Advanced Micro Devices: strong operating momentum, but the price needs far more
Sell at $615.53
A strong and improving business, but the price sits above even the Bull value once warrant dilution is counted.
- Downside to base value
- 59.4%
- Maximum buy price
- $125
- Required margin of safety
- 50%
- EV to trailing free cash flow
- 131x
- Market-implied 2030 FCF after SBC
- about $83 billion
- Potential warrant dilution
- up to 320 million shares
- Business quality
- 68/100
- Valuation score
- 18/100
- Confidence
- 13/25, grade C
AMD has rarely looked stronger operationally. Q2 2026 revenue reached $11.54 billion, up 50% year over year, and Data Center revenue more than doubled to $6.72 billion. The shares closed at $615.53 on September 21, 2026, giving the company an equity value of roughly $1.02 trillion.
The question is not whether AMD is a good business. It is. The question is how much future success the price already assumes.
At today's price, AMD trades at about 131 times trailing free cash flow. My Base case assumes management broadly delivers its ambitious long-term plan. Even so, it supports a value of about $250 per share. The market is paying for an outcome larger than my Bull case.
Economic classification: Quality Growth. Primary valuation method: growth-adjusted multiple on 2030 free cash flow after stock-based compensation, per fully diluted share. Secondary method: 10-year two-stage DCF.
The investment thesis
AMD is changing from a diversified CPU and graphics company into a core supplier of AI infrastructure.
EPYC server processors keep gaining share, Instinct GPUs are ramping, and the Helios rack-scale platform begins shipping in the second half of 2026. OpenAI and Meta have each committed to deploy up to 6 gigawatts of AMD accelerators over several product generations.
Management's November 2025 plan targeted revenue growth above 35% per year, a non-GAAP operating margin above 35% and non-GAAP earnings per share above $20 within three to five years. After Q2, management said it is tracking materially ahead of that plan.
The investment case therefore depends on three things:
- Data Center growth turns into durable free cash flow per share.
- Margins expand while large AI systems scale.
- The value created by the OpenAI and Meta agreements exceeds the shares given away to win them.
Routing and economic metric
AMD is routed to the SF-01 framework for hardware technology with positive free cash flow, with cycle normalization because semiconductors are cyclical.
The hardest honest metric is free cash flow after stock-based compensation, from continuing operations. Stock-based compensation is a real and recurring cost of about 5% of revenue, so neither headline free cash flow nor non-GAAP earnings per share is used as the valuation anchor.
Trailing twelve-month free cash flow was $7.73 billion. After $1.90 billion of stock-based compensation, it was $5.84 billion.
Current financial performance
| Metric | Value |
|---|---|
| Q2 2026 revenue | $11.54 billion, up 50% |
| Q2 2026 Data Center revenue | $6.72 billion, up 107% |
| Q2 2026 GAAP operating income | $1.99 billion |
| Q3 2026 revenue guidance | about $13.0 billion |
| Free cash flow, trailing twelve months | $7.73 billion |
| Stock-based compensation, trailing twelve months | $1.90 billion |
| Cash and short-term investments | $13.11 billion |
| Debt | $3.23 billion |
| Estimated Q3 diluted shares | 1.661 billion |
Free cash flow for 2025 was $5.52 billion from continuing operations. Some data services show about $6.7 billion because they include cash flow from discontinued operations. This analysis uses the continuing operations figure.
The warrant question
AMD issued OpenAI (October 2025) and Meta (February 2026) warrants for up to 160 million shares each, at an exercise price of $0.01. Vesting depends on GPU purchase milestones up to 6 gigawatts per customer, on AMD share price targets that rise to $600 for the final tranche, and on further technical and commercial conditions. As of June 27, 2026, no warrant shares had vested.
Together, the warrants could add 320 million shares, about 19% of today's diluted count. At a share price above $600, the final price hurdle is already met, so deployment volume is now the main vesting driver.
The same gigawatts that create revenue also create shares. This analysis therefore links the two in every scenario. Counting the revenue while ignoring the dilution would overstate value by roughly $80 to $100 per share in the Bull case.
| Scenario | Gigawatts deployed | Warrant shares | Fully diluted shares 2030 |
|---|---|---|---|
| Bear | about 2 of 12 | 53 million | 1.734 billion |
| Base | 8 of 12 | 213 million | 1.894 billion |
| Bull | 12 of 12 | 320 million | 2.001 billion |
Bear, base and bull valuation
| Scenario | Revenue 2027 | Revenue 2030 | FCF after SBC 2030 | Per share 2030 |
|---|---|---|---|---|
| Bear | $63.7 billion | $71.2 billion | $10.0 billion | $5.75 |
| Base | $80.8 billion | $141.5 billion | $31.1 billion | $16.43 |
| Bull | $87.5 billion | $188.3 billion | $48.9 billion | $24.46 |
Bear assumes an AI accelerator digestion period in 2029 and 2030, share loss to Nvidia and custom chips, and margins that stay near today's level. Revenue still doubles from 2025. Base assumes management's plan is broadly met: Data Center more than doubles in 2027, revenue grows 32.5% per year to 2030, and free cash flow after SBC reaches a 22% margin. Bull assumes the plan is exceeded, all 12 gigawatts are deployed and the mature margin reaches 26%.
Primary method: growth-adjusted multiple
The multiple starts from an 11x base for stable free cash flow companies and adds premiums for growth, returns, balance sheet and repeat revenue, less penalties for dilution and risk. The result is 16x in Bear, 20x in Base and 26x in Bull, applied to 2030 free cash flow after SBC. Free cash flow retained until 2030 and today's net cash are added, and the total is discounted back at 10%.
| Scenario | Multiple | Value |
|---|---|---|
| Bear | 16x | $80 |
| Base | 20x | $254 |
| Bull | 26x | $471 |
Secondary method: 10-year two-stage DCF
| Scenario | Discount rate | Terminal growth | Value |
|---|---|---|---|
| Bear | 11% | 2.5% | $74 |
| Base | 10% | 3.0% | $244 |
| Bull | 9% | 3.0% | $459 |
The terminal value accounts for 59% of the Base DCF, and the implied terminal multiple of 14.7x free cash flow is far below today's multiple.
Valuation triangulation
A 60% weight goes to the primary method and 40% to the DCF. This gives a Bear Value of about $77, a Base Value of about $250 and a Bull Value of about $466. The two methods differ by only 4% in Base, but they share the same operating forecast, so the agreement overstates their independence. A historical check that uses no forecast at all points to about $325: AMD's median year-end price to free cash flow from 2021 to 2024 was about 69x, against 131x today.
What is the market already pricing in?
At $615.53, and using the Base multiple, required return and share count, the price requires about $83 billion of free cash flow after SBC in 2030. The Base case produces $31 billion and the Bull case $49 billion.
At a 22% margin, $83 billion of free cash flow would require roughly $376 billion of revenue in 2030, a compound growth rate of about 61% per year from 2025. Management's own target is above 35%.
A simpler reverse test gives the same message. To justify today's enterprise value, AMD would need sustainable free cash flow of about $20.3 billion at 50x, $25.3 billion at 40x or $33.8 billion at 30x. That is 2.6 to 4.4 times trailing free cash flow, before counting warrant shares.
Where the stock becomes attractive
The Confidence Score of 13/25 (grade C) sets a starting margin of safety of 40%. Operating leverage and cyclicality add 5 points, and customer concentration adds another 5. The required margin of safety is therefore 50%.
Applied to the $250 Base Value, the maximum buy price is $125. At $615.53, the shares trade far above the buy zone and above the Bull value.
Reasons to own or follow AMD
- Data Center momentum. Segment revenue doubled year over year, and management expects it to more than double again in 2027.
- Server CPU leadership. EPYC keeps gaining share, and management targets more than 50% of server CPU revenue.
- Contracted visibility. Two 6 gigawatt customer commitments give AMD multi-year AI visibility it never had before.
- Balance sheet. Net cash of about $9.9 billion and record free cash flow.
- Execution. Management has repeatedly met or beaten its guidance.
What could go wrong?
- AI capex slowdown. If hyperscaler spending pauses, the growth priced into the shares does not arrive.
- Dilution. Up to 320 million warrant shares vest precisely when the business succeeds.
- Competition. Nvidia's rack-scale lead and customers' in-house chips can cap share and pricing.
- Supply and policy. AMD depends on TSMC and on export rules. An $800 million export-control charge hit Q2 2025.
Management execution
AMD has beaten revenue expectations quarter after quarter, and the long-term plan presented in November 2025 already looks conservative to management. The weaker parts of the record are capital allocation: the Xilinx acquisition added about $25 billion of goodwill that still depresses returns on capital, buybacks were made at high multiples, and the warrant agreements hand up to 19% of the company to two customers.
Management credibility is assessed as good: 15/20.
Stock Analyza scorecard
| Business quality | Score |
|---|---|
| Moat and sector positioning | 15/20 |
| Economic return versus cost of capital | 8/20 |
| Balance-sheet resilience | 13/15 |
| Earnings and cash flow quality | 9/15 |
| Growth and reinvestment runway | 13/15 |
| Management and capital allocation | 8/10 |
| Dilution and governance | 2/5 |
| Total | 68/100 |
AMD is a strong business with a long runway, but reported returns on capital remain low because of acquisition goodwill, and the warrant structure weighs on shareholder alignment.
| Valuation | Score |
|---|---|
| Discount to triangulated Base Value | 0/40 |
| Protection versus Bear Value | 0/20 |
| Cross-method convergence | 12/15 |
| Historical and peer valuation | 2/10 |
| Market-implied expectation asymmetry | 1/10 |
| Data and model quality | 3/5 |
| Total | 18/100 |
Confidence score
| Confidence | Score |
|---|---|
| Data quality and provenance | 4/5 |
| Business predictability | 2/5 |
| Valuation robustness | 3/5 |
| Accounting transparency | 3/5 |
| Scenario dispersion | 1/5 |
| Total | 13/25, grade C |
The main limitation is the width of the outcomes. The Bull value is six times the Bear value, because AI accelerator demand and customer concentration can move results sharply in either direction.
What I would watch from here
Green
- Data Center revenue growth above 60% year over year
- Free cash flow after SBC per diluted share rising more than 40% per year
- Non-GAAP gross margin at or above 56%
- Diluted shares including vested warrants below 1.75 billion
- OpenAI and Meta gigawatt deployments on schedule
Yellow
- Data Center growth between 30% and 60%
- Gross margin between 53% and 56%
- Diluted shares between 1.75 and 1.90 billion
- Deployments delayed by up to two quarters
- Stock-based compensation between 4% and 6% of revenue
Red
- Data Center growth below 30%
- Gross margin below 53%
- Diluted shares above 1.90 billion
- Deployments delayed by more than two quarters or cancelled
- Inventory rising faster than revenue
The three most important thesis breakers
These are the developments that would prove this cautious view wrong:
- 2027 revenue above $90 billion with free cash flow after SBC above 20% of revenue.
- Gigawatt milestones reached while diluted shares stay below 1.8 billion.
- Guidance implying 2030 free cash flow above $50 billion.
Adversarial review
The arithmetic review independently recomputed the scenario values, the valuation weights, the reverse valuation and the maximum buy price.
The review found and corrected one critical issue: an earlier draft counted revenue from the customer agreements without the warrant shares that come with it. Two further corrections moved values upward: forward rather than backward normalization of free cash flow, and a Bear multiple that avoids counting the same downturn twice.
The strongest bullish interpretation was tested directly. Combining Bull operations with a 30x exit multiple, an 8% discount rate and 3.5% terminal growth produces $579 to $608 per share, still below the current price. A case built on management's 38% growth path produces $234 to $244.
The gate status is PASS_WITH_WARNING because the primary and secondary methods share one operating forecast and the intermediate warrant price hurdles are not fully disclosed.
Final verdict: sell at $615.53
AMD is executing well, and its AI business may become very large. But at $615.53, the price already assumes about $83 billion of annual free cash flow after SBC by 2030, well above even the Bull case once warrant dilution is counted.
Verdict: SELL at $615.53. Triangulated Base Value $250, downside 59.4%, required margin of safety 50%, maximum buy price $125.
For current holders, this means reducing or selling. For new money, it means avoiding the stock at this price.
The main thesis breaker for this view is AMD delivering Bull-case volumes with margins and a valuation multiple above the Bull assumptions, the only combination that reaches today's price.
Sources
Primary sources include AMD's Q2 2026 earnings release, June 2026 Form 10-Q, 2025 Form 10-K, Financial Analyst Day release and the OpenAI partnership announcement.
The reference market price is the September 21, 2026 closing price of $615.53.
Disclaimer. This analysis is provided solely for research and educational purposes. It is not personalized investment, financial, legal or tax advice. Semiconductor companies are cyclical and exposed to customer concentration, competition, supply-chain and export-control risks. Valuation assumptions and management targets may prove incorrect. Investors can lose part or all of their invested capital.
Framework: SF-01, cyclical high-growth hardware tech subtype. Economic class: Quality Growth. Engine: growth-adjusted multiple with DCF cross-check. Confidence: C/13. Data status and adversarial gate: PASS_WITH_WARNING. Version: Master v3.1.