- Equity Research Cheat Sheets
- Posts
- AMD 3Q25 - Beat and raise all around. Is it enough? Bull case $350 Bear case $180
AMD 3Q25 - Beat and raise all around. Is it enough? Bull case $350 Bear case $180
Cheat Sheets
ADVANCED MICRO DEVICES (AMD)
📊 Revenue $9.25B (+36% YoY BEAT) | Adj EPS $1.20 vs $1.17 Est | Data Center $4.3B | "Well Over $100B" OpenAI Potential | AI TAM Expanding
Q4 Guide $9.6B (25% Growth) | FCF $1.5B Record | Turin 50% of EPYC Already | "Tens of Billions by 2027" | Hyperscalers Planning "Significant CPU Builds 2026" | Stock -4.17% AH
💰 Market Cap: $416B | 🏢 28K Employees | 🌍 Global Semiconductor Leader
👨💼 CEO Lisa Su | 🎯 AI-First Strategy | 🇺🇸 Santa Clara, CA
$239.62
📉 -$10.43 (-4.17%) After Hours
+107% YTD | Extended Selloff Continues
Price Targets (12-18 Months)
Current Price: $239.62 (After Hours)
$350.00
Bull Case (+46%)
2026 EPS: $6.50 | Forward P/E: 54x
AI DOMINANCE
🚀 Needs:
MI350/MI450 GPU ramp exceeds $15B annual run rate by 2027 with 20%+ AI accelerator market share • OpenAI 6GW deployment (starting H2 2026) expands to full scale with additional "multiple customers at similar scale" as CEO targets • EPYC Venice 2nm launch dominates 2026-27 as "strongest customer pull ever" converts to 45%+ server share • Hyperscaler "significant CPU builds into 2026" sustain as "durable multi-quarter phenomenon" drives server revenue • ROCm ecosystem achieves full parity enabling smooth customer experience across all frameworks and workloads
$280.00
Base Case (+17%)
2026 EPS: $5.80 | Forward P/E: 48x
STEADY EXECUTION
⚖️ Needs:
Data center AI business scales to $8-10B annually with MI350 continuing through H1 2026, MI450 ramping sharply H2 2026 per guidance • EPYC maintains 39-42% server CPU share with Venice "performing very well in labs" and early partner platforms already online • Hyperscaler CPU buildouts materialize as "durable demand" for general-purpose compute supporting AI infrastructure • Gross margins stabilize 54-55% as GPU margins "continue to improve" each generation, offsetting mix headwinds • Q4 $9.6B achieved, followed by mid-20s% growth in 2026 as data center momentum sustains • China MI308 situation remains status quo with "some licenses" but limited revenue contribution
$180.00
Bear Case (-25%)
2026 EPS: $4.20 | Forward P/E: 43x
AI SLOWDOWN
⚠️ Risk:
MI355 to MI450 transition creates H1 2026 revenue pause as "some customers wait for MI455" per past reports • Hyperscaler AI spending slowdown impacts AMD disproportionately given OpenAI concentration risk (potentially "half of revenue 2027-28" per analyst question) • CPU "significant builds" forecast fails to materialize if AI workload requirements were overstated or Intel recovers share • China MI308 remains "dynamic situation" with WIP inventory stranded and $1.5B+ revenue permanently lost • Gross margins compress toward 50% as GPU pricing competition intensifies and CFO admits margins "below corporate average" • Stock multiple crashes from 48x to low-30s P/E as execution concerns mount and valuation normalizes post-rally
The TL;DR
💰
What Happened
Record Revenue: $9.25B, up 36% YoY and 20% QoQ, crushing $8.74B estimate by $510M
Adj EPS: $1.20 vs $1.17 expected, up 30% YoY with strong operating leverage at 24% op margin
Data Center: $4.3B (+22% YoY, +34% QoQ) - servers up "a little bit better" than GPUs YoY per CFO
Free Cash Flow: Record $1.5B from $1.8B operating cash flow, up 3x YoY
Turin Acceleration: 5th-gen EPYC CPUs already 50% of total EPYC revenue in Q3, fastest ramp ever
Adj EPS: $1.20 vs $1.17 expected, up 30% YoY with strong operating leverage at 24% op margin
Data Center: $4.3B (+22% YoY, +34% QoQ) - servers up "a little bit better" than GPUs YoY per CFO
Free Cash Flow: Record $1.5B from $1.8B operating cash flow, up 3x YoY
Turin Acceleration: 5th-gen EPYC CPUs already 50% of total EPYC revenue in Q3, fastest ramp ever
📈
Why It Matters
AI Validation: OpenAI deal "well over $100 billion in revenue potential" - CEO Su confirmed "tens of billions annually by 2027"
CPU Surprise: Hyperscalers forecasting "significant CPU builds into 2026" - AI workloads driving unprecedented server demand
Multi-Quarter Visibility: "Durable, not short-term" CPU demand as AI workloads require massive general-purpose compute buildouts
Customer Momentum: Interest "accelerated" post-OpenAI with "engagements at higher scale" - multiple customers at "similar scale" targeted 2027-28
Cloud Expansion: 160+ new EPYC instances launched in Q3, total now 1,350+ globally (up 50% YoY)
CPU Surprise: Hyperscalers forecasting "significant CPU builds into 2026" - AI workloads driving unprecedented server demand
Multi-Quarter Visibility: "Durable, not short-term" CPU demand as AI workloads require massive general-purpose compute buildouts
Customer Momentum: Interest "accelerated" post-OpenAI with "engagements at higher scale" - multiple customers at "similar scale" targeted 2027-28
Cloud Expansion: 160+ new EPYC instances launched in Q3, total now 1,350+ globally (up 50% YoY)
🎯
What's Next
Q4 2025 Guide: $9.6B (±$300M) = 25% YoY, 4% QoQ growth - data center up double-digits, client up, gaming down strongly
MI355 Continues: Expected to "ramp in first half of 2026" with "sharper ramp in second half" as MI450 comes online
OpenAI Deployment: First 1GW of MI450 starts "second half of 2026" - multi-quarter planning underway with CSPs on power/supply
Venice Silicon: "In labs performing very well" - "strongest customer pull ever" with partners already online, 2nm launch 2026
Supply Chain Ready: "Feel very good" on capacity for "significant growth rates" - working 2 years out on power plans
MI355 Continues: Expected to "ramp in first half of 2026" with "sharper ramp in second half" as MI450 comes online
OpenAI Deployment: First 1GW of MI450 starts "second half of 2026" - multi-quarter planning underway with CSPs on power/supply
Venice Silicon: "In labs performing very well" - "strongest customer pull ever" with partners already online, 2nm launch 2026
Supply Chain Ready: "Feel very good" on capacity for "significant growth rates" - working 2 years out on power plans
💡
Bottom Line for Retail Investors
AMD delivered a masterclass quarter with record results across the board, but the 4.17% after-hours plunge to $239.62 (now down from $267 ATH) shows Mr. Market had baked in perfection after the 110% YTD run. The real story from the call: CEO Su sees "unprecedented" AI compute demand driving multi-year CPU and GPU growth, with OpenAI potentially worth "well over $100 billion" and multiple customers targeted at "similar scale" by 2027-28. The jaw-dropper? Hyperscalers are now forecasting "significant CPU builds into 2026" because AI workloads need massive general-purpose servers—a "durable, multi-quarter phenomenon" that could sustain 25%+ data center growth. The deeper selloff to $239 creates a better risk-reward setup than $250—at 46% upside to bull case and only 25% downside to bear case, the asymmetry has improved. For believers in the AI infrastructure buildout, this earnings-triggered weakness is exactly the entry point patient investors were waiting for. Lisa Su's confidence on supply chain readiness and the "strongest customer pull ever" for Venice CPUs suggests AMD isn't slowing down—the market just needed to reset expectations after the October moonshot.
🐂 Bull Thesis
🤖
AI Accelerator Breakthrough
OpenAI Scale: "Well over $100B revenue potential over next few years" - first 1GW MI450 deploys H2 2026, multi-year roadmap collaboration
Helios Reception: OCP demo was "phenomenal" with customers bringing engineering teams - "lot of excitement" driving "broader engagement at higher scale"
Multiple Hyperscalers: "Numerous customers" in deep engagements - dimensioning supply for "multiple customers at similar scale" as OpenAI by 2027-28
ROCm Progress: V7 delivers 4.6x inference, 3x training vs V6 - "most customers starting with AMD now have very smooth experience"
Helios Reception: OCP demo was "phenomenal" with customers bringing engineering teams - "lot of excitement" driving "broader engagement at higher scale"
Multiple Hyperscalers: "Numerous customers" in deep engagements - dimensioning supply for "multiple customers at similar scale" as OpenAI by 2027-28
ROCm Progress: V7 delivers 4.6x inference, 3x training vs V6 - "most customers starting with AMD now have very smooth experience"
⚡
CPU Market Share Domination
AI-Driven CPU Surge: Hyperscalers forecasting "significant CPU builds into 2026" - "durable, multi-quarter phenomenon" as AI needs general compute
Turin Fastest Ramp: Already 50% of EPYC revenue in Q3 - "extremely fast" adoption with ASP expansion driving mix improvement
Venice Demand: "Strongest customer pull we've seen" - multiple cloud/OEM partners already have first platforms online ahead of 2026 launch
Enterprise Traction: Commercial momentum accelerated 30%+ YoY with Fortune 500 wins across healthcare, financial services, manufacturing
Turin Fastest Ramp: Already 50% of EPYC revenue in Q3 - "extremely fast" adoption with ASP expansion driving mix improvement
Venice Demand: "Strongest customer pull we've seen" - multiple cloud/OEM partners already have first platforms online ahead of 2026 launch
Enterprise Traction: Commercial momentum accelerated 30%+ YoY with Fortune 500 wins across healthcare, financial services, manufacturing
📊
Financial Momentum Building
Revenue Acceleration: Q3 +36% YoY, Q4 guided +25% YoY - "clear step up in growth trajectory" with 20% sequential in Q3
Cash Generation: $1.5B free cash flow from $1.8B operating CF - returned $1.3B to shareholders via buybacks in 9 months
Margin Resilience: 54% Q3 gross margin maintained despite GPU ramp, guiding 54.5% Q4 - GPU margins "continue to improve" each generation
Operating Leverage: $2.2B operating income (24% margin) despite R&D up 42% to $2.8B - "strong momentum across business"
Cash Generation: $1.5B free cash flow from $1.8B operating CF - returned $1.3B to shareholders via buybacks in 9 months
Margin Resilience: 54% Q3 gross margin maintained despite GPU ramp, guiding 54.5% Q4 - GPU margins "continue to improve" each generation
Operating Leverage: $2.2B operating income (24% margin) despite R&D up 42% to $2.8B - "strong momentum across business"
🌐
Ecosystem Diversification
Client Record: $2.8B revenue (+46% YoY, +10% QoQ) - desktop CPU sales at all-time high with Ryzen 9000 leading
Gaming Surge: $1.3B (+181% YoY) driven by semi-custom (Sony/Microsoft holiday prep) and Radeon 9000 discrete GPU strength
Embedded Recovery: $857M, up 4% QoQ as demand strengthened - on track for second consecutive year of record $14B+ design wins
TAM Expansion: CEO Su: "AI compute TAM just going up" - 500B silicon TAM will be updated higher at Analyst Day next week
Gaming Surge: $1.3B (+181% YoY) driven by semi-custom (Sony/Microsoft holiday prep) and Radeon 9000 discrete GPU strength
Embedded Recovery: $857M, up 4% QoQ as demand strengthened - on track for second consecutive year of record $14B+ design wins
TAM Expansion: CEO Su: "AI compute TAM just going up" - 500B silicon TAM will be updated higher at Analyst Day next week
🐻 Bear Thesis
📈
Valuation Stretched Post-Rally
Continuing Selloff: Down 4.17% to $239.62 after-hours, now -10.3% from $267 ATH just days ago - momentum traders exiting aggressively
Expectations Reset: Even "$100B+" OpenAI quote and "tens of billions by 2027" guidance couldn't stop selling - market wanted more
Technical Breakdown Risk: Breaking below $240 support could trigger stop-losses and accelerate decline toward $220-230 range
Sentiment Shift: Broader tech weakness (Palantir -8%, NVDA -3.96% same day) suggests sector rotation away from AI names
Expectations Reset: Even "$100B+" OpenAI quote and "tens of billions by 2027" guidance couldn't stop selling - market wanted more
Technical Breakdown Risk: Breaking below $240 support could trigger stop-losses and accelerate decline toward $220-230 range
Sentiment Shift: Broader tech weakness (Palantir -8%, NVDA -3.96% same day) suggests sector rotation away from AI names
🎯
NVIDIA's Continued Dominance
Still Second Fiddle: OpenAI signed 10GW with NVIDIA vs 6GW with AMD - "world needs more compute" but AMD not first choice
Customer Concentration: Analyst asked if OpenAI could be "half of GPU revenue in 2027-28" - CEO deflected but didn't deny concentration risk
CUDA Moat Remains: While ROCm 7 shows progress, Su admits "obviously always more work to do" on libraries and frameworks
Supply Chain Parity: Everyone planning "multiple quarters out" on power/components - AMD's supply advantage unclear vs NVIDIA scale
Customer Concentration: Analyst asked if OpenAI could be "half of GPU revenue in 2027-28" - CEO deflected but didn't deny concentration risk
CUDA Moat Remains: While ROCm 7 shows progress, Su admits "obviously always more work to do" on libraries and frameworks
Supply Chain Parity: Everyone planning "multiple quarters out" on power/components - AMD's supply advantage unclear vs NVIDIA scale
🚧
China Export Restrictions
Still No Revenue: Q3 and Q4 exclude MI308 China sales - CFO: "pretty dynamic situation," received "some licenses" but demand uncertain
Work-in-Process Limbo: Have WIP inventory but "working with customers on demand environment" - unclear if/when converts to revenue
$1.5B Annual Hole: Potential revenue completely sidelined with no clear path to recovery in near term
Competitive Gap Widens: While AMD wrestles with MI308 licenses, NVIDIA and others capturing China demand through alternatives
Work-in-Process Limbo: Have WIP inventory but "working with customers on demand environment" - unclear if/when converts to revenue
$1.5B Annual Hole: Potential revenue completely sidelined with no clear path to recovery in near term
Competitive Gap Widens: While AMD wrestles with MI308 licenses, NVIDIA and others capturing China demand through alternatives
⏱️
Execution and Timing Risks
MI355 to MI450 Gap Risk: Su confirmed MI355 "continues to ramp in H1 2026" before MI450 in H2 - potential revenue air pocket if customers wait
Margin Headwinds: CFO acknowledged GPU margins "a little bit below corporate average" - scaling top-line prioritized over margin expansion
Q4 Gaming Decline: Gaming guided "down strong double digits" sequentially after strong Q3 - console cycle volatility continues
Helios Complexity: CEO admits rack-scale systems "certainly are" complex - execution risk on first major rack deployment with new ZT partnership
Margin Headwinds: CFO acknowledged GPU margins "a little bit below corporate average" - scaling top-line prioritized over margin expansion
Q4 Gaming Decline: Gaming guided "down strong double digits" sequentially after strong Q3 - console cycle volatility continues
Helios Complexity: CEO admits rack-scale systems "certainly are" complex - execution risk on first major rack deployment with new ZT partnership
This analysis is for informational purposes only and should not be considered investment advice. Please consult with a financial advisor before making investment decisions.

