← Advanced Micro Devices, Inc.
AMD · Forward model · Bull case
The Bull case, 20 quarters out
Each segment is projected from its own operating driver, rolled up into consolidated cash flow, and discounted back to a fair value per share. The assumptions below are editable — change them and every number on this page moves with them.
AMD reports THREE operating segments — Data Center, Client and Gaming, Embedded — and discloses Client and Gaming dollars separately inside the second of those. This model carries FOUR verticals so Gaming's console trough is not hidden inside Ryzen. Every historical point is copied from the 8-Ks or derived from two disclosures (Q4 2025 Client = combined $3,940M less Gaming $843M; Q3 2025 Data Center = consolidated $9,246M less the three IR-rounded lines, an $11M remainder). Instinct versus EPYC inside Data Center is NOT split: AMD does not publish it. Helios racks have no ASP or unit count, so Data Center is a growth driver rather than a unit driver — inventing a rack price would look more rigorous and be less true. Fiscal quarters are 13 weeks; 2026 Q2 ended 27 June 2026 and is labelled 2026 Q2. Segment 'EBITDA' here is the disclosed segment operating margin. Client and Gaming OP is published only as a pair ($582M, 15.2%); the 16% and 12% on those two lines are assumed mix. Corporate 9.4% is the All other bucket (acquisition-related amortisation, stock-based compensation, deal costs), not a fourth business. netCash is cash $5,086M plus short-term investments $8,025M less total debt $3,226M. Capex $808M is 7% of the basis quarter and may be a Helios stand-up spike; terminal 4% assumes that. Q2 2025 Data Center OP of −$155M included $800M of MI308 China-export charges and is not a clean year-ago.
Helios ramps on schedule at Anthropic, Microsoft, Meta and Oracle. Data Center keeps a high sequential rate into 2027, mix lifts margin a few points, and the multiple holds a duration premium versus a generic semiconductor. What this case does NOT reach is Su's year-over-year double as a 2027 print — that is the named case.
Latest: $68.85B (2031Q2E)
| Period | Value |
|---|---|
| 2025Q1 | $7.44B |
| 2025Q2 | $7.68B |
| 2025Q3 | $9.25B |
| 2025Q4 | $10.27B |
| 2026Q1 | $10.25B |
| 2026Q2 | $11.54B |
| 2026Q3E | $13.43B |
| 2026Q4E | $15.42B |
| 2027Q1E | $17.50B |
| 2027Q2E | $19.67B |
| 2027Q3E | $21.90B |
| 2027Q4E | $24.21B |
| 2028Q1E | $26.59B |
| 2028Q2E | $29.04B |
| 2028Q3E | $31.59B |
| 2028Q4E | $34.22B |
| 2029Q1E | $36.97B |
| 2029Q2E | $39.83B |
| 2029Q3E | $42.83B |
| 2029Q4E | $45.97B |
| 2030Q1E | $49.28B |
| 2030Q2E | $52.76B |
| 2030Q3E | $56.44B |
| 2030Q4E | $60.34B |
| 2031Q1E | $64.47B |
| 2031Q2E | $68.85B |
What drives each segment
Data Center
Growth pathEPYC server CPUs and Instinct GPUs, now sold increasingly as Helios racks. 58% of the basis quarter, up 107% year over year. What paces it is packaging, HBM and customer-site power, none of which AMD publishes as a unit or megawatt series, so the projection is sequential growth on the reported segment. Lisa Su's 2027 double is the named case, not the opening rate.
Latest: $57.05B (2031Q2E)
| Period | Value |
|---|---|
| 2025Q1 | $3.67B |
| 2025Q2 | $3.24B |
| 2025Q3 | $4.29B |
| 2025Q4 | $5.38B |
| 2026Q1 | $5.78B |
| 2026Q2 | $6.72B |
| 2026Q3E | $8.24B |
| 2026Q4E | $9.89B |
| 2027Q1E | $11.63B |
| 2027Q2E | $13.45B |
| 2027Q3E | $15.36B |
| 2027Q4E | $17.34B |
| 2028Q1E | $19.40B |
| 2028Q2E | $21.54B |
| 2028Q3E | $23.76B |
| 2028Q4E | $26.07B |
| 2029Q1E | $28.49B |
| 2029Q2E | $31.01B |
| 2029Q3E | $33.67B |
| 2029Q4E | $36.46B |
| 2030Q1E | $39.41B |
| 2030Q2E | $42.53B |
| 2030Q3E | $45.84B |
| 2030Q4E | $49.35B |
| 2031Q1E | $53.08B |
| 2031Q2E | $57.05B |
Assumptions & reasoning
- A growth driver rather than units, and that is the honest answer. AMD publishes no GPU shipments, no Helios rack count and no ASP. Backing out a rack at an assumed $2-3M would look like the Nvidia model and would be fiction: even the customer GW headlines (Anthropic 2 GW, Meta 6 GW) are not a disclosed MW series.
- Opening 18% is the Q3 guide, not the trailing 16%. Combined with Client +6%, Gaming flat and Embedded +8% it prints $13.0B, the midpoint of $12.7-13.3B. Jean Hu said Data Center sales accelerate in the second half; this is that, not Su's 2027 double.
- Segment operating margin 31.3% ($2,103M on $6,718M). Q2 2025 was −$155M because of $800M of MI308 China-export charges, so the year-ago margin is not a base. Terminal 34% is a few points of Instinct mix, not a walk to Nvidia 70%.
- Q3 2025 $4,289M is the $11M remainder after subtracting the IR-rounded Client, Gaming and Embedded lines from consolidated $9,246M. IR printed Data Center as $4.3 billion.
Client
Growth pathRyzen CPUs for notebooks and desktops. $3.06B in the basis quarter, +23% year over year and +6% sequential. A PC-cycle line with share gains, disclosed every quarter inside Client and Gaming. Operating income is not split from Gaming, so the 16% margin is an assumption about mix.
Latest: $7.61B (2031Q2E)
| Period | Value |
|---|---|
| 2025Q1 | $2.29B |
| 2025Q2 | $2.50B |
| 2025Q3 | $2.80B |
| 2025Q4 | $3.10B |
| 2026Q1 | $2.88B |
| 2026Q2 | $3.06B |
| 2026Q3E | $3.31B |
| 2026Q4E | $3.55B |
| 2027Q1E | $3.78B |
| 2027Q2E | $4.00B |
| 2027Q3E | $4.22B |
| 2027Q4E | $4.43B |
| 2028Q1E | $4.64B |
| 2028Q2E | $4.85B |
| 2028Q3E | $5.06B |
| 2028Q4E | $5.27B |
| 2029Q1E | $5.48B |
| 2029Q2E | $5.69B |
| 2029Q3E | $5.91B |
| 2029Q4E | $6.13B |
| 2030Q1E | $6.36B |
| 2030Q2E | $6.60B |
| 2030Q3E | $6.84B |
| 2030Q4E | $7.09B |
| 2031Q1E | $7.34B |
| 2031Q2E | $7.61B |
Assumptions & reasoning
- Client and Gaming operating income is $582M, 15.2% of the combined $3,841M. Client is 80% of that revenue. 16% is the blend plus a point, an assumption, not a disclosure.
- Q4 2025 $3,097M is combined Client-and-Gaming $3,940M from the Q1 2026 slides less disclosed Gaming $843M. IR printed Client as a record $3.1 billion.
- Opening 6% is the trailing sequential rate, not the 23% year-over-year print. A reader who thinks AI PCs are a new cycle should take the slider up; a reader who thinks Q2 was a catch-up should take it down.
Gaming
Growth pathSemi-custom console SoCs plus Radeon discrete GPUs. $779M in the basis quarter, down 31% year over year on lower semi-custom. This is a console-cycle trough, not a franchise collapse. Opening growth is zero: the Q1-to-Q2 bounce is not a new cycle.
Latest: $1.24B (2031Q2E)
| Period | Value |
|---|---|
| 2025Q1 | $647M |
| 2025Q2 | $1.12B |
| 2025Q3 | $1.30B |
| 2025Q4 | $843M |
| 2026Q1 | $720M |
| 2026Q2 | $779M |
| 2026Q3E | $795M |
| 2026Q4E | $811M |
| 2027Q1E | $828M |
| 2027Q2E | $847M |
| 2027Q3E | $866M |
| 2027Q4E | $885M |
| 2028Q1E | $906M |
| 2028Q2E | $927M |
| 2028Q3E | $949M |
| 2028Q4E | $972M |
| 2029Q1E | $995M |
| 2029Q2E | $1.02B |
| 2029Q3E | $1.04B |
| 2029Q4E | $1.07B |
| 2030Q1E | $1.10B |
| 2030Q2E | $1.12B |
| 2030Q3E | $1.15B |
| 2030Q4E | $1.18B |
| 2031Q1E | $1.21B |
| 2031Q2E | $1.24B |
Assumptions & reasoning
- The 2025 H2 prints ($1.3B, $843M) were semi-custom. Q2 2026 at $779M is that cycle rolling off, which is why opening growth is zero rather than the +8% sequential bounce.
- Margin 12% is below the 15.2% Client-and-Gaming blend because semi-custom is the weaker half. AMD does not publish Gaming operating income. If you think discrete GPUs are the future of this line, the margin slider is where that lives.
- A next-generation console win would show up as a delayed jump, not as 0.5% terminal growth. That is a bull-case claim this line does not make.
Embedded
Growth pathXilinx FPGAs, embedded CPUs and adaptive SoCs. $977M in the basis quarter, +19% year over year, 39.5% segment operating margin. The recovery after 2025 customer-inventory digestion. A disclosed reportable segment with its own operating profit, unlike the Client/Gaming sub-lines.
Latest: $2.96B (2031Q2E)
| Period | Value |
|---|---|
| 2025Q1 | $823M |
| 2025Q2 | $824M |
| 2025Q3 | $857M |
| 2025Q4 | $950M |
| 2026Q1 | $873M |
| 2026Q2 | $977M |
| 2026Q3E | $1.08B |
| 2026Q4E | $1.17B |
| 2027Q1E | $1.27B |
| 2027Q2E | $1.36B |
| 2027Q3E | $1.46B |
| 2027Q4E | $1.55B |
| 2028Q1E | $1.64B |
| 2028Q2E | $1.73B |
| 2028Q3E | $1.82B |
| 2028Q4E | $1.91B |
| 2029Q1E | $2.01B |
| 2029Q2E | $2.10B |
| 2029Q3E | $2.20B |
| 2029Q4E | $2.30B |
| 2030Q1E | $2.40B |
| 2030Q2E | $2.51B |
| 2030Q3E | $2.61B |
| 2030Q4E | $2.72B |
| 2031Q1E | $2.84B |
| 2031Q2E | $2.96B |
Assumptions & reasoning
- Highest-margin disclosed segment and the smallest. Segment operating income $386M is 39.5% of $977M, published, not assumed. Terminal 40% holds that franchise rather than expanding it.
- FY2025 Embedded was down 3% on customer inventory. Q2 2026 +19% is digestion ending. Opening 8% is below the Q1-to-Q2 jump on purpose.
- Physical AI (Kria, Ryzen AI Embedded X100) is a content mix inside this line, not a fifth vertical. AMD has not published a robotics revenue series.
Where each case comes from
Su Double case — primary sources
The primary sources this case is built from — filings, calls and posts. Where they are claims by an interested party rather than disclosures, every number in the Su Double column is what happens if they are taken at face value.
AMD Q2 2026 earnings call, 4 August 2026
- Aug 4, 2026 During Tuesday's earnings call, AMD CEO Lisa Su said the company expects “data center segment revenue to more than double year-over-year in 2027,” as well.
- Aug 4, 2026 Taken together, we now expect data center segment revenue to more than double year-over-year in 2027.
- Aug 4, 2026 We expect Data Center sales to accelerate in the second half of 2026, driving stronger overall revenue growth and continued earnings expansion.
From cash flow to fair value
| Present value of free cash flow, 20 quarters | $96.12B |
| Terminal-year revenue | $250.11B |
| Terminal-year EBITDA | $63.94B |
| Exit multiple, on revenue | 10.0x |
| Terminal value | $2.50T |
| Discounted at 9.5% a year, terminal value becomes | $1.59T |
| Enterprise value | $1.68T |
| Net cash | $9.88B |
| Equity value | $1.69T |
| Shares | 1.66B |
| Fair value per share | $1,021.57 |
| Against the current price of $476.67 | +114% |
7x terminal revenue on a mid-20s blended operating margin — about 25x that margin — for a fabless semiconductor whose Data Center line has stopped compounding at 18% a quarter. Today the equity is ~$758B at $456.75, EV ~$748B after $9.9B net cash, 18x trailing sales of $41.3B and ~15x a ~$49B FY2026. Nvidia is richer on 60-70% margins this model does not give AMD. The exit is a de-rate, and deliberately. Discount rate is 11%, a point above Nvidia, because Instinct share at Helios scale is unproven and Q2 just showed capex can double in a quarter. Move the exit multiple before anything else: at 4x and at 12x the answer moves by more than Client, Gaming and Embedded combined. Nothing here prices Anthropic's 2 GW or Meta's 6 GW as contracted cash flow; they show up only as the Data Center growth rate they can sustain.
Read the other way round: at $476.67 the market is paying 4.3x terminal-year revenue, holding every other assumption on this page fixed. That is the number to argue about.
The projected path
| Quarter | Data Center | Client | Gaming | Embedded | Revenue | YoY | EBITDA | Capex | FCF | R40 | PV of FCF |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 2026 Q3E | $8.24B | $3.31B | $795M | $1.08B | $13.43B | +45% | $2.82B | $891M | $1.68B | +58 | $1.64B |
| 2026 Q4E | $9.89B | $3.55B | $811M | $1.17B | $15.42B | +50% | $3.32B | $975M | $2.04B | +63 | $1.95B |
| 2027 Q1E | $11.63B | $3.78B | $828M | $1.27B | $17.50B | +71% | $3.86B | $1.06B | $2.43B | +85 | $2.27B |
| 2027 Q2E | $13.45B | $4.00B | $847M | $1.36B | $19.67B | +70% | $4.42B | $1.14B | $2.85B | +85 | $2.60B |
| 2027 Q3E | $15.36B | $4.22B | $866M | $1.46B | $21.90B | +63% | $5.00B | $1.22B | $3.28B | +78 | $2.93B |
| 2027 Q4E | $17.34B | $4.43B | $885M | $1.55B | $24.21B | +57% | $5.61B | $1.31B | $3.74B | +72 | $3.26B |
| 2028 Q1E | $19.40B | $4.64B | $906M | $1.64B | $26.59B | +52% | $6.24B | $1.39B | $4.22B | +68 | $3.60B |
| 2028 Q2E | $21.54B | $4.85B | $927M | $1.73B | $29.04B | +48% | $6.89B | $1.48B | $4.71B | +64 | $3.93B |
| 2028 Q3E | $23.76B | $5.06B | $949M | $1.82B | $31.59B | +44% | $7.57B | $1.56B | $5.23B | +61 | $4.26B |
| 2028 Q4E | $26.07B | $5.27B | $972M | $1.91B | $34.22B | +41% | $8.28B | $1.65B | $5.76B | +58 | $4.59B |
| 2029 Q1E | $28.49B | $5.48B | $995M | $2.01B | $36.97B | +39% | $9.02B | $1.75B | $6.32B | +56 | $4.93B |
| 2029 Q2E | $31.01B | $5.69B | $1.02B | $2.10B | $39.83B | +37% | $9.79B | $1.85B | $6.91B | +54 | $5.26B |
| 2029 Q3E | $33.67B | $5.91B | $1.04B | $2.20B | $42.83B | +36% | $10.61B | $1.96B | $7.52B | +53 | $5.60B |
| 2029 Q4E | $36.46B | $6.13B | $1.07B | $2.30B | $45.97B | +34% | $11.46B | $2.07B | $8.17B | +52 | $5.95B |
| 2030 Q1E | $39.41B | $6.36B | $1.10B | $2.40B | $49.28B | +33% | $12.36B | $2.19B | $8.85B | +51 | $6.30B |
| 2030 Q2E | $42.53B | $6.60B | $1.12B | $2.51B | $52.76B | +32% | $13.31B | $2.32B | $9.57B | +51 | $6.65B |
| 2030 Q3E | $45.84B | $6.84B | $1.15B | $2.61B | $56.44B | +32% | $14.32B | $2.45B | $10.32B | +50 | $7.02B |
| 2030 Q4E | $49.35B | $7.09B | $1.18B | $2.72B | $60.34B | +31% | $15.38B | $2.59B | $11.13B | +50 | $7.40B |
| 2031 Q1E | $53.08B | $7.34B | $1.21B | $2.84B | $64.47B | +31% | $16.52B | $2.75B | $11.98B | +49 | $7.78B |
| 2031 Q2E | $57.05B | $7.61B | $1.24B | $2.96B | $68.85B | +31% | $17.72B | $2.91B | $12.88B | +49 | $8.18B |
Every row is projected. A year-over-year change is shown only where the quarter it compares against exists — an em dash means there is no comparable quarter, not a flat year.
Model revisions
Assumptions are marked to reality as each quarter prints. Every change is appended here, with the fair value the model produced at the time, so the model's own history stays visible.
| Date | Changed | Fair value then | Note |
|---|---|---|---|
| 2026-08-25 | all | $360.64 | Initial model. Four verticals on the disclosed Data Center / Client / Gaming / Embedded split, basis the June quarter at $11,536M, Q3 calibrated to the $13.0B ± $300M guide. |