Global Semiconductors: Watts to Tokens: AI Power Semis and the Transition to 800V Data Centers

Research evidence excerpt

Global Semiconductors: Watts to Tokens: AI Power Semis and the Transition to 800V Data Centers

Accessible version

Global Semiconductors

Watts to Tokens: AI Power Semis and the

Transition to 800V Data Centers

Industry Overview

AI Power: transformative opportunity for analog semis 25 May 2026 Corrected

Power is becoming a critical constraint in AI scaling. We estimate rising compute density Equity

drives rack power capacity up 100x from 10-15kW in traditional cloud servers to 1.5MW Global

for Nvidia’s Feynman platform (CY29/30). Existing infrastructure cannot meet demand, Semiconductors

requiring wholesale revamp of power delivery from grid to rack to GPU/XPU core. This Vivek Arya

creates a transformative opportunity for analog semi vendors to shift mix away from Research Analyst

cyclical auto/industrial demand toward secular, durable AI markets, where diverse BofASvivek.arya@bofa.com

architectures, new components, novel materials (wide-bandgap semis), and products Duksan Jang

create an unparalleled chance to differentiate and take share. In this report, we build a Research Analyst

BofAS

bottoms-up AI analog semi industry model translating accelerator/rack demand into duksan.jang@bofa.com

content pools across components (VRM, IBC, SST, etc.) and device types (Si, SiC, GaN, Michael Mani

etc.) into supplier revenue share across low power (<200kW) and high power (600kW+) Research Analyst

racks. We estimate ~233GWs deployed through CY30 expands today’s $7.9bn AI analog michael.mani@bofa.com

market to $27bn by CY30 (28% CAGR) inside a $1.7tn AI data center systems LT TAM. Didier Scemama >>

Research Analyst

Data Centers: $25bn TAM led by 100x jump in rack power MLIdidier.scemama@bofa.com(UK)