Stock Analysis Expert - Comprehensive Equity Research & Valuation

Advanced Micro Devices (NASDAQ: AMD): Equity Research Note

Data as of September 22–23, 2026. Reference price is about $615–624 (record close of $623.77 on Sept 22). Market cap is about $1.0 trillion. Where sources disagree, I say so. Some ratios below are my own calculations and are labeled that way.

1. Executive Summary

AMD has become the clear #2 supplier of merchant AI accelerators. It also keeps taking share in server CPUs. In Q2 FY26, data center revenue grew 107% year over year to $6.7B, which was 58% of total revenue. Multi-gigawatt MI450/Helios commitments from OpenAI, Meta and Anthropic give it demand visibility into 2027. The problem is the price. The stock is up roughly 180% year to date and trades near 40x 2027 consensus EPS. It sits at or slightly above the average analyst target, and daily RSI is about 80. The long-term thesis looks strong, but a lot of it is already priced in.

Signal: HOLD (accumulate on pullbacks). Moderate conviction. For an aggressive investor holding 3–5 years, I'd build a position in stages as the stock pulls back rather than buy the full amount after a 5-day, roughly 20% run.

2. Fundamental Analysis

Financial health

Metric	FY2025	Q1 FY26	Q2 FY26	Q3 FY26 guidance
Revenue	$34.6B (record)	$10.3B (+38% YoY)	$11.54B (+50% YoY)	~$13.0B ±$0.3B (+41% YoY)
Gross margin (GAAP / non-GAAP)	50% / 52%	n/a	54% / 56%	~56% non-GAAP
Operating income (GAAP / non-GAAP)	$3.7B / $7.8B	n/a	$2.0B / $3.09B	n/a
Diluted EPS (GAAP / non-GAAP)	$2.65 / $4.17	$0.84 / $1.37	$1.38 / $1.66	n/a
Free cash flow	Record (Q4 alone: $2.1B)	n/a	$1.56B	n/a

Q2 segments:

Data Center: $6.72B (+107%)
Client: $3.06B (+23%)
Gaming: $779M (−31%)
Embedded: $977M (+19%)

Liquidity and leverage:

Cash and short-term investments were $13.1B at the end of Q2.
Debt is modest. From my background knowledge, total debt was about $3–4B; I did not re-verify this. Net cash is positive and debt-to-equity is well under 0.1x.
Watch item: capex has reportedly almost tripled year over year. That explains why FCF conversion trails net income right now.

Margins versus peers: A 56% non-GAAP gross margin is a record for AMD. Non-GAAP operating margin is about 27%. That is still far below NVIDIA, whose gross margins are roughly in the 70s% based on my background knowledge. The gap is both the bull case (room to expand) and the bear case (weaker pricing power).

Valuation (reference price about $620)

Metric	AMD	Comment
Trailing P/E (non-GAAP)	~108x (my calc: TTM EPS ≈ $5.76)	Published trailing P/Es range from 128x to 235x depending on GAAP vs. non-GAAP
Trailing P/E (GAAP)	~160x (my calc: TTM GAAP EPS ≈ $3.89)	Depressed by acquisition amortization
Forward P/E, 2027	~36–40x on $15.57 consensus	2027 consensus rose from $13.10 to $15.57 in 90 days; the Street-high estimate is ~$20.25
PEG	~0.4–0.6	24/7 Wall St. cites 0.57. Looks cheap only if ~100% EPS growth in 2027 happens
EV/EBITDA (TTM)	~80–90x (my rough estimate)	About 30x on 2027 estimates if consensus holds
P/B	~15x (estimate; book equity ≈ $65B, not verified)	Inflated by Xilinx goodwill; not very meaningful
P/FCF / FCF yield	152x / 0.66%	Weighed down by the capex surge
Dividend yield	0%	AMD pays no dividend. Buybacks were small in H1 FY26 (~$221M)

Peer comparison. Peer data was inconsistent across sources:

NVDA: forward P/E is cited anywhere from about 24x (Motley Fool, Aug 2026) to about 43x (another source). Either way it has a lower growth-adjusted multiple and much higher margins than AMD.
AVGO: forward P/E about 19–25x. The market treats it as the steadier, contract-backed AI custom-silicon play.
INTC / QCOM: I could not verify current multiples. Qualcomm is described as cheaper than Intel on forward P/E and P/S. INTC's market cap is about $648B, so AMD is now worth about 1.5 times Intel.
History: AMD has usually traded at about 30–45x forward earnings (from my background knowledge, not verified here). Today's 36–40x on 2027 earnings is at the top of that range, but not beyond it.

Bottom line on valuation: AMD is the most expensive large-cap AI chip stock relative to current earnings. It is only reasonably priced if the 2027 EPS ramp actually happens.

Competitive position

Moats:
- EPYC server CPU leadership. Analysts see AMD possibly passing Intel in data center CPU revenue within about 2 years.
- Chiplet and packaging know-how.
- Rack-scale systems capability (Helios, through the ZT Systems acquisition).
- ROCm software is improving but still trails NVIDIA's CUDA. That is AMD's biggest structural disadvantage.
Tailwinds:
- Hyperscalers want a second source to NVIDIA.
- Inference workloads favor AMD's memory-heavy GPUs.
- Customer commitments: OpenAI (6 GW), Meta (6 GW) and Anthropic (up to 2 GW of MI450, starting H1 2027).
- Helios began initial shipments late in Q3, with the bigger ramp in Q4 and through 2027.
Threats:
- NVIDIA's pace of new products (Rubin).
- Custom ASICs from Broadcom and hyperscaler in-house chips.
- Limited HBM and TSMC CoWoS supply.
- Gaming is weak (−31%).

Management & governance

CEO Lisa Su has one of the best execution records in semiconductors: the Zen turnaround, the Xilinx and ZT Systems deals, and selling ZT's manufacturing arm to stay asset-light.
Flags:
- Insider ownership is low (under 1%, typical for mega-caps).
- Buybacks have been small. Capital is going into capex and rack-scale buildout instead.
- Dilution: the October 2025 OpenAI agreement included a warrant for up to ~160M shares that vest on deployment and share-price milestones. This is from my background knowledge; check the latest 10-Q. It is dilutive if milestones are hit, but only in scenarios where the stock is doing well.
- No accounting red flags found. GAAP and non-GAAP differ mainly because of Xilinx amortization.

3. Technical Analysis (as of Sept 22, 2026)

Trend: A strong primary uptrend. The stock is in a rising channel, broke out above $584.90, and made a 5-day run to an all-time high. It is up about 180% year to date and about 290% over 12 months.
Resistance: $617.80 (upper channel), then $636.87, then $651.94.
Support: $584.90 (breakout level), then $541–560, then the $444–462 zone.
Moving averages: reported values are 50-day about $548 and 200-day about $498, with the 20-day above the 50-day and the 50-day above the 200-day (bullish stacking).
- ⚠️ These figures look inconsistent with the stock's path (about $223 in January, about $470 on Sept 1). The true 200-day is probably meaningfully lower. Check them on your own charting platform.
Momentum: RSI(14) is about 80–82, deeply overbought. One source's reading of 51.5 from mid-September is now stale. MACD is positive.
Volume: heavy on the breakout to $1T. Positive volume bias.
Read: The trend is healthy, but a short-term overbought reading after a vertical move usually leads to consolidation or a retest of the breakout level (about $585) or the 50-day.

4. Scorecard

Factor	Score (1–5)	Rationale
Valuation	2	About 40x 2027 EPS and about 108x trailing. PEG under 1 only if the ramp is delivered
Growth	5	Revenue +50%, data center +107%, guidance +41%, 2027 EPS possibly roughly doubling
Profitability	3	Record 56% gross margin, but well below NVDA/AVGO. GAAP operating margin about 17%
Balance Sheet	4	Net cash, $13.1B liquidity. Capex surge lowers FCF conversion
Momentum	5 (overbought)	All-time highs and bullish MAs, but RSI about 80 argues for patience

5. 12-Month Target Range (to about Sept 2027)

Scenario	Target	vs. $620	Key assumption
Bear	~$380	−39%	Helios/MI450 yields or HBM supply slip. 2027 EPS falls to about $13–14. Multiple compresses to about 27x as AI capex sentiment cools
Base	~$680	+10%	2027 EPS about $16 is met, 2028 view about $20. The market pays about 34x forward. Consensus target averages about $616–647
Bull	~$900	+45%	MI450 share gains beat expectations (2027 EPS about $20+, near the Street high), 2028 EPS about $26 at about 35x. More gigawatt-scale deals signed

The payoff is roughly symmetric at today's price: about 10% base-case upside against about 39% bear-case downside. That is why the call is HOLD/accumulate instead of an outright BUY.

6. Catalysts & Risks

Upside catalysts

Q3 FY26 earnings (expected early November): a beat on the ~$13B guidance and a strong Q4 guide as Helios ramps.
New hyperscaler or sovereign AI deals, and MI500 roadmap updates.
Further 2027 estimate revisions (33 up vs. 3 down in the last 30 days).
EPYC share gains as Intel's roadmap stalls.
Gross margin expanding past 56%.

Downside risks

Execution: Helios rack-level yields, HBM/CoWoS allocation, and customer deployment timing.
Concentration: a few hyperscaler and AI-lab customers drive a large share of GPU backlog. Commitments of 6 GW or more depend on those customers' own funding.
Regulatory: U.S. export controls. AMD took charges on China-bound MI308 chips.
Competition: NVIDIA Rubin, Broadcom custom ASICs, and in-house hyperscaler chips.
Macro/sector: a pause in AI capex, rate shocks, cyclical semiconductor drawdowns. AMD has historically fallen 50%+ in sector corrections.
Valuation: the stock sits at or above the average analyst target. Any disappointment could cause a sharp de-rating.
Dilution: vesting of the OpenAI warrant.

7. Suggested Entry/Exit & Position Sizing (general framework)

Total position size (aggressive profile): 4–7% of equity portfolio at cost, with a hard cap around 10% including gains. AMD has a beta of roughly 1.8–2.0 (my estimate), so size for volatility.
Staged entry:
- Tranche 1 (~1/3): on a retest of $580–590 (the breakout level) or after RSI cools below 60.
- Tranche 2 (~1/3): $530–560 (around the 50-day).
- Tranche 3 (~1/3): $450–500 (200-day zone / prior support), or after Q3 earnings confirm the Helios ramp.
Thesis review / risk stop:
- A weekly close below the 200-day MA, or a fundamental break: Helios/MI450 delays, data center growth under about 30%, or loss of a major gigawatt customer.
- For a 3–5 year holder, a thesis-based stop makes more sense than a tight price stop.
Trimming: trim above about $800–900 (bull-case area) or whenever the position exceeds your cap. Rebalance back to target weight instead of exiting fully while the thesis holds.

Assumptions & data caveats

Reference price is about $615–624 (Sept 22, 2026). Shares outstanding are about 1.63B diluted, implied from the $1T market cap.
I calculated trailing EPS, EV/EBITDA and P/B from reported quarterly results. Debt and book equity come from my background knowledge (cutoff June 2026) and were not re-verified.
Conflicting data:
- Forward P/E is quoted anywhere from 30x to 63x depending on source date and estimate year.
- The moving-average values look questionable.
- Peer multiples for INTC and QCOM could not be verified.
Check everything against live data before acting. Q3 results in early November will materially update this analysis.

Sources

This is educational analysis, not personalized financial advice.

Stock Analysis & Valuation Expert

Example run

Inputs used