← AVGO/AI Semiconductor — Scenario Model

Blue = input assumption|Black = computed

AI Semiconductor — Scenario Model

Current Price

$301

AI Semiconductor Component

$241/sh

Market Implies

86% chance of upside scenarios

7% reversion probability

Scenario Overview

Scenario	EV	$/sh	Prob	Key Gate
mega bull	$1.5T	$320	60%	next gen 4d
strong bull	$970B	$205	19%	100b target
base	$459B	$97	7%	—
bear	$157B	$33	7%	mediatek google ramp
reversion	$74B	$16	7%	customer insources

Gate Milestones

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

2027

100b exceeded

AI revenue exceeds $100B in FY2027 -- Hock Tan's target validated

Unlocks: Confirms ASIC platform model at hyperscale, justifies infrastructure investments

CEO: 'line of sight to >$100B AI chip revenue in 2027' [growth_trajectory.json]; $73B backlog supports conversion [growth_trajectory.json]

2028

8th customer

8th hyperscaler signs XPU deal (Apple Baltra confirmed + another)

Unlocks: Reduces Google 78% concentration risk; validates multi-customer platform model

Apple Baltra AI server chip for 2027 deployment [xpu_customers.json]; MediaTek CEO confirms negotiations with 2nd hyperscaler [risks.json]

2030

asic 40pct share

ASICs capture >40% of AI compute silicon (up from ~20% inference today)

Unlocks: Structural shift from GPUs to custom silicon for inference-dominated workloads

ASICs already ~20% of inference [competitive_dynamics.json]; inference becoming dominant AI workload [competitive_dynamics.json]

2033

next gen 4d

Next-gen 4D packaging extends lead vs Marvell/Alchip/MediaTek

Unlocks: Sustains design advantage for next decade of ASIC programs

3.5D XDSiP already shipping: 6,000mm2, 12 HBM stacks [competitive_dynamics.json]

2027

100b target

AI revenue hits $100B in FY2027

Unlocks: Confirms Hock Tan's target; validates backlog conversion

CEO: '>$100B' [growth_trajectory.json]; $73B backlog over 18mo [growth_trajectory.json]

2027

vera rubin impact

NVIDIA Vera Rubin narrows ASIC efficiency gap for inference workloads

Unlocks: Hyperscalers shift marginal workloads back to GPUs; ASIC growth decelerates

Vera Rubin: 6 chips, Rubin GPU + Vera CPU + NVLink 6 + full stack [competitive_dynamics.json]

2029

mediatek google ramp

MediaTek captures significant share of Google TPU v8e/v9 production

Unlocks: Google diversifies ASIC supply; Broadcom loses pricing power on largest customer

MediaTek 20-30% cheaper on TPU v7e/v8e [risks.json]; Google planning 5M->7M TPU units [risks.json]

2027

google diversifies

Google shifts 30%+ of TPU volume to MediaTek, reducing Broadcom share

Unlocks: Breaks Broadcom's near-monopoly on Google TPU design

MediaTek 20-30% cost advantage on TPU v7e/v8e [risks.json]; Google planning 5M TPU units 2027, 7M 2028 [risks.json]

2029

customer insources

Google/Meta build internal ASIC design capability, reducing Broadcom IP premium

Unlocks: Commoditizes the ASIC design function; Broadcom becomes contract designer

SemiWiki: Google testing ASIC design commoditization [risks.json]

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 $241/share?

Scenario	AI Semi EV	Per Share	Implied Prob.	Weighted Contrib.
mega bull	$1.5T	$320	59.8%	$191
strong bull	$970B	$205	19.5%	$40
base	$459B	$97	6.6%	$6
bear	$157B	$33	7.0%	$2
reversion	$74B	$16	7.2%	$1
Total			100%	$241/sh

Note: These probabilities apply to AI Semiconductor's contribution to Broadcom equity. Infrastructure Software and Non-AI Semi contribute separately.

Target contribution: $241/sh. Residual: $0.0/sh.