Summary

Initiating at Buy, $457 target: the price already pays for most of the AI ramp, and we see 13% more

InitiationBy Penmark Team
AMDBuy$457.00+13%
Page 1 of 28
Equity research | Initiation of coverage | United States
Semiconductors | Coverage: Initiated | Published 6 November 2026, 08:00 ET
Penmark Research Team | research@penmarkresearch.com
Advanced Micro Devices, Inc. (Nasdaq: AMD)

Initiating at Buy, $457 target: the price already pays for most of the AI ramp, and we see 13% more

Rating
Initiation
Buy
12-month target
$457
Price, 5 Nov 2026
$404.40
Price return
+13.1%
Total return
+13.1%
Bull case, 12m
$576
Bear case, 12m
$312

We initiate at Buy with a $457 target, 13.1% total return and low conviction. AMD designs server and PC processors, Instinct AI accelerators and adaptive chips; Data Center is 81% of FY27E revenue. The $404 price implies FY27E Data Center AI growth of 148%; we model 170%.

  • Data Center AI grows 170% to $36.6bn in FY27E. Helios racks ramp at five lead customers; at the Street’s +137% the value is $394, at ours $425. Debate 1.
  • Server CPU doubles to $40.6bn on Venice. Guidance is about 2x; each extra 10pp of growth is worth $14 a share.
  • Operating margin hits 36% in FY27E; mix and warrants cost $20. Gross margin 54.5%, 1.1pp under the Street. Debates 2 and 3.

What the market is missing. Guidance of 80%+ growth implies FY27E revenue of $93.9bn or more; consensus is $90.8bn, we are $95.3bn (+5.0%); EPS $17.43 vs $16.27 (+7.2%).

Catalysts. NVIDIA results 17 Nov 2026; AMD Q4’26 on 2 Feb 2027 ($16.96bn vs $16.31bn consensus); Q1’27 Helios volume in May.

Valuation and target. 24.4x FY27E EPS (peer PEG 0.22x times 109% growth) gives $425 base. Bear $276, bull $510 at 25/55/20% weights is $405, or $457 rolled one year. DCF: $223.

Risks. US data-center policy −$74, Arm CPUs and NVIDIA Vera −$41, rack mix −$18 a share.

Exhibit 1: Non-GAAP EPS quadruples to $17.43 by FY27E while the P/E falls from 97x to 23x
Year end Dec, US$bnFY24AFY25AFY26EFY27EFY28E
Revenue25.834.652.195.3121.3
Growth y/y13.7%34.3%50.5%82.9%27.3%
Non-GAAP op. income6.17.815.134.045.1
Margin23.8%22.4%29.0%35.6%37.2%
Non-GAAP net income5.46.813.429.638.9
Non-GAAP EPS ($)3.314.178.0417.4322.12
Growth y/y–26.0%92.9%116.7%26.9%
Free cash flow2.46.74.719.632.2
At $404.40FY24AFY25AFY26EFY27EFY28E
P/E122.2x97.0x50.3x23.2x18.3x
EV / non-GAAP EBIT108.6x85.8x44.0x19.6x14.8x
P/B11.5x10.5x9.2x6.8x5.1x
FCF yield0.4%1.0%0.7%2.9%4.8%
Dividend yield0.0%0.0%0.0%0.0%0.0%
ROE (GAAP)2.9%7.2%14.8%29.2%28.9%
Net debt / EBITDAnet cashnet cashnet cashnet cashnet cash
P/E at our $457 target138.1x109.7x56.8x26.2x20.7x

Source: Company filings (A); Penmark Research model (E); Nasdaq close, 5 Nov 2026. Non-GAAP excludes stock compensation and acquisition amortisation. Shaded = estimates. Fuller ratios in Exhibits 8-9.

Base value is $425 a share today and $480 in 12 months, with a bull-to-bear range of +42% to −23%

Exhibit 3: Base $480, bull $576, bear $312 in 12 months, from $404
$100$200$300$400$500$600$700Jan-26Jul-26Jan-27Jul-27ActualEstimates →Price $404Target $457Bull $576 (+42%)Base $480 (+19%)Bear $312 (−23%)

Source: Nasdaq closes via Yahoo Finance; Penmark Research model. Case values at 5 Nov 2026 rolled forward one year at the 13.0% cost of equity; target is the probability-weighted value rolled forward.

Scenarios and price target basis
Bull$576+42%20% prob.

Server +110%, DC AI +220%; GM 55.5%. FY27E EPS $20.91; $510 today.

Base$480+19%55% prob.

Server +100%, DC AI +170%; GM 54.5%. FY27E EPS $17.43; $425 today.

Bear$312−23%25% prob.

Server +70%, DC AI +90%; GM 53.0%. FY27E EPS $11.31; $276 today.

Target, probability-weighted$457+13.1%
Upside / downside1.8xup $172, down $93
Exhibit 4: Non-GAAP EBIT rises 4.4x to $34.0bn by FY27E
Income, US$bnFY24AFY25AFY26EFY27EFY28E
Revenue25.834.652.195.3121.3
Gross profit13.818.229.052.066.7
Operating costs7.610.413.818.021.6
EBIT6.17.815.134.045.1
Stock comp.1.41.62.12.93.3
Net inc., adj.5.46.813.429.638.9
Net inc., GAAP1.64.310.025.434.7
D&A3.13.03.13.23.8
Exhibit 5: FCF dips to $4.7bn in FY26E on working capital
Cash flow, US$bnFY24AFY25AFY26EFY27EFY28E
Op. cash flow3.07.77.223.937.0
Capex(0.6)(1.0)(2.5)(4.3)(4.9)
Free cash flow2.46.74.719.632.2
FCF margin9.3%19.4%8.9%20.5%26.5%
Capex / sales2.5%2.8%4.9%4.5%4.0%
Stock comp.1.41.62.12.93.3
SBC / sales5.5%4.7%3.9%3.0%2.7%
FCF / share ($)1.474.122.8011.5118.27
Exhibit 6: Net cash grows to $60bn by FY28E with no buybacks
Balance, US$bnFY24AFY25AFY26EFY27EFY28E
Cash, investments5.110.615.735.367.5
Receivables6.26.312.518.422.4
Inventories5.77.914.820.923.6
Goodwill, intang.43.841.840.038.036.2
Total assets69.276.9102.4136.9178.2
Debt1.73.27.17.17.1
Net cash3.47.38.728.260.4
Equity57.663.072.9101.1139.0
Exhibit 7: Data Center is 81% of FY27E sales; EPS $17.43
Per share, KPIsFY24AFY25AFY26EFY27EFY28E
Shares (m)1,6371,6361,6601,7001,761
EPS, adj. ($)3.314.178.0417.4322.12
EPS, GAAP ($)1.002.656.0414.9319.69
DPS ($)0.000.000.000.000.00
BVPS ($)35.1738.5143.8859.4678.95
Data Center, $bn12.616.633.977.2102.0
DC % of sales48.8%48.0%64.9%81.0%84.1%
Revenue growth13.7%34.3%50.5%82.9%27.3%

Source: Company filings (A); Penmark Research model (E). Shaded columns are estimates (E). Non-GAAP excludes stock compensation and acquisition amortisation. Net cash = cash and short-term investments less debt. FY26E-FY28E diluted shares include customer warrants as they vest.

Margins widen to 37% EBIT and ROE to 29% as revenue nearly doubles; we sit 3% to 7% above consensus EPS through FY27E

Exhibit 8: EBIT margin expands 13pp to 36% by FY27E
Growth, margins (%)FY24AFY25AFY26EFY27EFY28E
Growth y/y
Sales13.7%34.3%50.5%82.9%27.3%
Non-GAAP EBIT–26.6%94.9%124.3%32.9%
Non-GAAP EPS–26.0%92.9%116.7%26.9%
Free cash flow–180%−31%320.8%64.5%
Margins
Gross margin53.4%52.4%55.6%54.5%55.0%
EBIT margin23.8%22.4%29.0%35.6%37.2%
Net margin21.0%19.7%25.6%31.1%32.1%
FCF margin9.3%19.4%8.9%20.5%26.5%
Capex / sales2.5%2.8%4.9%4.5%4.0%
Exhibit 9: P/E falls to 23x FY27E, 26x at our target
Ratios, valuationFY24AFY25AFY26EFY27EFY28E
Returns and leverage
ROE (GAAP)2.9%7.2%14.8%29.2%28.9%
ROA (GAAP)2.4%5.9%11.2%21.2%22.0%
ROCE10.0%12.0%19.9%34.3%34.2%
Net debt / equitynet cashnet cashnet cashnet cashnet cash
Net debt / EBITnet cashnet cashnet cashnet cashnet cash
Valuation at $404.40
P/E122.2x97.0x50.3x23.2x18.3x
EV / EBIT108.6x85.8x44.0x19.6x14.8x
P/B11.5x10.5x9.2x6.8x5.1x
FCF yield0.4%1.0%0.7%2.9%4.8%
Dividend yield0.0%0.0%0.0%0.0%0.0%
P/E at target138.1x109.7x56.8x26.2x20.7x

Source: Company filings (A); Penmark Research model (E). Non-GAAP EBIT stands in for EBITDA (the model’s EBITDA-equivalent line). ROE, ROA and ROCE on average GAAP capital. Multiples at the 5 Nov 2026 close except the last row. AMD pays no dividend and holds net cash, so dividend and net-debt rows read nil.

Exhibit 10: We are 3.1% above consensus on FY26E EPS and 7.2% above on FY27E
US$m, $/shareQ4’26EFY26EFY27EFY28E
Revenue
Penmark16,96452,13095,336121,319
Consensus16,31051,47690,830126,790
Prior year10,27034,63952,13095,336
Diff. vs consensus+4.0%+1.3%+5.0%−4.3%
Growth y/y+65.2%+50.5%+82.9%+27.3%
Non-GAAP EPS ($)
Penmark2.918.0417.4322.12
Consensus2.687.8016.2723.00
Prior year1.534.178.0417.43
Diff. vs consensus+8.6%+3.1%+7.2%−3.8%
Growth y/y+90.1%+92.9%+116.7%+26.9%

Source: Company filings, guidance of 3 Nov 2026; consensus per InvestingPro, 5 Nov 2026; Penmark Research model. Order: Penmark, consensus, prior-year period, difference, growth.

Exhibit 11: EPS compounds to $22.12 by FY28E, 7% above consensus in FY27E
Consensus$0$5$10$15$20$25FY24AFY25AFY26EFY27EFY28EActualEstimates →

Source: Company filings (A); Penmark Research model (E); consensus per InvestingPro, 5 Nov 2026.

What it means for valueEach $1 of FY27E EPS is worth $24 a share at our 24.4x multiple, so our $1.16 premium to consensus is worth about $28. FY27E EPS bridge: revenue above consensus adds $1.18, a gross margin 1.1pp lower takes off $0.54, a higher share count $0.25, and opex and other lines add $0.77. FY28E does not enter the value.

Next catalyst is NVIDIA on 17 November; AMD trades 66% above peers on P/E but 11% below on PEG using our EPS

Exhibit 12: AMD trades at 24.9x consensus FY2 EPS against 15.0x for AI-compute peers, with 38pp more EPS growth
FY2, at 5 Nov priceAMDPeersDiff.
Valuation
P/E, consensus24.9x15.0x+66%
P/E, our EPS23.2x15.0x+55%
PEG, consensus0.23x0.22x+2%
PEG, our EPS0.20x0.22x−11%
Growth and returns
EPS growth109%71%+38pp
Gross margin, TTM GAAP50%67%−17pp
ROE, TTM GAAP8%33%−25pp
Debt / market cap1.0%3.2%−2.2pp
Beta, 5Y, raw2.301.96+0.34
What the comparison saysOn our EPS AMD trades at a PEG of 0.20x against a 0.22x peer median. The 24.4x target multiple is the 0.22x peer median PEG times AMD’s 109% consensus growth; at NVIDIA’s 0.16x PEG the value would be $291 today ($329 in 12 months).

Source: Company filings, S&P Global and other consensus (7-30 Oct 2026), Nasdaq closes 5 Nov 2026; Penmark Research model. Peers: NVIDIA, Broadcom, Marvell median. Peer beta is the raw 5-year median; the model relevers it to 1.93 at AMD’s leverage. Difference in % for multiples, pp for ratios.

Exhibit 13: Four dated catalysts in the next six months
  • 17 Nov 2026NVIDIA FQ3 results: AI demand read-across; Bull 1 and Bear 3 context.
  • 2 Feb 2027AMD Q4’26: we expect $16.96bn revenue and $2.91 EPS vs consensus $16.31bn and $2.68; FY27 gross margin commentary.
  • Feb 2027FY26 10-K: warrant vesting, purchase commitments and Helios customer disclosure.
  • May 2027Q1’27: Helios volume and GW shipped; Server CPU doubling on track.

Source: Company calendar, Penmark Research.

Exhibit 14: Policy and Arm risks cost $115 a share; Helios adds $46
  • US data-center policy. Permit pauses slow 2027 GW. Likelihood Low (20%); −$74/share.
  • Arm CPUs and NVIDIA Vera. Arm and NVIDIA CPUs take sockets. Likelihood Medium (30%); −$41/share.
  • Rack mix dilutes margin. Racks pull gross margin below 54.5%. Likelihood Medium (35%); −$18/share.
  • Upside, against the cautious view: Helios ramps faster. Lead customers pull volume forward. Likelihood High (60%); +$46/share.
The thesis is invalidated ifFY27E Data Center AI growth runs below the 148% the price implies, or the target PEG falls to 0.16x (−18.7%, a Sell).

Source: Penmark Research model, catalyst table.

Exhibit 15: Trailing P/E of 60.8x is 12 turns above its 48.6x average since 2022
P/E, trailing four quarters non-GAAP0x25x50x75x100x125x2023202420252026Mean 49x; bands ±1σ, ±2σ61x

Source: Nasdaq closes via Yahoo Finance; reported non-GAAP quarterly EPS (trailing four quarters); weekly, Jan 2022 to 5 Nov 2026.

Exhibit 16: P/E against EPS growth: R² of 0.04 means the peer line explains little
10x20x30x40x−25%0%25%50%75%100%125%P/E on FY2 consensusEPS growth, FY1 to FY2R² 0.04, n=5AMDNVDAAVGOMRVLINTCQCOM

Source: Company filings, consensus (7–30 Oct 2026), Nasdaq closes 5 Nov 2026; Penmark Research. NVIDIA, Broadcom, Marvell, Intel, Qualcomm are not covered by Penmark.

PART 1 · SUMMARY · 3 THE THESIS IN CHARTS

Data Center revenue drives the stock, AI is 53% of FY27E growth, and a 24.4x target multiple gives 13% to our target

Exhibit 17: Share price follows Data Center revenue: correlation 0.91 over eight quarters
Data Center revenue ($bn), leftShare price at quarter end ($), right$0bn$2bn$4bn$6bn$8bn$10bn$0$200$400$600$8004Q241Q252Q253Q254Q251Q262Q263Q26

Source: Company filings (Data Center segment revenue); Nasdaq quarter-end closes via Yahoo Finance; Penmark Research. Eight quarters, 4Q24 to 3Q26.

Exhibit 18: Data Center AI adds $23bn of the $43bn rise in FY27E revenue, 53% of growth in our base case
Server CPUData Center AIClient and GamingEmbedded$0bn$25bn$50bn$75bn$100bn$125bnFY26EFY27EFY28E$52bn$95bn$121bn

Source: Penmark Research model, revenue build (US$bn). Server CPU and Data Center AI split is not disclosed; FY26E split 60/40 from management commentary.

Exhibit 19: Cumulative Instinct shipments of 2.8GW cover the 2GW of warrants vested in FY27E
Instinct GW shipped, cumulative from FY27ECustomer warrant GW vested, cumulative0.02.55.07.510.012.5FY27EFY28EFY29E2.86.811.7

Source: Penmark Research model: Data Center AI revenue per GW; OpenAI and Meta warrants (8-Ks, 6 Oct 2025 and 24 Feb 2026). GW, cumulative.

Exhibit 20: Revenue is 1% above consensus in FY26E, 5% above in FY27E and 4% below in FY28E
Consensus$0bn$50bn$100bn$150bnFY26EFY27EFY28E

Source: Penmark Research estimates (E); consensus per InvestingPro, 5 Nov 2026. Nine months reported plus Q4 consensus for FY26E.

Exhibit 21: P/E methods span $276–$510; the DCF at a 13% discount rate, at $188–277, is the outlier
Bear to bull, P/E valuePeer PEG range, P/E valueDCF, WACC and growthgridStreet targets, low to high$0$500$1000$1500$276$510$291$446$188$277$430$1250Price $404Target $457

Source: Penmark Research model; Street targets per stockanalysis.com, 4 Nov 2026. Dots are point values: base, weighted, DCF, Street mean.

Exhibit 22: Each 3 turns of P/E is worth about $52 a share; each $0.87 of EPS about $21
$288$304$320$336$352$335$354$373$391$410$382$404$425$446$467$429$453$477$501$525$477$503$529$556$582FY27E non-GAAP EPS$15.69$16.56$17.43$18.30$19.18Target P/E18.4x21.4x24.4x27.4x30.4x

Source: Penmark Research model, sensitivity grid; centre is base-case value of $425 at 24.4x and $17.43.

PART 1 · SUMMARY · 4 KEY DEBATES

We differ from the market on four debates; the three we can price net to +$11 a share, against a $20 gap to base value

Debates 1 to 3 are priced in the model. Debate 1 is covered in section 10, debate 2 in section 11 and debate 3 in section 12; debate 4 sits in the bear case.

Exhibit 23: We are above the market on growth and below it on margin and dilution
DebateMarket viewPenmark viewEvidenceWhat would change our viewResolves
1. Data Center AI growth in FY27E+137% (value $394). Consensus revenue $90.8bn sits below the 80%+ guide.+170% to $36.6bn (value $425, +$31). Helios ramps at five lead customers; Q4 guide has Data Center up 50% q/q.Q3’26 Data Center $8.5bn, +97% y/y; AI passes server in Q4; 2.8GW shipped in FY27E.Q4’26 Data Center below +45% q/q, or FY27 AI guide at or below 2x.Feb 2027
2. FY27E gross margin55.6% (value $438). Q3’26 was 56%.54.5% (value $425, −$13). Rack-scale Helios carries more third-party content; no FY27 margin guide.Q4’26 guide is 55% against 56% in Q3; non-GAAP margin was 52% in FY25.Q4’26 margin of 56% or above with racks already shipping.Feb 2027
3. Warrant dilutionNo dilution before 2028 (value $432); models use about 1.68bn shares.2GW vests in FY27E, 53m more shares and −$7 a share. OpenAI and Meta can take up to 320m shares at $0.01.Up to 160m shares each, vesting with GW shipped; shipments cover vested GW.FY26 10-K shows no vesting before the first full GW ships.Feb 2027
4. US data-center policy5 Nov selloff (−17%) prices a real chance of curbs after the midterms.We carry it as the largest bear case: −$74 at 20% probability, already inside the 25% bear weight.Texas grid and permit pauses (Aug, Sep); federal bill S.4214 still in committee.Passage of a federal bill, or a statewide order in a top-three AI market.2027 sessions

Source: Company reports, S&P Global and InvestingPro consensus (28 Oct-5 Nov 2026), Penmark Research model.

Exhibit 24: Debates are worth +$31, −$13 and −$7 a share against the market's view
Value at market's viewValue at our view1. Data Center AI +137% vs+170%2. Gross margin 55.6% vs54.5%3. Warrant GW vested 0 vs2$394$425$425$438$425$432

Source: Penmark Research model, Market Check: value per share with one driver at the market view and at ours, others at base.

How the debates map to the thesis

Growth is worth +$31; margin and dilution take back $20. The net of $11 leaves base value $425, 5% above the price; the 13% return to target adds the cost-of-equity roll.

Why the market may be wrong

The price already assumes 148% AI growth. Guidance of 80%+ growth implies at least $93.9bn of FY27E revenue, 3% above consensus; the 17% fall on 5 November was policy fear.

Exhibit 25: Debates 1 to 3 resolve in February 2027, three months from now
17 Nov 2026NVIDIA FQ3AI demand read-acrossDebates 1, 2AMD Q4’262 Feb: DC AI, margin, guideDebate 3FY26 10-K26 Feb (est.): warrants4 May 2027Q1’27Helios volume (est.)3 Aug 2027Q2’27MI450 ramp (est.)2 Nov 2027Q3’27FY28 framework (est.)

Source: Company calendar and Yahoo Finance earnings dates (2 Feb 2027); the 10-K and later dates are Penmark estimates of the usual reporting cadence.

PART 2 · THE BUSINESS · 5 HOW THE COMPANY MAKES MONEY

AMD sells chips into four end markets, and Data Center is moving from 48% of revenue in FY25A to 65% in FY26E

Exhibit 26: Of $52.1bn FY26E revenue, $15.1bn (29 cents in each dollar) becomes non-GAAP operating profit
Server CPU$20.3bnData Center AI$13.5bnClient$11.2bnGaming$2.9bnEmbedded$4.2bnRevenue$52.1bnCost of revenue$23.1bnGross profit$29.0bnNon-GAAP opex (R&D, marketing, admin)$13.8bnOperating profit$15.1bn

Source: Penmark Research model, FY26E (nine months reported plus Q4 guide); non-GAAP. US$bn. Server CPU and Data Center AI split is not disclosed by AMD; the model splits it 60/40 from management commentary.

Exhibit 27: Q3’26 shows the mix shift: Data Center is 64% of revenue and grew 97% y/y
$13.4bn+44.7% y/yQ3’26 revenue$8.5bn+97% y/yData Center55.9%vs 54.0% in Q3’25Non-GAAP gross margin29.3%vs 24.2% in Q3’25Non-GAAP operatingmargin$2.09vs $1.20 in Q3’25Non-GAAP EPS

Source: Company filings, Q3’26 release (26 Sep 2026 quarter); Penmark Research model.

What it sells, and why customers buy

AMD designs the chips and TSMC builds them. Data Center sells EPYC server CPUs and Instinct accelerators (MI455X in Helios racks) to cloud and AI customers; Client sells Ryzen PC processors; Gaming sells console chips; Embedded sells FPGAs and adaptive chips. Customers buy for performance per watt and a second source to NVIDIA and Intel.

How it is priced

Revenue is units times price times mix. Accelerators are priced per rack or per GW; our build gives about $13bn of Data Center AI revenue per GW shipped.

Cash versus profit

Profit runs ahead of cash in the ramp. FY26E free cash flow is $4.7bn on $15.1bn of operating profit because working capital absorbs $8.2bn. Inventory reached $11.5bn at Q3’26 and unconditional purchase commitments are $48.8bn.

What this showsOperating leverage is the story: 29% of revenue reaches profit in FY26E and 36% in FY27E, because opex grows 30% against revenue growth of 83%.
Exhibit 28: AMD earns its margin in chip design; wafers, memory and packaging are bought in, and $48.8bn of purchase commitments are disclosed off balance sheet
Design (EDA, IP)AMD, Arm and x86 rivalsR&D $8.1bn in FY25AWafers and packagingTSMC and outsourcedassemblySupply sets 2027 volumeMemoryHBM and DRAM suppliersCost inflation hit PCsSystemsAMD Helios racks, OEMand ODM partnersRack mix dilutes marginCustomersHyperscalers, AI labs, PCmakers, consolesNo 10% customer in FY25Chip designRacks

Source: Company filings (FY25 10-K; Q3’26 10-Q commitments at 26 Sep 2026); Penmark Research. AMD is fabless: it does not own wafer fabrication.

PART 2 · THE BUSINESS · 5 HOW THE COMPANY MAKES MONEY (CONTINUED)

Revenue is a tree with two branches that matter: server CPU and Data Center AI explain $43bn of FY27E growth

Exhibit 29: Data Center is $33.9bn of the $52.1bn FY26E revenue tree and carries both of the drivers we forecast
Server CPU$20.3bnFY27E +100%: units x priceData Center AI$13.5bnFY27E +170%: GW x $/GWData Center$33.9bn, +103.5%Client$11.2bn, +5.2%FY27E −8%Gaming$2.9bn, −26.5%Client and Gaming$14.1bn, −3.3%Embedded$4.2bn, +21.6%FY27E +25%Revenue FY26E$52.1bn, +50.5%

Source: Company filings (FY25A); Penmark Research model (FY26E). Highlighted boxes are the forecast drivers in sections 10–12 and 14. Server CPU and Data Center AI are model estimates.

Exhibit 30: Server CPU and Data Center AI add $43.3bn to FY27E revenue; Client, Gaming and Embedded net to −$0.1bn
$0bn$20bn$40bn$60bn$80bn$100bnFY26EServerDC AIClientGamingEmbed.FY27E$52.1bn+$20.3bn+$23.0bn−$0.9bn−$0.3bn+$1.1bn$95.3bn

Source: Penmark Research model, revenue build, US$bn. Components may not sum due to rounding.

Exhibit 31: Q3’26 y/y growth runs from +97% in Data Center to −40% in Gaming
Data CenterEmbeddedClientGaming+97%+29%+7%−40%

Source: Company filings, Q3’26 release; growth versus Q3’25.

Exhibit 32: Data Center is 81% of FY27E revenue; Embedded is 6%
Server CPU$40.6bn (43%)Data Center AI$36.6bn (38%)ClientEmbedded$5.3bn (6%)Gaming

Source: Penmark Research model, FY27E revenue, US$bn. Client $10.3bn (11%), Gaming $2.6bn (3%).

Cost to deliver

Fabless means variable cost of revenue and fixed R&D. Cost of revenue is 44% of FY26E revenue and scales with units; non-GAAP opex of $13.8bn is mostly engineers (about 31,000 staff) and does not. R&D was $8.1bn in FY25A.

Concentration

No customer was 10% of revenue in FY25 or FY24. TSMC supplies wafers, and memory and packaging cost is the supply risk the Street tracks.

PART 2 · THE BUSINESS · 6 THE MARKET: AI ACCELERATORS AND DATA-CENTER CPUs

The market is growing about 55% a year, but a $51bn FY28E AMD AI line rests on customers who are still raising capex

Exhibit 33: AI spending is still rising: three hyperscalers plan about $560bn of 2026 capex
$193bn+55% on $125bn in 2025Data-center GPU market,2026E$560bnrange midpointsAmazon, Alphabet, Metacapex, 202627.8%units, TrendForceASIC share of AI servers,2026EUp to 12GW1GW firm commitmenteachOpenAI and Meta dealswith AMD11after 3 Nov votesOhio local data-centerbans

Source: Global Market Insights (GPU market, excludes custom ASICs); company capex guidance (Amazon, Alphabet, Meta); TrendForce; AMD 8-Ks (6 Oct 2025, 24 Feb 2026); press reports on Ohio. Vendor market sizes differ widely by scope.

Exhibit 34: Amazon, Alphabet and Meta alone guide to $560bn of 2026 capex; Oracle adds $92bn
Amazon (2026)Alphabet (2026)Meta (2026)Oracle (FY27)$220bn$200bn$140bn$92bn

Source: Company guidance; Alphabet $195–205bn, Meta $135–145bn, Oracle $90–95bn gross (FY27) shown at midpoints; Amazon about $220bn cash capex. Microsoft excluded: year basis unclear.

Exhibit 35: Third-party estimates put AMD at 7% to 10% of AI accelerators in 2026, NVIDIA near 75%
AMDNVIDIAOther202580%12%7%2026E8%75%16%

Source: Third-party estimates (2025: NVIDIA 80–81%, AMD 6–8%; 2026: NVIDIA 75%, AMD 7–10%), shown at midpoints with other as remainder. Not AMD disclosures; estimates vary by source.

Exhibit 36: US policy risk is a string of local pauses so far; the federal moratorium bill has not left committee
25 Mar 2026S.4214 introducedFederal moratorium bill, incommittee3 Aug 2026Texas grid pauseERCOT interconnection21 Sep 2026Texas permit pauseEnvironmental permits3 Nov 2026Ohio votes11 local bans reported5 Nov 2026Chip selloffAMD −17%, SOX −12%

Source: ERCOT and Texas agency notices; congress.gov (S.4214); press reports (3–5 Nov 2026). A broader Texas moratorium was proposed, not issued. Ohio and Maine measures are local, not statewide.

Exhibit 37: Five platforms compete for the same AI budgets through 2028, and AMD is the only merchant alternative at volume
VendorPlatform and timingWhy it matters to AMD
NVIDIAVera Rubin NVL72: 72 Rubin GPUs and 36 Vera CPUs per rackRack-scale rival to Helios; Vera also contests server CPU sockets
AMDMI455X in Helios racks shipping in volume since Q3’26; MI500 planned for 2027Five lead customers: Anthropic, Meta, Microsoft, OpenAI, Oracle
BroadcomCustom accelerators with OpenAI: 10GW target, H2 2026 to end-2029Takes AI budget that could go to merchant GPUs
Google and AWSIronwood TPU now; Trainium4 expected 2027In-house silicon at the largest buyers
QualcommAI250 sampling expected mid-2027; Dragonfly C1000 in 2028Later entrant; limited effect before 2028

Source: Company announcements and press reports, Q3’26 call (3 Nov 2026); timing for 2027 and 2028 products is company or press expectation.

PART 2 · THE BUSINESS · 7 COMPETITIVE POSITION

AMD has taken about 10 points of x86 server revenue share between Q4’24 and Q1’26, but its AI position still sits behind a 75% margin leader

Exhibit 38: AMD's x86 server revenue share rose from 36% to 46% between Q4’24 and Q1’26, third-party estimates
AMD unit share (%)AMD revenue share (%)Q4’24Q4’25Q1’26Q2’2626362941334634

Source: Third-party x86 server share estimates via press reports; AMD does not disclose share. Only quarters with published estimates are shown (Q4’24, Q4’25, Q1’26, Q2’26); revenue share for Q2 2026 is not available.

Exhibit 39: Client unit share is 30%: notebook is gaining, desktop slipped from its 36% peak
Desktop (%)Notebook (%)Q4’24Q4’25Q2’26272436263529

Source: Third-party PC CPU unit share estimates via press reports; only quarters with a published figure are shown. Total client unit share was 30.3% in Q2 2026.

Exhibit 40: AMD leads or challenges in every segment; Arm CPUs and NVIDIA are the threats we price
SegmentMain competitorsAMD positionThreat to the thesis
Server CPUIntel Xeon, Arm CPUs (Graviton, Axion), NVIDIA VeraShare leader on revenue gains; 540+ public-cloud EPYC instance typesArm CPUs and Vera: Bear 3, −$41
AI acceleratorsNVIDIA (about 75%), Broadcom and Google custom ASICs, TrainiumSecond source with five lead customers on MI455XNVIDIA’s annual cadence and 75% margin
ClientIntel, Qualcomm Arm PCs, Apple30% unit share; Intel supply limits are a tailwindMemory costs cut PC units; FY27E client −8%
GamingConsole cycle, NVIDIA discrete graphicsConsole chips; 8% discrete add-in share in Q2 2026Late-cycle console demand: Q3’26 −40% y/y
EmbeddedAltera, Microchip, LatticeVersal; more than $18bn design wins expected in 2026Design wins are not booked revenue

Source: Company filings; third-party share data; Penmark Research. Threat values per share from the model scenarios.

Exhibit 41: NVIDIA earns a 75% gross margin; AMD earns 56%, and racks will pull it lower
NVIDIA, Q2 FY27AMD, Q3’26AMD, FY2575%56%52%

Source: Company results, non-GAAP gross margin. NVIDIA Q3 FY27 guide is 74.0% plus or minus 50bp; AMD Q4’26 guide is about 55%.

Why AMD wins

Customers want a second source and more performance per watt. Turin is in AWS, Google Cloud, Azure and Tencent; Helios gives AI labs an alternative to NVIDIA racks.

What does not last

Share gains in PCs depend on Intel’s supply. We carry Client at −8% in FY27E for that reason.

PART 2 · THE BUSINESS · 8 MANAGEMENT, GOVERNANCE AND CAPITAL ALLOCATION

Management beat its revenue guide in Q1 and Q3 of FY26, and buybacks have covered stock pay; warrants are the new dilution

Exhibit 42: The team is long-tenured and 2026 pay is tied to revenue, free cash flow and relative return
ExecutiveRoleRecord2026 salary / target equity
Lisa SuChair and CEOCEO since Oct 2014; Chair since Feb 2022$1.375m / $36m
Jean HuEVP, CFO and TreasurerJoined Jan 2023$0.85m / $9m
Mark PapermasterEVP and CTOJoined Oct 2011$0.90m / $10m
Forrest NorrodEVP, Data Center SolutionsRuns the largest segment$0.80m / $8m

Source: AMD proxy statement 2026 and Form 8-K; pay is target, not realised.

Exhibit 43: Buybacks of $8.6bn over five years exceeded $5.9bn of stock compensation
Buybacks ($m)Stock compensation ($m)Free cash flow ($m)$0$2,000$4,000$6,000$8,000FY21AFY22AFY23AFY24AFY25A

Source: Company 10-K filings; Penmark Research model. Buybacks as reported in the cash flow statement.

Exhibit 44: Revenue beat the guided range in Q1’26 and Q3’26: $13.4bn against $12.7–13.3bn
Q3’26 revenue ($m)Q1’26 revenue ($m)13,377 vs guide 12,700–13,30010,253 vs guide 9,500–10,100

Source: Company releases (Q4’25 on 3 Feb 2026, Q2’26 on 4 Aug 2026); reported results. Bar = reported revenue; grey band = guidance range.

Exhibit 45: Two executives sold $40.6m of stock in two months, all under 10b5-1 plans
Lisa Su (44,405 sh)Forrest Norrod (34,523 sh)$22.6m$18.0m

Source: SEC Form 4 filings, Sep–Oct 2026. No open-market purchases, Sep–Oct 2026. Lisa Su also received a 35,000-share grant on 10 Sep 2026.

Governance

Chair and CEO roles are combined. Nora Denzel is lead independent director; seven of eight board nominees are independent.

Pay and dilution

Pay follows per-share outcomes. Long-term awards use relative TSR to August 2029 and an FY28 EPS modifier. The 2026 plan asks for 65m more shares, and customer warrants add up to 320m.

Exhibit 46: AMD has $8.7bn of net cash, but $48.8bn of purchase commitments and warrants for up to 320m shares
$15.8bn26 Sep 2026Cash and investments$7.1bnincl. $4.75bn notes, Aug2026Debt$4.7bnFY25A: $6.7bnFY26E free cash flow$48.8bn$33.4bn due by FY27Purchase commitments1,636m1,229m in FY21FY25 diluted shares

Source: Q3’26 release and 10-Q (26 Sep 2026); FY25 10-K; Penmark Research model.

PART 3 · THE THESIS · 9 WHAT MOVES THE STOCK

AMD trades on guidance and the chip sector, not on the quarter: EPS beats moved the price the wrong way in 3 of 8 prints

Four drivers pass at least two of three tests. The stock is a high-beta AI chip proxy, and the reaction to the next guide matters more than the reported EPS.

Exhibit 47: Four drivers pass two tests or more, and each maps to a numbered debate
DriverWhat moves the stockTests passedValue in our modelDebateNext number
1. AI sector demandSOX explains 26% of variance; AMD fell 17.3% on 5 Nov against 13.9% implied by SOXEvent days; factor testPolicy bear case: −$74 at 20% odds4NVIDIA FQ3, 17 Nov
2. Data Center AI growthPrice tracks Data Center revenue, correlation 0.91 over eight quarters; the price implies +148%Price link; return split+$31 a share vs the market’s +137%1Q4’26, 2 Feb 2027
3. Forward guide, not the EPS beatReaction to EPS surprise: correlation −0.22; moves of 17% to 19% in Feb and May 2026Event days; return splitMultiple fell 12% as EPS rose 78%1, 2Q4’26, 2 Feb 2027
4. Gross margin and dilutionQ4 guide of 55% against 56%; warrants vest from 2027Event days; return split−$13 and −$7 a share2, 3FY26 10-K, Feb 2027

Source: Nasdaq closes and EPS surprises via Yahoo Finance; company filings; Penmark Research model. Tests: event days, return split, factor regression.

Exhibit 48: Share price reaction did not follow the EPS surprise: correlation −0.22, right sign in 5 of 8 prints
Next-day share price move (%)EPS surprise (%)−20%−10%0%10%20%Feb 25May 25Aug 25Nov 25Feb 26May 26Aug 26Nov 26

Source: Nasdaq closes and EPS surprises via Yahoo Finance, results days 4 Feb 2025 to 3 Nov 2026. Average absolute move 7.6%.

Exhibit 49: The stock rose 58% in a year on 78% EPS growth, so the multiple fell 12%
Share priceTrailing EPSP/E changeS&P 500−100%0%100%200%300%1 year3 years58%78%−12%8%260%159%39%69%

Source: Nasdaq closes via Yahoo Finance to 5 Nov 2026; trailing four-quarter non-GAAP EPS from company releases ($6.65 now; $3.73 a year ago; $2.57 three years ago). No dividend is paid.

Exhibit 50: Market and chip-sector moves explain 53% of weekly price variance; rates and AI peers add almost nothing
S&P 500SOX (chips)UnexplainedAI peersShare of weekly variance27%26%46%1%

Source: Penmark Research regression, 157 weekly returns, Nov 2023 to Nov 2026; S&P 500 alone, then SOX, NVIDIA-Broadcom-Marvell basket and 10-year yield added in turn. Total R-squared 0.54; the 10-year yield adds 0.0%.

What the tests say

The market sets the tone, the sector sets the size. The S&P 500 and SOX explain 53% of weekly moves; the other 46% is company news, mostly guidance.

AMD fell 17.3% on 5 November, 3.3pp more than its chip-sector beta implied (z-score −1.4, not extreme).

PART 3 · THE THESIS · 9 WHAT MOVES THE STOCK (CONTINUED)

The drivers we keep are worth +$31 to −$74 a share; rates, peers and EPS beats do not move the price

Exhibit 51: Value at stake per share ranges from +$31 on AI growth to −$74 if US policy curbs deployment
US policy limits 2027 GW(20%)Data Center AI +170% vs+137%Gross margin 54.5% vs55.6%Warrants vest (2GW) vsnone−$74+$31−$13−$7

Source: Penmark Research model, Market Check and Thesis tabs; first three are market view against ours, policy is the bear case at 20% probability. US$ per share.

Exhibit 52: Candidates tested and dropped
CandidateResultVerdict
10-year Treasury yieldR-squared 0.02% after market, sector and peersDropped
AI peer basket (NVIDIA, Broadcom, Marvell)0.8% beyond SOX, which already holds themDropped
EPS beat sizeCorrelation −0.22 with the next-day moveDropped
Consensus revisionsFY27E EPS up 17% in three months; the stock fell 36% in a monthDropped

Source: Penmark Research regression and event study; Yahoo Finance.

DRIVER 1 · DEBATE 4AI sector demand and policy decide the size of every move.

What the market does: SOX explains 26% of variance beyond the market, and AMD fell 17.3% on 5 Nov on data-center policy fears. Our model: a 2027 policy limit costs $74 a share at 20% probability. Next number: NVIDIA FQ3 on 17 Nov 2026; our estimate is not set, as NVIDIA is not covered.

DRIVER 2 · DEBATE 1Data Center AI growth is worth $31 a share more than the market allows.

What the market does: price implies +148% against our +170%. Our model: value $425 against $394 at the market’s +137%; the gap to the price is $20. Next number: Q4’26 on 2 Feb 2027; we model $16.96bn revenue and $2.91 EPS against consensus $16.31bn and $2.68.

DRIVER 3 · DEBATE 1 AND 2The guide, not the EPS beat, sets the reaction.

What the market does: a 15.5% EPS surprise in Feb 2026 met a 17.3% fall; a 5.8% surprise in May met a gain of 18.6%. Our model: the FY27 guide is already above 80% growth and we are at 83%. Next number: FY27 Data Center AI guide of “significantly more than double”; average absolute move on results is 7.6%.

DRIVER 4 · DEBATES 2 AND 3Margin and dilution are small but real: −$13 and −$7 a share.

What the market does: consensus implies 55.6% gross margin and about 1.68bn shares. Our model: 54.5% and 1.70bn average shares. Next number: Q4’26 gross margin guided about 55%; FY26 10-K on warrant vesting.

LimitsEight results and 157 weekly observations are small samples; R-squared figures describe the past three years, a period dominated by AI spending. Event-day moves include news released with the results.
PART 3 · THE THESIS · 10 THESIS: DATA CENTER AI

Data Center AI more than doubles in FY27E because lead customers have committed and Helios is shipping

What the market believes

Consensus has FY27E revenue of $90.8bn, below the 80% growth AMD guided on 3 November. The price implies +148% Data Center AI growth, or $92.3bn of revenue, against our +170% and $95.3bn.

The evidence

  • Reported data. Data Center revenue was $8.5bn in Q3’26, +97% y/y; the Q4 guide of +50% q/q implies $12.8bn.
  • Filings. OpenAI and Meta each hold a 1GW firm commitment for MI450 and warrants on up to 160m shares; our build ships 2.8GW in FY27E.
  • Call statements. AMD named Anthropic, Meta, Microsoft, OpenAI and Oracle as Helios lead customers and guided Data Center AI to “significantly more than double”.
  • Industry data. Amazon, Alphabet and Meta guide to $560bn of 2026 capex, and 2027 is expected higher.
Exhibit 53: Data Center revenue has risen for five straight quarters to $8.5bn, +97% y/y
$0$2,000$4,000$6,000$8,000$10,000Q4'24Q1'25Q2'25Q3'25Q4'25Q1'26Q2'26Q3'26$3,859$3,674$3,240$4,341$5,380$5,775$6,718$8,546

Source: Company filings and Q3’26 release; segment revenue, US$m. Q4’26 guide is +50% q/q, or about $12.8bn.

Exhibit 54: Penmark revenue against consensus, with the gap in numbers
US$mFY26EFY27EFY28E
Revenue, Penmark$52,130$95,336$121,319
Revenue, consensus$51,476$90,830$126,790
Difference+1.3%+5.0%−4.3%

What it is worth

Each 10pp of Data Center AI growth is worth about $9 a share. Our +170% gives a value of $425; the market’s +137% gives $394. This is the largest single gap to consensus and drives 53% of FY27E revenue growth.

Signposts

  • 2 Feb 2027. Q4’26: Data Center near $12.8bn confirms; below +45% q/q disproves.
  • Feb 2027. FY27 AI guide at more than 2x supports; at or below 2x breaks the claim.
  • Mid-2027. Q1 and Q2’27 GW shipped against 2.8GW for FY27E.
Exhibit 55: Our FY27E Data Center AI growth of +170% sits between the market-implied +148% and the bull case
BearPrice-impliedBase (ours)Bull+90%+148%+170%+220%

Source: Penmark Research model, Scenarios and Market Check tabs. Price-implied growth holds the target P/E and other drivers at base.

What would prove us wrongQ4’26 Data Center growth below 45% q/q on 2 February 2027, or a FY27 AI guide of 2x or less. Either would cut our FY27E AI growth toward the bear case’s +90%, which is worth $276 a share.
PART 3 · THE THESIS · 11 THESIS: MARGINS

Operating leverage lifts the operating margin to 36% even though rack-scale Helios dilutes gross margin

What the market believes

Consensus implies a 55.6% FY27E gross margin and about 1.68bn shares. The Q4 guide of 55% is a point below Q3’s 56%, and the CFO did not guide FY27 margin.

The evidence

  • Reported data. Non-GAAP gross margin was 55.9% in Q3’26 and 56.2% in Q2’26 against 52% in FY25; operating margin was 29.3% in Q3’26.
  • Call statements. The CFO expects “very significant operating margin expansion in 2027”; Q4 opex is guided at about $3.75bn, up 5% q/q on revenue up 27%.
  • Long-term targets. The November 2025 analyst day set FY28 gross margin at 55% to 58% and operating margin above 35%; we model 55.0% and 37.2%.
  • Peer data. NVIDIA earns a 75% gross margin on rack-scale systems, so the ceiling is higher than AMD shows today.
Exhibit 56: Non-GAAP gross margin has held at 55% to 56% for three quarters; the Q4 guide is 55%
0204060Q4'24Q1'25Q2'25Q3'25Q4'25Q1'26Q2'26Q3'2654.153.743.354.057.055.456.255.9

Source: Company releases, non-GAAP gross profit over revenue; Q2’25 includes an $800m MI308 inventory charge; Q4’26 guide is about 55%.

Exhibit 57: Our margin and share count differ from consensus in the direction of caution
FY27EPenmarkConsensusDiff.
Revenue ($m)95,33690,830+5.0%
Gross margin54.5%55.6%−1.1pp
Average diluted shares (m)1,7001,680+1.2%
Non-GAAP EPS$17.43$16.27+7.2%

What it is worth

Margin is a $13 drag, leverage is the offset. At the consensus margin our value would be $438, $13 above base. Revenue above consensus adds $1.18 to FY27E EPS, margin takes off $0.54 and shares $0.25; opex and other lines add the remaining $0.77 of the $1.16 gap.

Signposts

  • 2 Feb 2027. Q4’26 margin of 56% or above with racks shipping would confirm the upside.
  • Q1’27. First full quarter of Helios mix: below 53% for two quarters breaks the claim.
  • FY26 10-K. Opex growth against the 30% we model.
Exhibit 58: Our 54.5% FY27E gross margin is 1.1pp below consensus; each point is worth about $12 a share
Bear 53.0Bull 55.5Base (ours) 54.5Consensus-implied 55.6

Source: Penmark Research model, Scenarios tab; consensus-implied margin per Market Check tab. Per-share value from the sensitivity grid.

What would prove us wrongGross margin below 53% for two quarters, which is our bear case and costs $18 a share, or opex growth above 35% against revenue.
PART 3 · THE THESIS · 12 THESIS: DILUTION

Customer warrants dilute EPS by a visible 3% a year, and every vested share is paid for by shipments

What the market believes

Most models use about 1.68bn shares and no warrant dilution. We use 1,700m average shares in FY27E and 1,761m in FY28E because vesting follows GW shipped.

The evidence

  • Filings. OpenAI and Meta can each take up to 160m shares at $0.01; vesting is tied to GW shipped, up to 6GW each.
  • Reported data. Q3’26 GAAP net income took a $1.1bn warrant charge, excluded from non-GAAP EPS.
  • Model test. Implied shipments of 2.8GW cover the 2GW vested in FY27E, and cover holds every year.
  • Proxy. The 2026 plan seeks 65m extra employee shares on top of the warrants.
Exhibit 59: Customer warrants add 160m shares by FY29E, 9% on top of 1,694m before them
Shares before warrants (m)Warrant shares vested (m)05001,0001,5002,000FY26EFY27EFY28EFY29E1,6701,7301,7921,854

Source: Penmark Research model, Statements tab; year-end diluted shares. OpenAI and Meta warrants at $0.01, up to 160m shares each (8-Ks, 6 Oct 2025 and 24 Feb 2026); model vests 2GW a year to a modelled 160m total by FY29E.

Exhibit 60: The warrant schedule is covered by implied shipments in every year
Warrant scheduleFY27EFY28EFY29E
GW vested, cumulative2.04.06.0
GW shipped, cumulative2.86.811.7
Warrant shares vested (m)53107160
Cover: shipped over vested1.4x1.7x1.9x

What it is worth

Dilution costs $7 a share, a 1.6% haircut to value. It is the smallest of the three debates but the surest: the warrants are contractual and vest as the revenue we forecast arrives.

Signposts

  • Feb 2027. FY26 10-K: first vesting tranche conditions and share count.
  • Q1’27. Q1 diluted share count; our 1,700m FY27E average implies about 1,690m.
  • 2027. Any change in the 6GW tranche terms.
Exhibit 61: Average diluted shares rise from 1,636m in FY25A to 1,761m in FY28E, after three years of near-flat counts
05001,0001,5002,000FY21AFY22AFY23AFY24AFY25AFY26EFY27EFY28EActualEstimates →1,2291,5711,6251,6371,6361,6601,7001,761

Source: Company 10-K filings (A); Penmark Research model (E). FY22 includes the Xilinx acquisition shares.

What would prove us wrongVesting that runs ahead of shipments, or a share count above 1.72bn in FY27E, would cut FY27E EPS by more than $0.25 and move value by over $7 a share.
PART 4 · THE NUMBERS · 13 TRACK RECORD AND QUALITY OF REPORTED NUMBERS

Revenue has grown 20% a year since FY21A, but cash conversion swings with working capital and non-GAAP excludes $5bn of real costs

Exhibit 62: Revenue grew at 20% a year over four years while free cash flow ranged between $1.1bn and $6.7bn
Revenue ($m)Free cash flow ($m)$0$10,000$20,000$30,000$40,000FY21AFY22AFY23AFY24AFY25A$16,434$3,220$23,601$3,115$22,680$1,121$25,785$2,405$34,639$6,735

Source: Company 10-K filings. FY22 includes the Xilinx acquisition (completed February 2022).

Exhibit 63: Non-GAAP EPS runs 57% above GAAP in FY25A; the gap narrows to 33% in FY26E and 17% in FY27E
Non-GAAP EPS ($)GAAP EPS ($)$0$5$10$15$20$25FY25AFY26EFY27EFY28EActualEstimates →$4.17$2.65$8.04$6.04$17.43$14.93$22.12$19.69

Source: Company filings (A); Penmark Research model (E). GAAP EPS includes amortisation of acquired intangibles and warrant fair-value changes, warrant fair-value changes are not forecast.

Exhibit 64: FCF conversion falls from 99% in FY25A to 35% in FY26E on $8.2bn of working capital
0255075100FY24AFY25AFY26EFY27EFY28EActualEstimates →8399%35%

Source: Company filings (A); Penmark Research model (E). Free cash flow over non-GAAP net income, %.

Exhibit 65: Five adjustments separate non-GAAP from GAAP; SBC is a cost in the DCF and warrants sit in the share count
ItemTreatment and size
Amortisation of acquired intangibles$2.2bn in FY26E excluded from non-GAAP; non-cash, but it is the cost of buying growth.
Stock-based compensation$1.6bn in FY25A (4.7% of revenue); excluded from non-GAAP EPS, treated as a cash cost in our DCF.
Warrant fair value$1.1bn charge in Q3’26 excluded from non-GAAP; the dilution is in our share count.
Inventory charge$800m MI308 charge in Q2’25 cut that quarter’s non-GAAP margin to 43%.
TaxFY25 GAAP tax was a $103m benefit on reserve releases; non-GAAP rate is 13% to FY27E and 15% from FY28E.

Source: Company filings and releases; Penmark Research model.

PART 4 · THE NUMBERS · 14 FORECASTS

Our FY27E revenue is 5% above consensus and 1.5% above the guided floor, and EPS is 7% above on two drivers

Exhibit 66: Server CPU units and Data Center AI gigawatts explain all of the $43bn FY27E revenue growth
Server CPU$40.6bn, +100%Guide: about 2x. Venice, agenticCPU demandData Center AI$36.6bn, +170%2.8GW shipped x about $13bnper GWData Center$77.2bn, +128%Client$10.3bn, −8%PC units down double digits;memory costsGaming$2.6bn, −10%Console cycleClient and Gaming$12.9bn, −8.4%Embedded$5.3bn, +25%Design wins above $18bn in 2026Revenue FY27E$95.3bn, +82.9%

Source: Company filings; Penmark Research model, FY27E. Highlighted boxes are the two forecast drivers. Server CPU and Data Center AI split is a model estimate.

Exhibit 67: We are above consensus in FY26E and FY27E and below it in FY28E; each gap has a stated reason
LinePenmarkConsensusGuidanceGapReason
Revenue FY26E ($m)52,13051,47651,656 to 52,676+1.3%Q4 built from segment guidance; Data Center +50% q/q
Revenue FY27E ($m)95,33690,830above 93,899+5.0%Guide is above 80% growth; consensus has not caught up
Revenue FY28E ($m)121,319126,790CAGR above 35%−4.3%Server and AI growth slow to 25% and 40%; consensus is higher
Gross margin FY27E54.5%55.6%none−1.1ppRack-scale mix; no FY27 margin guide
Gross margin FY28E55.0%–55% to 58%–Low end of the analyst-day range
EPS FY26E$8.04$7.80none+3.1%Revenue beat on Q4 guide midpoint
EPS FY27E$17.43$16.27none+7.2%Revenue +$1.18, margin −$0.54, shares −$0.25, opex +$0.77
EPS FY28E$22.12$23.00above $20−3.8%Warrant shares reach 107m; consensus revenue higher

Source: Company guidance (3 Nov 2026) and analyst-day targets (11 Nov 2025); consensus per InvestingPro and S&P Global, 5 Nov 2026; Penmark Research model. Consensus gross margin is implied, not published.

Exhibit 68: Q4’26 revenue of $16.96bn is inside the $16.49–17.51bn guide and 4% above consensus
Revenue ($m)Non-GAAP gross margin(%)Non-GAAP EPS ($)16,964 vs guide 16,490–17,51055.0 vs guide 54.5–55.5$2.91

Source: Company guidance (3 Nov 2026); consensus per InvestingPro, 5 Nov 2026; Penmark Research model. EPS is not guided. Bar = Penmark Q4’26 estimate; grey band = guidance range; amber tick = consensus.

Guidance recordRevenue landed above the guided range in Q1’26 ($10.25bn against $9.5-10.1bn) and Q3’26 ($13.38bn against $12.7-13.3bn), and gross margin landed inside the range both times. Our Q4 sits near the guide midpoint.
PART 4 · THE NUMBERS · 14 FORECASTS (CONTINUED)

Value moves most with the target multiple and least with margin; free cash flow lags profit until FY27E as inventory builds

Exhibit 69: Free cash flow is $4.7bn in FY26E, then $19.6bn and $32.2bn as working capital eases
US$mFY25AFY26EFY27EFY28E
Operating cash flow7,7097,19523,85937,036
Capital expenditure(974)(2,545)(4,290)(4,853)
Free cash flow6,7354,65019,56832,183
Change in working capital–(8,222)(7,597)(4,687)
FCF margin19.4%8.9%20.5%26.5%
Capex % of revenue2.8%4.9%4.5%4.0%
Inventories ($m)7,92014,79820,93123,623
Receivables ($m)6,31512,52418,40122,446

Source: Company filings (A); Penmark Research model (E).

Costs and margins

Cost of revenue varies with units; opex is mostly fixed. FY27E non-GAAP opex is $18.0bn, up 30% on revenue up 83%, so operating margin rises to 35.6%. Gross margin falls to 54.5% on rack mix.

Cash and balance sheet

Working capital absorbs $8.2bn in FY26E. Inventory reaches $14.8bn and receivables $12.5bn. Cash builds to $35.3bn including investments by FY27E and debt stays at $7.1bn.

Exhibit 70: Target P/E moves value by $209 across its range; FY27E EPS and AI growth move it by about $85 and $73
Target P/E 18.4x to 30.4xFY27E EPS $15.69 to$19.18DC AI growth +130% to+210%Gross margin 52.5% to56.5%Base $425−$105+$105−$42+$42−$37+$37−$24+$24

Source: Penmark Research model, Scenarios tab sensitivity grids; value per share at the price date, one input moved with others at base.

Exhibit 71: Each 20pp of Data Center AI growth is worth about $18; each 1pt of gross margin about $12
$366$377$388$399$411$383$395$407$418$430$401$413$425$437$449$419$431$443$455$468$437$449$462$474$487FY27E gross margin52.5%53.5%54.5%55.5%56.5%FY27E AI growth+130%+150%+170%+190%+210%

Source: Penmark Research model, Scenarios tab; value per share at the price date. Centre is the base case, $425.

What it means for valueEPS sensitivity: $0.87 of FY27E EPS is worth about $21 a share at 24.4x. A one-turn move in the target multiple is worth about $17. The multiple, which comes from peer PEG, is the largest single input and is the reason the rating changes at the lowest AI peer PEG.
Exhibit 72: Operating margin rises from 22% to 37% by FY28E while gross margin stays near 55%
Gross margin (%)Operating margin (%)FCF margin (%)0102030405060FY24AFY25AFY26EFY27EFY28EActualEstimates →553726

Source: Company filings (A); Penmark Research model (E). Non-GAAP gross and operating margin.

Exhibit 73: Cash and investments build from $11bn to $68bn by FY28E on free cash flow, against $7.1bn of debt
Cash and investments ($bn)Debt ($bn)$0bn$20bn$40bn$60bn$80bnFY25AFY26EFY27EFY28EActualEstimates →$11bn$3bn$16bn$7bn$35bn$7bn$68bn$7bn

Source: Company filings (A); Penmark Research model (E). The model assumes no buybacks or acquisition cash outflow, so the cash build is a ceiling.

PART 4 · THE NUMBERS · 15 VALUATION

Forward P/E on a peer PEG gives a $425 base value and a $457 target; the DCF at a 13% discount rate is the cross-check

Method

AMD is profitable and earnings are ramping, so we value it on FY27E P/E. The multiple is 24.4x: the median PEG of NVIDIA, Broadcom and Marvell (0.22x) times AMD’s consensus EPS growth of 109%. EV/EBITDA adds nothing for a net-cash company, and sum of the parts does not fit one fabless model.

Rating

Total return of 13.1% is Buy, with low conviction. One of eight robustness tests flips the rating: the lowest AI peer PEG (0.16x) gives $329 and −19%.

Methods compared

The P/E methods span $276 to $510 and the DCF sits far below at $188 to $277. The football field is Exhibit 21 on page 5. The gap is the discount rate: the P/E multiple embeds the market’s pricing of peers, while the DCF uses a 13.0% cost of capital.

Street

The Street range is $430 to $1,250, with a mean of $649. Our $457 target is $27 above the lowest Street target.

Exhibit 74: Probability-weighted value of $405 plus a $53 one-year roll gives the $457 target
$0$100$200$300$400$500Weighted valueRoll at 13.0% cost ofequity12-month target$405+$53$457

Source: Penmark Research model, Valuation tab. No dividends. Total return to target is 13.1% against a 13.0% cost of equity.

Exhibit 75: At the target, FY27E P/E is 26.2x; at the price it is 23.2x
MultipleAt $404.40At target $457.28
P/E, FY27E23.2x26.2x
P/E, FY28E18.3x20.7x
Target forward P/E24.4x24.4x before roll
Peer median PEG0.22x0.22x
AMD EPS growth used109%109%

Source: Penmark Research model, Valuation tab. Multiples at the actual price are used elsewhere in this report.

Reconciliation to page 1Target $457.28 = weighted value $404.51 rolled one year at 13.05%. Page 1 shows the same target, a 13.1% price return and, with no dividend, the same total return.
Exhibit 76: AMD trades at 23-25x FY2 earnings against 15x for AI peers, but with 38-46pp more EPS growth
CompanyPrice, 5 NovP/E, FY2EPS growthPEGIn target set
NVIDIA$181.6011.4x70.9%0.16xYes
Broadcom$291.5315.0x66.8%0.22xYes
Marvell$133.5518.2x73.7%0.25xYes
Intel$52.2535.8x39.0%0.92xNo
Qualcomm$156.7515.4x−2.7%n/mNo
AI compute median15.0x70.9%0.22x
AMD, consensus$404.4024.9x108.6%0.23x
AMD, our EPS$404.4023.2x116.7%0.20x

Source: Company filings, consensus (7-30 Oct 2026), Nasdaq closes 5 Nov 2026; Penmark Research. FY2 is the second forecast fiscal year; fiscal year ends differ. Peers are not covered by Penmark.

PART 4 · THE NUMBERS · 15 VALUATION (CONTINUED)

The DCF gives $223 because it uses a 13.0% discount rate; at the 8.7% the price implies it gives $404

Exhibit 77: The DCF values the equity at $438bn on SBC as a cash cost and 1,990m shares
DCF, US$mValue
PV of FY26E stub to FY36E cash flows245,767
PV of terminal value183,475
Enterprise value429,241
Terminal value share of EV42.7%
Plus cash and investments15,776
Less debt(7,079)
Equity value437,938
Diluted shares incl. warrants (m)1,990
Value per share at price date$223.05

Source: Penmark Research model, Valuation tab. Equity value of $438bn over 1,990m shares is $220.07 at 26 Sep 2026; rolled forward 0.11 years at the cost of equity to the 5 Nov price date it is $223.05. Terminal growth 3%; ten full years to FY36E.

Exhibit 78: The price of $404 needs an 8.7% discount rate; our 13.0% WACC gives $223
$0$200$400$600$8006.0%7.0%8.0%9.0%10.0%11.0%12.0%13.0%14.0%15.0%$18313.0%: $223Price $404

Source: Penmark Research model, reverse DCF on the Valuation tab; model cash flows and 3% terminal growth held. WACC tested from 6.0% to 15.0%.

Exhibit 79: DCF value stays between $188 and $277 across WACC of 11.4% to 14.5% and growth of 2% to 4%
$259$263$267$272$277$237$240$243$247$251$219$221$223$226$228$202$204$206$208$210$188$190$191$192$194Terminal growth2.0%2.5%3.0%3.5%4.0%WACC11.4%12.2%13.0%13.7%14.5%

Source: Penmark Research model, Scenarios tab; value per share at the price date. Centre $223.

Exhibit 80: Bull $510, base $425 and bear $276 weight to $405; the price sits just below the weighted value
$0$200$400$600$276$425$510Bear25%Base55%Bull20%Weighted $405Price $404

Source: Penmark Research model, Scenarios tab, values at the price date before the one-year roll. Probabilities are analyst judgements: bear 25% reflects policy risk and an unguided FY27 margin.

Why the two methods differ by 48%

The gap is the discount rate, not the cash flows. The DCF equals the P/E value at a WACC of 8.4%, or a beta of 0.84, against the 1.93 we use from peers (AMD’s own 5-year beta is 2.3).

What the price impliesAt the price, the market assumes FY27E Data Center AI growth of +148%, or revenue of $92.3bn and EPS of $16.59, at our 24.4x multiple. That is 22pp below our +170%.
Exhibit 81: The three cases differ most on Data Center AI growth, which ties to debate 1 and drivers 1 and 2
CaseProbabilityFY27E revenueFY27E EPSValueStory and key assumptions
Bear25%$76.7bn$11.31$276Policy limits US build; rack margin 53.0%; Arm CPUs take sockets
Base55%$95.3bn$17.43$425FY27 framework delivered; warrants vest with GW
Bull20%$105.1bn$20.91$510Helios ramps faster at five lead customers; Venice beats guide

Source: Penmark Research model, Scenarios tab. Probability-weighted value is $404.51; probabilities are analyst judgements and do not carry statistical confidence.

PART 4 · THE NUMBERS · 16 CATALYSTS

The February 2027 results and 10-K decide three of four debates, and we are 4% above consensus on revenue into them

Exhibit 82: Seven dated events in the next 12 months; the February 2027 print and 10-K carry three of the four debates
17 Nov 2026NVIDIA FQ3AI demand read-across31 Dec 2026Deals closeWorld Labs, TaalasDebates 1, 2AMD Q4’262 Feb: $16.96bn, $2.91Debate 3FY26 10-K26 Feb (est.): warrants4 May 2027Q1’27Helios volume (est.)3 Aug 2027Q2’27MI450 ramp (est.)2 Nov 2027Q3’27FY28 framework (est.)

Source: Company calendar; Yahoo Finance earnings dates (2 Feb 2027); the 10-K and later quarterly dates are Penmark estimates of the usual cadence.

Exhibit 83: Four catalysts with our expectation against consensus and the value at stake
DateEventPenmark against consensusValue at stakeRating effect
17 Nov 2026NVIDIA FQ3 resultsRead-across on AI demand and rack supply; not covered by PenmarkSector tone; ±5% on the day is typicalNeutral
2 Feb 2027Q4’26 results: we expect $16.96bn revenue and $2.91 EPS; consensus $16.31bn and $2.68Revenue 4.0% and EPS 8.6% above consensus; guide midpoint $17.0bnBull 1 (+$46) or Bear 2 (−$18)Supports
Feb 2027FY26 10-K and FY27 guide: Data Center AI, margin, first warrant vestingWe model AI +170%, margin 54.5%, 2GW vestedDebates 1 to 3: +$31, −$13, −$7Could change rating
2027State and federal data-center policy after the 3 Nov midtermsWe carry a 20% chance of limits on 2027 GWBear 1: −$74 a shareCould change rating

Source: Company releases and calendar; consensus per InvestingPro, 5 Nov 2026; Penmark Research model.

Exhibit 84: Helios ramp is the largest catalyst at +$46 a share and 60% likely
Bull 1: Helios ramp (60%)Bull 2: server CPU (70%)Bull 4: opex leverage(40%)Bull 3: PC share (30%)Bull 5: Embedded (35%)+$46+$14+$5+$4+$3

Source: Penmark Research model, Thesis tab; US$ per share if the catalyst lands, before probability weights shown.

Events that could change the rating

A Q4 guide below 45% Data Center growth or a policy limit would end the Buy. The rating already flips at the lowest peer PEG, so conviction is low. A confirmed 56% margin and an AI guide above 2x would lift the case towards the bull value of $510.

Exhibit 85: Our Q4’26 revenue and EPS sit 4% and 9% above consensus, with revenue inside the guided range
Q4’26 revenue ($m)Q4’26 non-GAAP EPS ($)FY27E revenue ($bn)16,964 vs guide 16,490–17,510$2.9195.3

Source: Company guidance (3 Nov 2026), consensus per InvestingPro (5 Nov 2026), Penmark Research model. FY27E guide is a floor (growth above 80%). Bar = Penmark estimate; grey band = guidance; amber tick = consensus.

Next number to watch

DriverWhenOur estimate
Data Center AI growthQ4’26, 2 Feb 2027Data Center $12.8bn, +50% q/q
Gross marginQ4’26, 2 Feb 202755.0% (guide about 55%)
Warrant dilutionFY26 10-K, Feb 20270 vested at FY26; 2GW by FY27E
Policy2027 sessionsNo limit in base case
PART 4 · THE NUMBERS · 17 RISKS

Five downside risks cost $7 to $74 a share; the thesis ends if Data Center AI growth falls below the +148% in the price

Exhibit 86: Risks to the rating and target, by likelihood and value per share (Buy: downside risks)
RiskWhat happensLikelihoodValueSignpost to watch
Bear 1: US data-center policyPermit pauses and a federal moratorium slow 2027 GW deployedLow (20%)−$74Federal bill S.4214 moves; statewide order in a top-three AI market
Bear 3: Arm CPUs and NVIDIA VeraServer CPU growth is +70% not +100%Medium (30%)−$41Venice share data; hyperscaler Arm launches
Bear 2: Rack mix dilutes marginGross margin 53.0% not 54.5%Medium (35%)−$18Q4’26 margin below 55%; FY27 margin commentary
Bear 4: PC decline deepensClient −18% not −8% on memory costsMedium (40%)−$8Intel supply; memory prices; 1H27 sell-through
Bear 5: Warrants vest as modelled (market assumes none)2GW vested in FY27E: 53m sharesHigh (70%)−$7FY26 10-K vesting conditions

Source: Penmark Research model, Thesis tab; value per share if the risk occurs, before probability weights.

Exhibit 87: Policy is the only risk above $40 a share with a sub-30% chance; warrants are likely but small
$0$20$40$60$80203040506070Value lost per share ($)Likelihood (%)PolicyArm / VeraRack marginPC declineWarrants

Source: Penmark Research model, Thesis tab catalyst values and probabilities.

What would end the thesis

Data Center AI growth below +148% for FY27E. That is the growth the price already implies; below it our value falls under the price. The rating also falls to Sell if the target PEG drops to the lowest peer level of 0.16x ($329, −19%).

Regulatory, legal, key-person

Policy and export rules matter most. Lisa Su is Chair and CEO; no succession plan is disclosed in AMD’s filings. Insider sales of $40.6m were under 10b5-1 plans.

Probability limitsThe 20% to 70% likelihoods are analyst judgements, not estimates from a sample. Risks are not independent: a policy limit would also slow Helios volume and raise the odds of Bear 2.
Exhibit 88: Weighted for likelihood, policy and the Arm and Vera threat cost $15 and $12 a share; warrants cost $5
US data-center policy (20% x $74)Arm CPUs and NVIDIA Vera (30% x $41)Rack mix dilutes margin (35% x $18)Warrants vest as modelled (70% x $7)PC decline deepens (40% x $8)$14.7$12.4$6.2$4.7$3.0

Source: Penmark Research model, Thesis tab: value per share times analyst-judged probability. Risks overlap, so the sum ($41) is not additive to the target.

PART 4 · THE NUMBERS · 18 OWNERSHIP AND TRADING

Index funds and large managers own 75% of AMD, shorts are 2.3% of float, and the Street is 83% Buy

Exhibit 89: Institutions hold 75% of the shares and short interest is 2.3% of float; the stock trades $11.6bn a day
74.9%5,205 holdersHeld by institutions0.4%Yahoo FinanceHeld by insiders2.3%38.1m shares; 1.7 days tocoverShort interest, % of float$11.6bn3 months, 22.0m sharesAverage daily value traded$658.526 Oct 2026; low $188.2252-week high

Source: Yahoo Finance holders and short-interest data (13F as of 30 Jun 2026); Nasdaq price history to 5 Nov 2026.

Exhibit 90: Ten holders own 36% of the stock; JPMorgan and Fidelity added most in the quarter
Holder% heldShares (m)Change
BlackRock9.2%150.3+3.3%
Vanguard Capital Mgmt6.5%106.5+0.5%
State Street4.5%74.4−0.6%
Invesco3.2%53.2+5.4%
JPMorgan Chase2.5%40.3+66.2%
Geode Capital2.4%39.6−0.6%
Vanguard Portfolio Mgmt2.3%37.9+0.9%
T. Rowe Price2.2%35.9+24.6%
FMR (Fidelity)1.6%25.6+67.2%
Morgan Stanley1.5%24.3+2.1%

Source: Form 13F filings via Yahoo Finance, quarter to 30 Jun 2026. Change is in shares held versus the prior filing.

Exhibit 91: 29 of 35 analysts rate AMD Buy, none Sell, and the mean target is $649
BuyHoldStreet ratings296

Source: TipRanks, 4 Nov 2026 (ratings in the last three months); no Sell ratings.

Exhibit 92: Our $457 target is close to the lowest Street target of $430; the Street mean is $649
Price 5 Nov $404Penmark target $457Street high $1250Street low $430Street mean $649

Source: stockanalysis.com, 4 Nov 2026; Penmark Research model.

Exhibit 93: AMD fell 39% from its $659 peak on 6 October to $404 in one month, and the chip index fell 32%
AMD ($)SOX, rebased to AMD$200$300$400$500$600$700Jun-26Jul-26Aug-26Sep-26Oct-26Nov-26$404$292

Source: Nasdaq closes via Yahoo Finance, 5 May to 5 Nov 2026; SOX rebased to the AMD price on 5 May 2026. Peak $658.52 on 6 Oct 2026; 52-week low $188.22 on 3 Mar 2026.

Overhangs

Share supply is rising from three sources. Warrants for up to 320m shares to OpenAI and Meta; about $8.2bn of World Labs consideration paid in stock; and a request for 65m more employee shares. We model 1,761m average shares in FY28E.

Insiders and liquidity

Executives sold $40.6m under 10b5-1 plans and made no open-market purchases. Planned sales are not a signal on their own. Daily value traded of $11.6bn means position size is not a constraint.

PART 5 · APPENDIX · 19 FINANCIAL MODEL

Revenue rises 3.5 times from $34.6bn to $121.3bn by FY28E as non-GAAP EPS rises from $4.17 to $22.12

Exhibit 94: Income statement, non-GAAP, US$m
Income statementFY24AFY25AFY26EFY27EFY28E
Revenue25,78534,63952,13095,336121,319
Growth y/y13.7%34.3%50.5%82.9%27.3%
Cost of revenue(12,026)(16,474)(23,144)(43,378)(54,594)
Gross profit13,75918,16528,98651,95866,726
Gross margin53.4%52.4%55.6%54.5%55.0%
Operating expenses(7,621)(10,397)(13,849)(18,004)(21,604)
Operating income6,1387,76815,13733,95445,121
Operating margin23.8%22.4%29.0%35.6%37.2%
Net income5,4206,83113,35429,64038,950
Diluted shares (m)1,6371,6361,6601,7001,761
EPS ($)3.314.178.0417.4322.12
EPS growth–26.0%92.9%116.7%26.9%
GAAP EPS ($)1.002.656.0414.9319.69
Exhibit 95: Cash flow, US$m
Cash flowFY24AFY25AFY26EFY27EFY28E
Operating cash flow3,0417,7097,19523,85937,036
Capital expenditure(636)(974)(2,545)(4,290)(4,853)
Free cash flow2,4056,7354,65019,56832,183
FCF margin9.3%19.4%8.9%20.5%26.5%
Stock-based compensation1,4071,6382,0582,8603,276
Share repurchases(862)(1,316)–––
Change in working capital––(8,222)(7,597)(4,687)
Exhibit 96: Balance sheet, US$m
Balance sheetFY24AFY25AFY26EFY27EFY28E
Cash and investments5,13210,55215,74835,31767,500
Receivables6,1926,31512,52418,40122,446
Inventories5,7347,92014,79820,93123,623
Goodwill and intangibles43,76941,83140,04238,00236,248
Total assets69,22676,926102,373136,926178,236
Debt1,7213,2227,0797,0797,079
Shareholders' equity57,56862,99972,858101,095139,045
Net cash3,4117,3308,66928,23860,421
Book value per share ($)35.1738.5143.8859.4678.95
PART 5 · APPENDIX · 19 FINANCIAL MODEL (CONTINUED)

Quarterly revenue rises from $7.7bn to a modelled $17.0bn in Q4’26 as Data Center grows 3.3 times

Exhibit 97: Last eight quarters and next quarter, non-GAAP, US$m
US$mQ4’24AQ1’25AQ2’25AQ3’25AQ4’25AQ1’26AQ2’26AQ3’26AQ4’26E
Revenue7,6587,4387,6859,24610,27010,25311,53613,37716,964
Data Center3,8593,6743,2404,3415,3805,7756,7188,54612,819
Client and Gaming2,8762,9413,6214,0483,9403,6053,8413,7222,903
Embedded9238238248579508739771,1091,242
Gross profit4,1403,9923,3264,9925,8555,6856,4887,4839,330
Operating income2,0261,7798972,2382,8542,5403,0943,9235,580
Net income1,7771,5667811,9652,5192,2652,7603,4714,858
EPS ($)1.090.960.481.201.531.371.662.092.91
Gross margin54.1%53.7%43.3%54.0%57.0%55.4%56.2%55.9%55.0%
Operating margin26.5%23.9%11.7%24.2%27.8%24.8%26.8%29.3%32.9%

Source: Company releases (A); Penmark Research model (E). Q4’26E is built from the 3 Nov 2026 guidance; later quarters are not modelled quarterly.

Exhibit 98: Segment revenue, US$m: Data Center is 81% of FY27E revenue
SegmentFY24AFY25AFY26EFY27EFY28E
Server CPU (model est.)––20,31540,63050,787
Data Center AI (model est.)––13,54336,56751,193
Data Center12,57916,63533,85877,196101,980
Client7,05410,64011,19610,30110,713
Gaming2,5953,9102,8752,5872,587
Embedded3,5573,4544,2015,2516,039
Total revenue25,78534,63952,13095,336121,319

Source: Company filings (A); Penmark Research model (E). Server CPU and Data Center AI are model estimates; FY24A to FY25A Client and Gaming as recast.

Exhibit 99: Returns and multiples at the price of $404.40; ROE rises from 7% to 29% by FY27E
RatiosFY24AFY25AFY26EFY27EFY28E
ROE2.9%7.2%14.8%29.2%28.9%
ROA2.4%5.9%11.2%21.2%22.0%
ROCE10.0%12.0%19.9%34.3%34.2%
P/E at $404.40122.2x97.0x50.3x23.2x18.3x
EV/operating income108.6x85.8x44.0x19.6x14.8x
P/B11.5x10.5x9.2x6.8x5.1x
FCF yield0.4%1.0%0.7%2.9%4.8%
P/E at target $457.28138.1x109.7x56.8x26.2x20.7x
Net cash / equity5.9%11.6%11.9%27.9%43.5%

Source: Company filings (A); Penmark Research model (E). EV uses net cash of $8.7bn at 26 Sep 2026; no dividend is paid, so there is no dividend yield.

PART 5 · APPENDIX · 20 REFERENCE

Reference: one primary method, one cross-check, a 5 November 2026 data cut-off and a stated basis of presentation

Exhibit 100: Methodology and basis of presentation
Adjusted to reportedNon-GAAP EPS $4.17 in FY25A against GAAP $2.65: the $1.52 gap is amortisation of acquired intangibles, stock-based compensation, restructuring and tax effects.
Business modelProducts and units, fabless semiconductors: revenue is units times price times mix by segment. Data Center is split into server CPU and Data Center AI by the model, not by AMD.
Primary valuationForward P/E on FY27E non-GAAP EPS at a peer-PEG multiple of 24.4x. Target is probability-weighted value rolled forward one year at the 13.0% cost of equity, no dividends.
Cross-checkDCF: Q4’26 stub and FY27E to FY36E, mid-period discounting from 26 Sep 2026, SBC treated as a cash cost, 3% terminal growth, 1,990m shares including warrants.
Rating bandsBuy above +10%, Hold between −10% and +10%, Sell below −10% expected total return. High conviction needs a return above 20% and a rating that survives every robustness test.
Fiscal yearEnds the last Saturday of December: FY26 is 28 Dec 2025 to 26 Dec 2026. FY26E is nine months reported plus a Q4 built from guidance.
Basis of presentationPrice, consensus and peer data are at the 5 Nov 2026 close; the model date is 6 Nov 2026. Non-GAAP operating income is used where the template shows EBITDA; adjusted EBITDA is $16.1bn in FY26E and $35.1bn in FY27E. Consensus EBITDA is not published. AMD’s own forward P/E history is not shown; Exhibit 15 uses trailing P/E.
Exhibit 101: Glossary and ratio definitions
GWGigawatt of AI accelerator capacity; our build uses about $13bn of Data Center AI revenue per GW.
Helios, MI455XAMD rack-scale AI system and its accelerator, in volume from Q3’26.
EPYC (Venice, Turin)AMD server CPU brands and generations.
PEGP/E divided by EPS growth in percent, here FY2 P/E over FY1-to-FY2 growth.
Non-GAAPAMD basis excluding amortisation of acquired intangibles, stock-based compensation, warrant fair value and other items.
Cost of equityRisk-free 4.9% plus beta 1.93 times equity risk premium 4.23%, giving 13.05%.
Exhibit 102: Data sources and as-of dates
Company dataAMD 10-K and 10-Q filings, Q3’26 release and call (3 Nov 2026), proxy statement 2026; company filings data.
Market dataNasdaq closes via Yahoo Finance to 5 Nov 2026; holders and short interest as of the latest Yahoo Finance data.
ConsensusInvestingPro (5 Nov 2026), S&P Global via stockanalysis.com (28 Oct 2026), TipRanks (4 Nov 2026).
Industry dataCompany capex guidance; third-party share and market-size estimates (Global Market Insights, TrendForce and CPU unit and revenue share trackers). These are estimates, not AMD disclosures.
PART 5 · APPENDIX · 21 DISCLOSURES

Analyst certification and important disclosures

Ratings

Ratings reflect the expected total return to our 12-month price target, including dividends, weighted across bull, base and bear scenarios.

BuyExpected total return above +10%
HoldExpected total return between −10% and +10%
SellExpected total return below −10%

We review ratings at each set of results and whenever the facts change materially. A share price can move outside its band before the rating changes, and targets may not be reached.

Analyst certification

The Penmark Team certifies that (1) the views in this report accurately reflect its views about the subject company and its securities, and (2) no part of its compensation was, is or will be directly or indirectly related to the specific ratings, price targets or views in this report.

Contact

Penmark Research

Questions about our research:
research@penmarkresearch.com

All our research and coverage:
penmarkresearch.com

Basis of preparation

This report uses information publicly available up to its publication date. Share prices are closing prices on the trading day before the publication date, unless stated otherwise. Forecasts, valuations and price targets are Penmark Research’s own, and forecast periods are marked E. Sources are shown beneath each exhibit.

Valuation and price targets

Price targets are based on the valuation methods set out in each report and are weighted across bull, base and bear scenarios. The main risks to each rating and price target are described in the report. Penmark Research updates its coverage after each set of results but is not obliged to update a report at any other time.

Independence and conflicts of interest

Penmark Research is independent. It does not provide investment banking, brokerage, market-making or asset management services, does not trade securities, and does not accept payment from any company for research coverage. Companies do not see or review Penmark Research reports before publication.

No member of the Penmark Team holds a long or short position, directly or through derivatives, in the securities of any company Penmark Research covers.

Regulatory status

Penmark Research publishes general, impersonal investment research on a regular basis. It is not registered as an investment adviser with the US Securities and Exchange Commission or any state securities authority, is not a registered broker-dealer and is not a member of FINRA. Penmark Research does not give advice tailored to any person’s circumstances.

Not investment advice or an offer

This report is for information only. It is not personalised investment advice and is not an offer or solicitation to buy or sell any security or other financial instrument. It does not take into account any reader’s objectives, financial situation or needs. Readers should make their own investment decisions and seek independent professional advice where appropriate.

Forward-looking statements and risk

Forecasts, valuations and price targets are estimates based on assumptions that may prove wrong, and they may not be achieved. Opinions are current as of the publication date and may change without notice. Investing in equities involves risk, including the loss of capital. Share prices can be volatile, and past performance is not a reliable indicator of future results.

Data and sources

Company financials, operating data and filings come from company filings, earnings releases and earnings calls. Market data and any consensus estimates come from third-party sources named beneath the relevant exhibit; consensus figures from aggregators are unofficial. Penmark Research does not use other brokers’ research. Third-party data belong to their providers and may not be extracted from this report.

Accuracy and liability

Information is taken from sources believed to be reliable, but Penmark Research makes no representation or warranty as to its accuracy, completeness or timeliness. To the fullest extent permitted by law, Penmark Research accepts no liability for any loss arising from use of this report. Descriptions of litigation and regulatory matters reflect allegations and public records, not findings, unless stated. If a published report is corrected, the corrected version carries an “Updated” date at penmarkresearch.com.

Distribution and access

Penmark Research reports are available to registered users of penmarkresearch.com. Downloads and programmatic access are governed by the Penmark Research subscription terms. This report is not directed at, or intended for, any person in Singapore or India, and access from those countries is blocked. It is not intended for distribution to, or use by, any person in any jurisdiction where that would be contrary to law or regulation or would subject Penmark Research to registration or licensing requirements there.

Copyright and permitted use

© Penmark Research. All rights reserved. Penmark Research is a trade name of Brightriver Labs. Readers may share a link to this report but may not copy, reproduce, redistribute, sell or use its content, including for text and data mining, without written consent. Contact: research@penmarkresearch.com.

Keep reading

Sign in to Penmark