDA Moat
⚠️ Dominant
Biggest risk
MI450 Execution
⚠️ Not yet shipped
H2 2026 critical
7. Alternatives
Company
Ticker
Market Cap
P/E (Fwd)
Growth
AMD
AMD
~$750B
50-60x
+38%
NVIDIA
NVDA
$5.2T
27x
+85%
Intel
INTC
$100B
25x
-5%
Broadcom
AVGO
$900B
28x
+25%
ETF
P/E
Expense
AMD Allocation
SOXX (Semiconductor)
~28x
0.35%
AMD, NVDA, AVGO, INTC
QQQ (NASDAQ 100)
~30x
0.20%
Yes
8. Investment Thesis
Bull Case
Hyperscaler diversification away from NVIDIA. Meta ($60B), OpenAI, Microsoft buying AMD chips.
CPU + GPU combo. Only vendor with strong CPU (EPYC) + AI GPU (Instinct). Agentic AI needs both.
MI450/Helios (H2 2026). First serious system competitor to NVIDIA's DGX/GB200.
EPYC growth >50%. Gaining server CPU share from Intel. TAM raised to >$120B.
Bear Case
CUDA moat is extremely strong. ROCm is not competitive. 5M+ devs vs. ~500k.
P/E 50-60x is absurdly expensive. NVIDIA (leader) trades at 27x.
MI450 execution uncertain. Chip not yet delivered. Technical risks.
Customer concentration. Meta, OpenAI, Microsoft = too big to fail for AMD.
Scenario Summary
Scenario
Price
Prob.
Key Driver
🐂 Bull
$550-650
30%
MI450 succeeds, DC >$30B
📊 Base
$400-480
50%
DC +40%, MI450 solid
🐻 Bear
$250-300
20%
MI450 disappoints, DC slows
9. Final Assessment
Dimension
Rating
Assessment
Business Quality
★★★★☆
#2 in AI chips, strong CPU+GPU combo, good customers
Growth
★★★★☆
DC +57%, client +26%
Valuation
★★☆☆☆
P/E 50-60x — extremely expensive vs. NVIDIA 27x
Risk Level
★★★★☆
CUDA moat, MI450 execution risk
Risk/Reward
Unfavorable
Premium for #2 position vs. market leader
Verdict: Too Expensive / Speculative
AMD is operationally strong (+57% DC growth) with excellent customers (Meta, OpenAI, Microsoft). The MI450 launch in H2 2026 is a genuine catalyst that could establish AMD as a serious #2 in AI infrastructure.
However, at 50-60x forward earnings, AMD is significantly more expensive than NVIDIA (27x) despite having a smaller market share, slower growth, and a weaker software moat. The CUDA ecosystem remains the dominant barrier.
Recommendation: Too expensive at current levels. Wait for P/E <35x (~$300-350). Until then, NVIDIA offers better risk/reward (27x P/E, 85% growth, dominant CUDA moat).
This analysis was automatically generated. Not investment advice.