Quick Summary
Covering the latest research from top Wall Street investment banks

AI-Driven Power Semiconductor Revolution: 800V Architecture Unlocks a $27 Billion New Market

Institution
Bank of America
Date
20260525
Authors
Vivek Arya, Duksan Jang, Michael Mani, Didier Scemama, Mikio Hirakawa
Company
Texas Instruments, Infineon, Analog Devices, onsemi, STMicroelectronics, Renesas, Monolithic Power Systems, Microchip Technology, Texas Instruments, Analog Devices, onsemi, STMicroelectronics, Microchip Technology
Ticker
TXN, IFX, ADI, ON, STM, 6723, MPWR, MCHP
Industry
Semiconductors, AI, VR, Specialty Industrial Machinery, Semiconductors
Rating
BullishHigh confidenceLong-termThe report argues that rising AI compute density will drive structural growth in the power semiconductor market, favoring analog chip and wide-bandgap semiconductor manufacturers poised to benefit from data center transformation.
AuthorsVivek Arya, Duksan Jang, Michael Mani, Didier Scemama, Mikio Hirakawa
CoverageOther
Business segmentsData Centers、Power Infrastructure
Research firm divisions/subsidiariesBofA SECURITIES(Subsidiary/Legal Entity)

AI summary card

AI-Driven Power Semiconductor Revolution: 800V Architecture Unlocks a $27 Billion New Market

As AI rack power scales toward 1MW, traditional power delivery architectures become obsolete, making 800V DC power delivery inevitable and driving the analog semiconductor TAM to $27 billion by 2030.

— | Target Price —
AI Data Centers800V DC Power DeliveryPower SemiconductorsSilicon Carbide (SiC)Gallium Nitride (GaN)Solid-State TransformersAnalog Chips
  • AI rack power will surge from the traditional 10–15kW to 1.5MW by 2030, creating cumulative new power demand of 233GW for AI computing.
  • The AI-related analog semiconductor market is projected to grow from $7.9 billion in 2025 to $27 billion by 2030, at a CAGR of 28%.
  • The 800V DC architecture will replace legacy 48V/54V AC architectures, significantly reducing copper usage and improving energy efficiency.
  • Silicon Carbide (SiC) and Gallium Nitride (GaN) will capture the largest share gains in high-voltage conversion and protection applications.
  • Texas Instruments (TXN) maintains the top market share, Infineon sees the fastest share gains, and ADI strengthens its position through the Empower acquisition.
  • Emerging infrastructure technologies like Solid-State Transformers (SST) and Solid-State Circuit Breakers (SSCB) will experience explosive adoption between 2028 and 2030.

Report interpretation

Overview

This report provides an in-depth analysis of how AI compute expansion is reshaping data center power infrastructure. As GPU cluster scale grows, single-rack power is advancing from hundreds of kilowatts toward megawatt levels (1MW), pushing traditional power architectures beyond their physical limits. The report identifies 800V DC (800 VDC) power delivery as the key solution to address challenges related to high power density, energy efficiency, and copper costs. This transition creates structural opportunities for the analog semiconductor industry, with the relevant market size expected to reach $27 billion by 2030. The report details value chain shifts from the grid to the chip core and analyzes competitive dynamics among major semiconductor vendors in this transformation.

Core views

Exponential growth in AI compute density is the primary driver behind power architecture transformation. Traditional cloud server racks operate at only 10–15kW, while Nvidia’s Blackwell architecture already reaches 100–120kW, and future Feynman platforms are expected to exceed 1.5MW. This surge renders legacy 48V/54V AC distribution architectures unsustainable due to space constraints, copper consumption (a 1MW rack requires 200kg of copper busbars), and conversion inefficiencies. The 800V DC architecture reduces the number of voltage conversion stages and increases transmission voltage to lower current, significantly cutting copper usage (~45% reduction) and boosting system efficiency (~5% improvement). The report forecasts that cumulative new global AI data center power demand will reach 233GW by 2030, directly expanding the analog semiconductor TAM 'from rack to core' from $7.6 billion in 2025 to $25 billion in 2030—a 27% CAGR. Meanwhile, the infrastructure market 'from grid to data hall' will grow at a 49% CAGR to $1.8 billion. At the device level, silicon-based analog ICs remain the largest segment by market size ($14 billion by 2030), but wide-bandgap semiconductors will experience explosive growth. Silicon Carbide (SiC), with its superior performance under high voltage and temperature, will dominate front-end rectification, protection, and medium-voltage conversion; Gallium Nitride (GaN), leveraging its high-frequency switching capabilities, will play a critical role in high-density DC/DC conversion near compute units. SiC and GaN market shares will rapidly increase from current low bases to capture most of the incremental value. In terms of competitive positioning, vendors with full-stack product portfolios and system-level design capabilities hold the greatest advantage. Texas Instruments (TXN) currently leads in market share thanks to its broad portfolio of power management products; Infineon, with comprehensive coverage across Si, SiC, and GaN, is projected to gain share the fastest—from 12% in 2025 to 17% in 2030—becoming the second-largest supplier; ADI strengthened its position in processor-proximal power delivery (Integrated Voltage Regulator) through its acquisition of Empower; and onsemi exhibits high leverage in SiC and novel vertical GaN (vGaN) technologies.

Analysis framework

The report employs a bottoms-up industry modeling approach. First, it estimates per-rack power requirements and cumulative global GW deployments for successive generations of accelerator platforms based on roadmaps from Nvidia and other GPU/XPU vendors. Second, it breaks down the power delivery path into two segments—'grid-to-data-hall' and 'rack-to-core'—and further decomposes them into specific components such as PSUs, IBCs, VRMs, SSTs, and SSCBs. Finally, it calculates market sizes for various devices (Si, SiC, GaN, Analog ICs, etc.) and revenue shares for key suppliers by analyzing changes in semiconductor content per rack or per MW under different voltage architectures (48V vs. 800V). This methodology precisely captures value migration driven by technological iteration.

Methodology notes

  • Industry/Industrial Analysis FrameworkVolume-price decomposition

    Volume-Price Decomposition

    The report decomposes market growth into 'volume' (new GW of AI power demand, number of racks) and 'price' (ASP of semiconductor content per rack or per MW). Although some legacy component volumes may decline with higher voltage levels, increased per-unit value and the introduction of new technologies substantially raise overall content value.

  • Industry/Industrial Analysis FrameworkUpstream-Midstream-Downstream Value Chain Transmission

    Upstream-Midstream-Downstream Value Chain Transmission

    The report analyzes value distribution along the physical path of power flow—from grid → substation → data hall → rack → motherboard → chip core—and notes that as power density increases, value is shifting from traditional passive components and low-voltage distribution toward active high-voltage conversion, protection, and control functions closer to the load.

  • Industry/Industrial Analysis FrameworkSubstitution Effect Analysis

    Substitution Effect Analysis

    The report examines substitution logic between new and legacy materials. For example, in high-voltage scenarios, SiC is replacing traditional silicon MOSFETs due to its low on-resistance and high breakdown voltage; GaN is displacing silicon-based solutions in dense conversion applications thanks to its high-frequency characteristics. Such substitutions are driven not only by performance but also by physical constraints (e.g., copper weight, thermal dissipation space).

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

  • Texas Instruments (TXN)
    Beneficiary
    Strengths
    Boasts the broadest portfolio of power semiconductors, currently holds the highest market share in AI analog semiconductors, and maintains a strong position in protection and control segments.
    Comparison
    Leads in market share but faces intense competition from Infineon in wide-bandgap semiconductors.
  • Infineon (IFX)
    Beneficiary
    Strengths
    Offers the most comprehensive AI product portfolio covering Si, SiC, and GaN, with complete coverage across the entire chain from grid to core, and is projected to gain share the fastest.
    Comparison
    More aggressive than TXN in wide-bandgap and high-voltage domains; expected to become the second-largest supplier by 2030.
  • Analog Devices (ADI)
    Beneficiary
    Strengths
    Strengthened its competitiveness in processor-proximal power delivery ('last inch') through the acquisition of Empower, gaining integrated voltage regulators and silicon capacitor technology.
    Comparison
    Excels in high-performance analog and precision control, with solid market positions in optical and sensing applications.
  • onsemi (ON)
    Beneficiary
    Strengths
    Highly leveraged in SiC and novel vertical GaN (vGaN) technologies, well-positioned to increase wallet share in high-voltage conversion and protection segments.
    Comparison
    Strong competitiveness in discrete power devices, particularly in emerging infrastructure applications like SSCBs and SSTs.

Key data

  • AI Analog Semiconductor TAM (2030E)$27 BillionGrowing from $7.9B in 2025 at a 28% CAGR
  • Cumulative New AI Power Demand (2025–2030)233 GWAnnual additions rising from 17GW in 2025 to 60GW in 2030
  • Semiconductor Content per Rack (1MW+ Racks)~$917,000Far exceeding the current $36,000 for 100–160kW racks
  • SiC Market CAGR (2025–2030)63%Growing from $183M in 2025 to $2.098B in 2030
  • GaN Market CAGR (2025–2030)69%Growing from $118M in 2025 to $1.612B in 2030
  • Infineon Projected Market Share (2030E)17%Up from 12% in 2025, becoming the #2 supplier

Impact & implications

For the semiconductor industry, this marks a significant shift from cyclical automotive/industrial demand toward a hybrid model increasingly driven by structural AI demand. Vendors with high-voltage, high-reliability product portfolios and system-level reference design capabilities will gain pricing power and market share. For data center operators, adopting 800V architecture requires upfront capital investment but delivers substantial long-term operational cost savings (TCO reduction of ~30%) and overcomes power density bottlenecks. At the infrastructure level, the introduction of solid-state transformers and circuit breakers addresses legacy pain points of long lead times and large footprints, accelerating data center deployment speed.

Risks

  • Slower-than-expected maturity of the 800V ecosystem, leading to deployment delays.
  • Reliability and thermal management issues with Solid-State Transformers (SST) and Solid-State Circuit Breakers (SSCB) in large-scale deployments.
  • Slower-than-expected growth in AI compute demand, causing data center construction delays.
  • Insufficient capacity expansion or slower-than-expected cost reductions for wide-bandgap semiconductors (SiC/GaN).

What to watch

  • Actual power specifications and deployment timelines for Nvidia’s Rubin Ultra and Feynman platforms.
  • Pilot programs and scale adoption progress of 800V DC architecture in major cloud service provider data centers.
  • Commercialization status and key supplier orders for Solid-State Transformers (SST) and Solid-State Circuit Breakers (SSCB).
  • Penetration rate trends of SiC and GaN in data center power modules.
Zhejiang ICP No. 2022035445-5
Disclaimer: Market data, charts, indicators, research views, and other information provided on this website are intended solely for information display, research communication, and educational reference. They should not be regarded as personalized investment advice, securities recommendations, trading instructions, solicitations, or guarantees of return. While we strive to improve the reliability of our data and content, such information may still be subject to delays, errors, incompleteness, or untimely updates due to source differences, methodological limitations, system processing, or market volatility. Users should exercise independent judgment based on their own circumstances and bear all risks and responsibilities arising from the use of this website.

Settings

Sign in to view recent logins