← AMD/GPU Share Capture Premium — Scenario Model
Blue = input assumption|Black = computed

GPU Share Capture Premium — Scenario Model

Current Price
$202
GPU Share Capture Premium Component
$70/sh
Market Implies
94% chance of upside scenarios
3% bear probability

Scenario Overview

ScenarioEV$/shProbRevenueEbit Margin %FcfGpu Share Target %Key Gate
mega bull$157B$9763%full warrant vest
strong bull$53B$3229%deal expansion
moderate$6B$33%asic share grows
limited($0B)$03%asic dominant
bear($1B)$03%deal collapse

Gate Milestones

Growth is gated by milestones. Cost step-changes and capability unlocks are tied to specific milestones.

2026
Y1
mi450 validation
MI450/Helios validated in OpenAI/Meta 1 GW deployments
Unlocks: Proves AMD can deliver at hyperscale. Revenue recognition begins
OpenAI/Meta binding 1 GW each H2 2026 [AMD 8-K]
2027
Y2
openai meta 2gw
OpenAI+Meta exercise 2nd GW tranche each
Unlocks: Confirms performance satisfaction, unlocks remaining deal value
Deals structured with milestone-based deployment [AMD 8-K]
2028
Y3
third hyperscaler
Third hyperscaler deal (Google, AWS, or Microsoft)
Unlocks: AMD becomes industry standard, not just two-customer story
Microsoft already runs Copilot on MI300X [Microsoft Tech Community]
2030
Y5
rocm critical mass
ROCm reaches self-sustaining ecosystem
Unlocks: Developer momentum creates flywheel, reduces CUDA advantage
JAX postings +340% YoY; Triton enables hardware-agnostic AI [Indeed/OpenAI]
2033
Y8
full warrant vest
Full warrant vesting (320M shares)
Unlocks: Maximum dilution but also maximum revenue conversion
Requires AMD stock at $600 for final tranches [AMD 8-K]
2026
Y1
mi450 launch
MI450 launches on schedule Q3 2026
Unlocks: Revenue ramp begins with binding 1 GW per deal
AMD CES 2026: MI450 on track for H2 2026
2028
Y3
deal expansion
OpenAI/Meta exercise 3rd GW tranche
Unlocks: Deal conversion at ~50% of maximum potential
Milestone-based structure [AMD 8-K]
2027
Y2
binding gw only
OpenAI/Meta exercise only binding 1 GW each
Unlocks: Remaining 5 GW per deal does not convert — performance not met
Only first 1 GW binding [AMD 8-K]
2033
Y8
asic share grows
Custom ASICs capture most inference share growth
Unlocks: AMD squeezed — neither the best (NVIDIA) nor the cheapest (ASICs)
ASICs: 40-65% TCO advantage for inference [Introl/TrendForce]
2028
Y3
deal partial
OpenAI/Meta deploy only partial 1 GW, shift to ASICs
Unlocks: Performance doesn't justify full binding commitment
OpenAI 10 GW Broadcom ASIC deal as primary alternative [TrendForce]
2030
Y5
asic dominant
Custom ASICs capture 50%+ of non-NVIDIA AI compute
Unlocks: AMD GPUs become niche — too expensive vs ASICs, too limited vs NVIDIA
Broadcom: 5 hyperscaler XPU customers, $73B backlog [Q4 Earnings]
2027
Y2
mi450 underperforms
MI450 benchmarks disappoint vs Vera Rubin NVL72
Unlocks: OpenAI/Meta shift remaining budget to NVIDIA + ASICs
NVIDIA GTC 2026: Vera Rubin 10x inference cost reduction vs Blackwell
2029
Y4
deal collapse
OpenAI/Meta do not exercise beyond minimum
Unlocks: Binding 1 GW not fully utilized; rest shifts to ASICs/NVIDIA
Only first 1 GW binding; remainder milestone-dependent [AMD 8-K]

Full Year-by-Year Models

Blue = input · Black = computed · Click to expand

Market-Implied Probability Reconciliation

Given these scenario EVs, what probability weights would the market need to assign to produce the current market-implied value of $70/share?

ScenarioGPU Share Capture EVPer ShareImplied Prob.Weighted Contrib.
mega bull$157B$9762.7%$61
strong bull$53B$3228.5%$9
moderate$6B$32.9%$0
limited($0B)$-02.9%$-0
bear($1B)$-12.9%$-0
Total100%$70/sh

Note: These probabilities apply to the INCREMENTAL value of AMD capturing AI GPU share beyond the base case (~8%). This is the speculative bet: $70/share is what separates AMD at $202 from a ~$130 stock.

Target contribution: $70/sh. Residual: $0.0/sh.