HScreen · AMD

AMD · Advanced Micro Devices, Inc. 7d old

conviction3 → 3.00 trust-weighted

coverage tierT1-FULL

horizon12 months

Return distribution probability-weighted scenarios, never collapsed to a point

-30%25%

+14%42%

+50%33%

EV +14.88%

why these probabilities: PRICE BASIS 514.39 USD as of 2026-08-17 (fmp:/stable/profile?symbol=AMD). BASE (+14%, p=42) is the market's expectation delivered, not my forecast. Consensus for FY2026 (period 2026-12-27) is revenue +49.6% and EPS +91.9%, and FY2027 (2027-12-27) revenue +68.8% / EPS +98.7%, on 33 and 36 revenue contributors respectively (fmp:/stable/analyst-estimates?symbol=AMD&period=annual, as_of 2026-08-17T16:29:47Z, grade EST). Delivering that on an unchanged forward multiple compounds the equity at roughly the rate the out-year estimates are pulled forward over twelve months; I take +14%, which also sits below the street consensus target of 593.92 (+15.5% implied) and well below the median 625 (+21.5%) — i.e. my base is deliberately at the conservative end of the street's own dated mark rather than derived from it. BULL (+50%, p=33) assumes revenue growth MATERIALLY ABOVE consensus — call it FY2027 revenue ~+85% against consensus +68.8%, with the gap coming from Instinct/Helios units the street cannot yet model because AMD guides one quarter at a time (AMD/AMD_2025Q4.json:24). Mechanism: Data Center compounding off the 2026Q1 record USD 5.8bn base at the +57% y/y already delivered (AMD/AMD_2026Q1.json:12), Helios ramping Q4-weighted per management (AMD/AMD_2026Q1.json:28, AMD/AMD_2026Q2.json:21), 2nm capacity being brought up (AMD/AMD_2026Q2.json:58), and the demand funded by disclosed cu

not built — options-implied distribution on the same axis, with the KL number and where the disagreement sits
needs: desk-options implied bands (v1.4 Part 4a) — activates at 60% coverage; currently 33%

Visibility how far ahead this name can be seen

0.50 scored on 1/4 components

horizon0.500

dispersion_tightnessabsent — only 1 broker in unit $M — dispersion undefined

delivery_reliabilityabsent — guidance_vs_delivery returned nothing

tape_stabilityabsent — revision tape EMPTY (0 up, 0 down) — not flat, ABSENT

Trust how much weight this name's own record can carry (credibility Z)

guidance_deliveryunavailable

estimates_depth0.429

evidence_depth0.318

composite0.373

Sizing survival-first size (half-Kelly) vs the enforced ladder

kelly not computable — no data

Payout durability does the distribution survive a downturn?

PASS

DPS trajectoryinsufficient_history

interest cover now28.2x

interest cover at FORWARD rates17.65x

Falsifiers what would prove this wrong

Instinct/MI gross margin converges toward corporate average as the MI400/Helios generation reaches scale, rather than staying structurally below it
Data Center segment revenue continues compounding at or above +40% y/y off the 2026Q1 record base of USD 5.8bn
Helios ramps to volume in Q4 2026 as management has guided, without slipping into 2027
Top-5 hyperscaler customers continue to buy merchant accelerators alongside internal silicon rather than reallocating budget away from merchant supply
Hyperscaler aggregate capex tracks toward the disclosed/estimated USD 1,335-1,475bn trajectory rather than being cut
Leading-edge 2nm capacity is available to AMD in the volume its ramp assumes
Token spend at the demand origin keeps growing, so tier-2 capex has a payer behind it

Conflicts surfaced, never netted out (T4)

The pack reports 247 estimate rows for AMD but the eps_drift and rev_drift typed functions both return n_up=0, n_down=0, direction=flat, breadth=0.0, median_move_pct=null over a 365-day window. For a name with 70 coverin	high
Street consensus is Buy (50 buy / 20 hold / 0 sell of 70) with a consensus target of 593.92 and a median of 625 — implying +15.5% to +21.5%. My base leg is +14%, below both. I am not more bullish than the street on the l	medium
Management frames MI gross margin below corporate average as a deliberate choice to win TCO and share (AMD/AMD_2025Q2.json:35) and expects margin to progress by generation as scale arrives (AMD/AMD_2025Q4.json:62). I tre	high
The growth score leans on customer-disclosed backlog (MSFT USD 368bn, GOOGL USD 514bn cloud backlog, AMZN growing backlog) and on UBS hyperscaler capex estimates. None of this is AMD's contracted revenue, and the substit	high
SemiAnalysis TCO work says MI455X is competitive against GB300 NVL72 and even VR NVL72 on marketed dense FP8 TFLOPs, but the same document's framing ("Can AMD break the CUDA Moat") leaves the software-moat question open	medium

The case against independent adversary, different model family (P6) weakened

Version history from the research record

archived versions2

archived	hash	bytes
2026-08-17T10:34:17	5a760901eb23	33,260
2026-08-19T05:54:53	4aacbb5990ac	35,840

What changed in the latest version

+	provenance:
+	produced_by: analyst-aisemi
+	produced_at: 2026-08-17
+	evidence_pack: EVIDENCE PACK — AMD, generated 2026-08-17T16:29:17Z by build_evidence_pack.py
+	price_basis:
+	price: 514.39
+	currency: USD
+	as_of_date: 2026-08-17
+	source_id: fmp:/stable/profile?symbol=AMD
-	value: 'Advanced Micro Devices designs x86 server and client CPUs (EPYC, Ryzen), discrete and data-centre
-	GPUs (Instinct MI-series), and semi-custom/embedded silicon; it is fabless and depends on TSMC leading-edge
-	nodes plus CoWoS-class advanced packaging. Data Center is now the dominant segment: Q1 FY2026 Data
-	Center revenue was a record $5.8bn, +57% y/y and +7% q/q, driven by EPYC share gains and the Instinct
-	ramp, with segment operating income $1.6bn (28% of revenue). The 2026-27 thesis rests on the rack-scale
…	diff truncated

Rendered from canonical artifacts only (T1) — no terminal database exists. Every figure is read from the system's own files at render time; where an artifact is missing the panel names the dependency rather than showing an empty box. No execution path exists anywhere in this interface (T5), not even a link.