01 · Equity deep-dive — synthesized analyst desk
AMD
$517.41 ▼ 11.5% off Jul ’26 high
NASDAQ · SEMICONDUCTORSMKT CAP ≈ $844B52-WK $135.91 – $584.73AS OF JUL 8, 2026 (CLOSE)

Twelve gigawatts of proof. One multiple of doubt.

The perennial second source just signed the AI era’s two biggest supply deals — 6 GW each with OpenAI and Meta — and the stock has quadrupled in a year to sit above Wall Street’s own consensus target. The question the tape is pricing: is contracted demand at 59× forward earnings a franchise — or a cycle top bought with 20% of the company? Five analyst lenses, three scenarios, four time horizons.

The verdict · TL;DR
One question decides the stock: are the OpenAI and Meta gigawatts durable franchise revenue, or vendor-financed demand at the top of an AI-capex cycle? The business is compounding — revenue +38%, data center +57%, EPS on track to nearly double in 2026 — but at $517 the stock trades through the Street’s $512 consensus, at a 59× forward multiple, days after a sector-wide capex scare knocked 11% off the top. The prob-weighted math says roughly flat at 12 months; the payoff, if the ramp yields, is in years two through five.
5-yr · prob-weighted
$815
+58% vs $517.41
52-week playback · where the tape sits ❚❚ 11% off the top · through consensus
$517.41 · Jul 8, 2026 consensus $512 · −1%
$135.91 · 52-wk low · Jul ’25 $584.73 · 52-wk high · Jul 2, ’26
Price history + cone of outcomes · 2024 → 2031
HISTORICALBULLBASEBEARPROB-WTD
$1200$600$300 $150$75 202520262027 202820292030 2031 $78 trough · Apr ’25 $136 · 52-wk low · Jul ’25 $264 · Oct ’25 $585 high · Jul 2 ’26 $815 $545$625$680 street $512 $1380 $785 $390 TODAY · $517.41

Gray line = AMD’s actual ride into today ($78 post-tariff trough Apr ’25 → $264 OpenAI-deal peak Oct ’25 → $191 Feb ’26 → $585 all-time high Jul 2 ’26 → $517.41 after this month’s semiconductor selloff); colored paths = synthesized scenario midpoints forward, probability-weighted (base 45% · bull 25% · bear 30%). Log scale — every gridline is a doubling; mid-year marks. Wall Street 12-month consensus ≈ $512 (range $320–$700, 42 of 51 analysts at Buy or better) sits below today’s close.

Re-weight the scenarios

Those probabilities are a judgment call — so make them yours. Drag to set how likely the bear and bull cases are (base takes the remainder); the blended target below, the dotted line on the chart, and the prob-weighted row of the scenario cards all update live.

30% bear 45% base 25% bull
Blended 5-yr expected $815 +58% vs $517.41
+38%
Q1’26 Revenue ($10.25B)
+57%
Data Center ($5.78B)
+43%
Non-GAAP EPS ($1.37)
55%
Non-GAAP Gross Margin
$2.57B
Record Quarterly FCF
$8.5B
Net Cash Position
$11.2B
Q2 Guide (+46% YoY)
12 GW
Contracted AI Capacity (OpenAI + Meta)
02 · The panel — five ways to read the same wafer

Five analyst lenses, five answers

The same fundamentals support targets from $265 to $635 depending on which framework you trust. Each lens below is a synthesized expert perspective with its own 12-month target — three of the five land below today’s price.

Growth / Momentum PM

The Second Franchise

12 GW of contracted OpenAI + Meta capacity starts monetizing in H2 ’26 — booked demand, not TAM slides. Revenue +43% this year, Q2 guided +46% and accelerating, EPS +77% with street 2027 at $12–14.6. A second growth engine in EPYC (server CPU +70%, $120B TAM by 2030) compounds beneath the GPU story. PEG ~1.0 says the growth still isn’t fully priced.

12-MO TARGET $635 · 45× ~$14 FY27 EPS
Quant / Technical

The Extension Trade

Price is 82% above the 200-day average, +281% in twelve months, in the ~99th valuation percentile vs its own history — and the crowd is all-in (2.9% short interest, sell-side chasing the tape with $640–700 targets). Extensions like this resolve by time or −20–40% air pockets; three >50% drawdowns in six years says which is more likely. Trend intact above ~$470, but risk-reward at $517 is negative.

12-MO TARGET $480 · ~40× ~$12 blended ’27 EPS
Moat / Competitive Strategy

The Borrowed Moat

EPYC is a genuine moat — chiplet lead, +70% server-CPU growth, Intel still rebuilding. The AI position is different: hyperscalers pay AMD to exist as a second source, and second-source economics guarantee share while capping price. Nvidia’s mid-70s gross margin vs AMD’s 55% is a price-war trump card, and CUDA is still the default. The warrants buy volume and alignment, not switching costs.

12-MO TARGET $490 · 40× ~$12.25 dilution-adj ’27 EPS
Value / FCF / Quality

The Yield Problem

Quality is confirmed: $8.5B net cash, capex-light at ~$1.2B, FCF up seven-fold since 2023 with record conversion. But $835B of enterprise value against ~$7.4B trailing FCF is a ~0.9% yield — even $15B of 2027 FCF is only 1.8%. And the penny-strike warrants (up to 320M shares, ~20% of the company) are a ~$165B customer rebate netted against the “contracted” revenue. Great business; no margin of safety.

12-MO TARGET $390 · 30× ~$13 ’27 EPS on diluted count
Short-Seller / Cycle Skeptic

The Circular Trade

An $844B company paying its two biggest customers in stock to buy its chips, trading above its own consensus, into a decelerating capex cycle. Hyperscaler AI capex just jumped 67% to $650B — the UBS survey says ~60% of enterprises are already curbing AI spend, and Meta is reselling excess compute. If 2027 capex growth stalls, the $12–14.6 EPS bridge collapses and the multiple de-rates with it. The July SOX crack was the first tell.

12-MO TARGET $265 · 25× $10.50 miss-case ’27 EPS
03 · Wall Street's read

Wall Street 12-month price targets

What the sell-side expects over the next year. Bars are sorted low to high; the dashed line is today’s $517.41 — note that it cuts through the middle of the range. After a +142% YTD run, the market has already spent most of the Street’s upside.

Consensus ≈ $512 (−1%) · 51 analysts, range $320–$700
BUYHOLDSELL
Street low $320 KeyBanc $530 BofA Securities $560 Citi $575 Bernstein $600 Wells Fargo $615 Goldman Sachs $640 UBS $670 Cantor Fitzgerald $700 TODAY · $517.41

Selected 12-month targets from the ~51 firms covering AMD, as revised in late June–early July 2026 (Goldman $450→$640 on Jul 6; Wells Fargo $505→$615 on Jun 30; UBS’s $670 assumes 25%+ merchant-accelerator share). The full consensus is ≈ $512 — about 1% below today’s close — with 42 of 51 at Buy or better and 9 at Hold. The recent raises followed the price up: five of the targets shown were set with the stock already above $500. Firms, ratings, and targets as reported by aggregators; “Street low” is the lowest published target in the S&P Global panel.

04 · Price scenarios — 1 / 2 / 3 / 5 years

Where the ramp leads

Synthesized scenario midpoints (mid-year). Returns shown vs. today’s $517.41. These are illustrative frameworks, not forecasts — five-year outcomes hinge almost entirely on whether AI-infrastructure capex keeps compounding and whether the MI450/Helios ramp yields.

1 Year

Mid-2027
Bull$730+41%
Base$545+5%
Bear$290−44%
Prob-wtd$515−1%

2 Years

Mid-2028
Bull$900+74%
Base$625+21%
Bear$270−48%
Prob-wtd$587+13%

3 Years

Mid-2029
Bull$1060+105%
Base$680+31%
Bear$320−38%
Prob-wtd$667+29%

5 Years

Mid-2031
Bull$1380+167%
Base$785+52%
Bear$390−25%
Prob-wtd$815+58%
Bull case — show the assumptions & math
The AI supercycle extends: hyperscaler capex keeps compounding past $650B, MI450/Helios ships on schedule, AMD takes 25%+ of merchant accelerators, OpenAI and Meta expand beyond the first tranches, and EPYC keeps taking Intel share. EPS runs $7.4 → ~$14.6 (’27, Bernstein’s number) → ~$20 (’28) → ~$34+ by 2031, and the market keeps paying a premium multiple for a two-horse AI market.
1-yr: 50× $14.6 ’27E EPS ≈ $730 · 5-yr: ~40× ~$34.5 ’31E EPS ≈ $1,380 · 5-yr price CAGR ≈ +22%/yr
Base case — show the assumptions & math
The build-out continues but decelerates from 2028; AMD delivers ~$13 of 2027 EPS (street midpoint) and ~20% EPS growth thereafter; warrants partially vest (some dilution); the multiple compresses gradually from 59× toward ~32× as growth normalizes. The stock grinds higher on earnings, not re-rating — and goes roughly nowhere in year one while EPS catches up with the price.
1-yr: ~42× $13 ’27E EPS ≈ $545 · 5-yr: ~32× ~$24.5 ’31E EPS ≈ $785 · 5-yr price CAGR ≈ +9%/yr
Bear case — show the assumptions & math
The capex cycle cracks in 2027: hyperscalers digest, enterprises curb spend (the UBS survey made real), Nvidia cuts price into the MI450 ramp, and 2027 EPS lands near $10.50 instead of $13–14.6. The multiple de-rates toward the ~25–30× semis carried pre-mania. AMD has printed −63% on a milder scare (Mar ’24 → Apr ’25); this path assumes a trough near $270 in 2028 and only partial recovery by 2031.
1-yr: ~28× $10.50 ’27E EPS ≈ $290 · trough ~24× ≈ $270 in ’28 · 5-yr: ~24× ~$16 ’31E EPS ≈ $390 · 5-yr CAGR ≈ −5%/yr
05 · Follow the cash

Revenue, capex, free cash flow & debt ($B)

The fabless engine in one view. AMD spends almost nothing on plant — TSMC carries the capex — so revenue converts to cash fast once margins move. Both theses live in these bars: the bull sees the 2026 revenue bar, the bear asks what happens to it in 2027.

Annual revenue, capex, FCF & total debt · 2023 → 2026E
REVENUECAPEXFREE CASH FLOWTOTAL DEBT
$0$10$20$30$40$50 2023202420252026E

Revenue (sky) doubles from $22.7B (2023) to ~$49.5B (2026 consensus) while capex (clay) stays ~2–4% of sales — $0.55B → ~$2B — because TSMC owns the fabs. Free cash flow (olive) is the operating-leverage receipt: $1.1B → $2.4B → $5.5B (AMD-reported; one aggregator shows $6.7B on a broader definition) → ~$10B run-rate implied by Q1 ’26’s record $2.57B. Total debt (slate) is trivial — $3.9B against $12.3B cash, ~$8.5B net cash. The bear’s question isn’t the balance sheet; it’s whether the 2026 revenue bar has a 2027 sequel. 2026 capex/FCF are estimates.

06 · Earnings power

The EPS ladder the targets are printed on ($)

Every price target is an EPS estimate times an exit multiple. Here is the non-GAAP earnings ladder underneath them — three reported rungs, then the steep part the market is already paying for.

Non-GAAP EPS · reported vs. estimated, 2023 → 2031E
REPORTEDESTIMATE
$0$5$10$15$20$25 2023202420252026E2027E2028E2029E2030E2031E $2.65 $3.31 $4.17 $7.38 $13.00 $16.50 $19.50 $22.00 $24.50

Gray = reported non-GAAP EPS ($2.65 → $3.31 → $4.17, 2023–25). Olive = estimates: 2026’s $7.38 is the 51-analyst consensus (+77%); 2027’s $13 is the street midpoint of a $12–14.6 range (Bernstein above $14, BofA’s $560 target uses 42× 2027); 2028–2031 are this desk’s base-case extrapolation (~20% decelerating growth — Bernstein calls ~$20 by 2028 “plausible” in the bull case). The base case’s ~$24.5 of 2031 EPS at a ~32× exit multiple ≈ the $785 five-year base target. Note the shape: the market is paying today for rungs nobody has climbed yet.

07 · Growth scorecard — yield check

Every line is compounding

Q1 FY26, year-over-year — read these against a stock that already trades above the Street’s target. Unlike the classic “growth intact, stock down” disconnect, here growth and price ran together; the scorecard tells you what has to stay true.

Year-over-year growth by metric · Q1 FY26
COREAI FRONTIER
Embedded (Xilinx) +6% Gaming +11% Client (Ryzen) +26% Total revenue +38% Non-GAAP EPS +43% Data Center +57% Server CPU (Q2 guide) +70%

Core lines (olive) — Embedded, Gaming, Client, total revenue, EPS — are all growing, with the steady segments cycling back after 2025’s Embedded dip. The AI frontier (clay) is what the multiple is built on: Data Center +57% to $5.78B (56% of company revenue and rising), and server CPU guided +70% YoY for Q2 as EPYC takes Intel share toward a $120B 2030 TAM. GAAP EPS grew +91% on the same quarter. The scorecard’s message cuts both ways: nothing in the operating data has cracked — and a 59× forward multiple gives no room for the first crack.

08 · The debate

Bull vs. Bear

The entire valuation argument compresses into one disagreement: are the OpenAI and Meta gigawatts a durable second franchise — or vendor-financed demand at the top of the capex cycle?

▲ THE BULL CASE

  • Demand is contracted, not hoped for. 12 GW across OpenAI (Oct ’25) and Meta (Feb ’26, $100B+ scope); first gigawatt of each deploys H2 ’26 on MI450/Helios — booked ramp, not TAM math.
  • Growth is accelerating at scale. Q1 revenue +38%, Data Center +57%, Q2 guided +46% — the growth rate is rising on a ~$45B run-rate.
  • Earnings are compounding faster than revenue. Non-GAAP EPS +43% in Q1, consensus +77% to $7.38 for 2026, street 2027 at $12–14.6 — PEG ~1.0 despite the headline multiple.
  • A second engine under the GPU story. EPYC server CPU guided +70% YoY with Intel still rebuilding; management sees the server TAM at $120B by 2030.
  • Fabless cash machine. Record $2.57B quarterly FCF, ~$8.5B net cash, capex ~2–4% of sales — operating leverage lands in cash, and the warrants align the two biggest AI buyers with the stock reaching $600.
  • The Street keeps chasing it up, not down. Goldman $450→$640, Wells Fargo $505→$615, UBS $670, Cantor $700 — revisions are one-directional and the lead-customer forecasts “exceed initial expectations.”

▼ THE BEAR CASE

  • The stock trades above its own consensus. $517 vs a $512 mean and $500 median — after +142% YTD there is no sanctioned upside left; every recent target raise followed the price.
  • Circular, equity-subsidized demand. Penny-strike warrants for up to 320M shares (~20% of the company, ~$165B at spot) are the price of the OpenAI and Meta deals — customers were paid in stock to buy the chips.
  • The capex cycle is cresting. Hyperscaler AI capex +67% to $650B in 2026; ~60% of enterprises curbing AI spend (UBS survey); Meta reselling excess compute; Burry short and a −10.8% SOX week say the market noticed.
  • Nvidia holds the price lever. Mid-70s gross margin vs AMD’s 55% means Nvidia can cut price into the MI450 ramp; CUDA remains the default and ROCm adoption is customer-subsidized.
  • Customers are also competitors. Google TPU, Meta MTIA, and OpenAI-Broadcom custom silicon all scale in 2026–27 — the same buyers underwriting AMD’s ramp are building its replacement.
  • History and structure amplify the downside. Three >50% drawdowns in six years (−63% into Apr ’25), beta 2.47, 99th-percentile valuation (59× fwd, 112× EV/EBITDA, ~0.9% FCF yield), and China export whiplash already costing ~$100M a quarter.
09 · Risk map

Risk map — likelihood × impact

Where each risk sits over a 3–5 year horizon, not just how big it sounds. AMD’s problem is the hot corner: the most likely risk — an AI-capex digestion phase — is also the highest-impact one, because the multiple leaves no cushion.

Low impact
Medium impact
High impact
Likely
  • Crowded-momentum washouts
  • Nvidia price response
  • China export whiplash
  • AI-capex digestion & de-rate
Possible
  • Warrant dilution (~20%)
  • HBM / TSMC supply squeeze
  • MI450 / Helios ramp slip
  • Custom-ASIC substitution
Tail
  • Taiwan supply shock
  • AI demand bust

AI-capex digestion & de-rate

Likely × High

Hyperscaler capex pauses after 2026’s +67% surge; 2027 EPS misses and the 59× multiple compresses toward 30× — both blades cut at once.

MI450 / Helios ramp slip

Possible × High

Rack-scale systems are new territory for AMD; a two-quarter slip breaks the warrant-milestone cadence and the 2027 EPS bridge with it.

Custom-ASIC substitution

Possible × High

OpenAI-Broadcom, Meta MTIA, and Google TPU silicon absorb the inference workloads the MI-series was contracted to serve.

Nvidia price response

Likely × Medium

With mid-70s gross margins, Nvidia can cut price into every AMD ramp; AMD’s 55% GM can’t follow without gutting the EPS path.

China export whiplash

Likely × Medium

MI308 rules already cost ~$800M in 2025 charges and cap China at ~$100M/quarter; every rule change reprices guidance overnight.

Taiwan supply shock

Tail × High

Every AMD die is TSMC-made; a blockade or conflict severs the entire product line — low odds, existential consequence.

AI demand bust

Tail × High

A dot-com-style collapse in AI infrastructure ROI turns contracted gigawatts into renegotiated ones; the 2028 bear trough gets a zero added.

Warrant dilution

Possible × Medium

Up to 320M new shares (~20%) vest as OpenAI/Meta milestones hit — the better the ramp goes, the more the share count grows.

HBM / TSMC supply squeeze

Possible × Medium

AMD shares CoWoS and HBM capacity with Nvidia; allocation shortfalls cap the very ramp the deals require.

Crowded-momentum washouts

Likely × Low

Beta 2.47, 2.9% short interest, price 82% over the 200-day — routine −15/20% air pockets like this month’s are the cost of holding.

10 · Plain-language glossary

The jargon, decoded

Hover the dotted terms in the prose, or scan the desk’s working definitions here.

Gigawatt (GW) of compute
AI deals are now sized by the electrical power the GPU fleet draws. One GW is roughly a large nuclear plant’s output — AMD has 12 contracted across OpenAI and Meta.
Merchant accelerator
AI chips sold on the open market (AMD, Nvidia) as opposed to custom in-house silicon (Google TPU, Meta MTIA). The bull case needs AMD at 25%+ of this pool.
Penny warrant
The right to buy shares at $0.01. OpenAI and Meta each hold warrants for up to 160M AMD shares that vest as they buy chips — an equity rebate on the deals.
Second source
The alternative supplier buyers keep alive so the leader can’t dictate price. AMD’s historic role vs Intel in CPUs, now vs Nvidia in AI — share is granted, pricing power isn’t.
Forward P/E
Price ÷ next-twelve-month expected earnings. AMD trades near 59× vs its own 5-year norm of ~25–45× and a ~74× semis sector average.
FCF yield
Free cash flow ÷ enterprise value — the cash coupon you buy. ~0.9% trailing here: about 90¢ of annual cash per $100 of AMD.
Exit multiple
The P/E assumed at the end of the forecast. Multiply it by projected EPS to get a target: 32× on $24.50 of 2031 EPS ≈ the $785 base case.
Prob-weighted
Each scenario’s price × its probability, summed into one expected value. The sliders above let you set the weights yourself.