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.
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.
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.
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.
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.
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.
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.
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.
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.
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.
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.
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.
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.
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.
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.
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.
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.
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.
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?
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.
Hyperscaler capex pauses after 2026’s +67% surge; 2027 EPS misses and the 59× multiple compresses toward 30× — both blades cut at once.
Rack-scale systems are new territory for AMD; a two-quarter slip breaks the warrant-milestone cadence and the 2027 EPS bridge with it.
OpenAI-Broadcom, Meta MTIA, and Google TPU silicon absorb the inference workloads the MI-series was contracted to serve.
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.
MI308 rules already cost ~$800M in 2025 charges and cap China at ~$100M/quarter; every rule change reprices guidance overnight.
Every AMD die is TSMC-made; a blockade or conflict severs the entire product line — low odds, existential consequence.
A dot-com-style collapse in AI infrastructure ROI turns contracted gigawatts into renegotiated ones; the 2028 bear trough gets a zero added.
Up to 320M new shares (~20%) vest as OpenAI/Meta milestones hit — the better the ramp goes, the more the share count grows.
AMD shares CoWoS and HBM capacity with Nvidia; allocation shortfalls cap the very ramp the deals require.
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.
Hover the dotted terms in the prose, or scan the desk’s working definitions here.