Inferra AI / Company / MRVL / Slides
Marvell Technology — marvell investor day 2026 presentaion deck (company IR deck)
Marvell Technology (MRVL) · 161 slides · Filed Oct 27, 2026
Every slide of this earnings presentation, with the text as published. Open the original deck ↗ · All MRVL deck slides
Slide 1
Slide 2 — Ross Seymore
Ross Seymore SVP, Investor Relations
Slide 3 — Agenda
© 2026 Marvell. All rights reserved. 3 © 2026 Marvell. All rights reserved. 3 Matt Murphy Chairman and CEO Chris Koopmans President and COO Sandeep Bharathi President, Data Center Group Break Xi Wang SVP and GM, Connectivity Business Russ Esmacher SVP and GM, Data Center Interconnect Business Dave Lazovsky EVP and GM, Data Center Networking Business Will Chu EVP and GM, Custom Cloud Solutions Business Dan Durn EVP and CFO Q&A Agenda
Slide 4 — Basis of presentation
© 2026 Marvell. All rights reserved. 4 Basis of presentation Fiscal Year vs. Calendar Year ▪Financial measures in this presentation correspond to Marvell’s fiscal calendar ▪Marvell’s fiscal year is the 52- or 53-week period ending on the Saturday closest to January 31 ▪As an example, FY26 refers to the period February 2, 2025 through January 31, 2026 ▪We may also refer to calendar years in presenting external market data Revenue and addressable markets ▪Marvell reports revenue disaggregated in two broad end markets: Data Center and Communications and Other ▪During this presentation, we also provide additional information on revenue and addressable market trends by three businesses within Data Center: Interconnect, Switch and Storage, and Custom © 2026 Marvell. All rights reserved. 4
Slide 5 — Forward-looking statements
© 2026 Marvell. All rights reserved. 5 Forward-looking statements Except for statements of historical fact, this presentation contains forward-looking statements (within the meaning of the federal securities laws) including, but not limited to, statements related to market trends and to the company’s business and operations, business opportunities, growth strategy and expectations, and financial targets and plans, that involve risks and uncertainties. Words such as “anticipates,” “expects,” “intends,” “plans,” “projects,” “believes,” “seeks,” “estimates,” “can,” “may,” “will,” “would” and similar expressions identify such forward-looking statements. These statements are not guarantees of results and should not be considered as an indication of future activity or future performance. Actual events or results may differ materially from those described in this presentation due to a number of risks and uncertainties. For factors that could cause Marvell’s results to vary from expectations, please see the risk factors identified in Marvell’s Quarterly Report on Form 10-Q for the fiscal quarter ended August 1, 2026, as filed with the SEC on August 28, 2026, and other factors detailed from time to time in Marvell’s filings with the SEC. The forward‐looking statements in this presentation speak only as of the date of this presentation and Marvell undertakes no obligation to revise or update publicly any forward-looking statements.
Slide 6 — Non-GAAP financial measures
© 2026 Marvell. All rights reserved. 6 ▪During this presentation, we will refer to certain financial measures on a U.S. Non-GAAP basis. ▪We believe that the presentation of Non-GAAP financial measures provides important supplemental information to management and investors regarding financial and business trends relating to our financial condition and results of operations. ▪While we use Non-GAAP financial measures as a tool to enhance our understanding of certain aspects of our financial performance, we do not consider these measures to be a substitute for, or superior to, the information provided by GAAP financial measures. ▪Where Non-GAAP financial measures are presented, the most directly comparable financial measure calculated in accordance with GAAP and a reconciliation of the Non-GAAP financial measure to such GAAP measure are included in the Appendix to this presentation. Non-GAAP financial measures © 2026 Marvell. All rights reserved. 6
Slide 7 — Matt Murphy
Matt Murphy Chairman and CEO
Slide 8 — © 2026 Marvell. All rights reserved.
© 2026 Marvell. All rights reserved. 8
Slide 9 — © 2026 Marvell. All rights reserved.
© 2026 Marvell. All rights reserved. 9 © 2026 Marvell. All rights reserved.
Slide 10 — A strong point of view on where industry is going
© 2026 Marvell. All rights reserved. 10 2018 A strong point of view on where industry is going © 2026 Marvell. All rights reserved. 10
Slide 11 — A strong point of view on where industry is going
© 2026 Marvell. All rights reserved. 11 2020 A strong point of view on where industry is going 11 © 2026 Marvell. All rights reserved.
Slide 12 — A strong point of view on where industry is going
© 2026 Marvell. All rights reserved. 12 2021 A strong point of view on where industry is going 12 © 2026 Marvell. All rights reserved.
Slide 13 — A strong point of view on where industry is going
© 2026 Marvell. All rights reserved. 13 2024 A strong point of view on where industry is going 13 © 2026 Marvell. All rights reserved.
Slide 14 — A strong point of view on where industry is going
© 2026 Marvell. All rights reserved. 14 A strong point of view on where industry is going 14 © 2026 Marvell. All rights reserved. 2025
Slide 15 — © 2026 Marvell. All rights reserved.
© 2026 Marvell. All rights reserved. 15 Build organically Invest in Marvell’s existing technology foundation Strategic M&A Expand the platform through targeted M&A + One integrated platform, built and scaled inside Marvell Integrate + invest Integrate the platform around one strategic direction Architecting the Marvell platform © 2026 Marvell. All rights reserved. 15 ~$40B of investment in building and scaling the platform ~$20B in acquisitions ~$20B in subsequent investment Subsequent investment = Non-GAAP operating expenses + stock-based compensation from FY17 to FQ2’27
Slide 16 — AI workloads are reshaping the network
© 2026 Marvell. All rights reserved. 16 Greater connectivity demands AI workloads are reshaping the network Moving data efficiently is becoming a fundamental constraint on AI system performance Reasoning models Agentic AI Mixture of experts Inference at scale More data movement More diverse workloads 16 © 2026 Marvell. All rights reserved.
Slide 17 — Hyperscalers rearchitecting their networks
© 2026 Marvell. All rights reserved. 17 Hyperscalers rearchitecting their networks In the era of reasoning models and MoE, where any chip may need to talk to any other chip to route a token, this hop count matters.” “ Google Cloud April 2026 17 As we enter the era of frontier-scale intelligence, the economics of AI are increasingly determined not by peak compute performance but by how efficiently the system moves data.” Microsoft August 2026 “ The way networks had been built wasn't going to scale into the future and that something fundamentally different had to happen.” May 2026 “ AWS Meta Whether these clusters are training the next frontier model or serving inference at global scale, the network is in the critical path.” August 2026 “ It’s all about connectivity © 2026 Marvell. All rights reserved.
Slide 18 — Connectivity is the next defining unlock
© 2026 Marvell. All rights reserved. 18 Connectivity is the next defining unlock *Approximate revenue split Compute-first Revenue split* Connectivity Company C Company B Company A Compute 18 © 2026 Marvell. All rights reserved. Memory-first Revenue split* Memory Company C Company B Company A
Slide 19 — Connectivity is the next defining unlock
© 2026 Marvell. All rights reserved. 19 Connectivity is the next defining unlock *Approximate revenue split Compute-first Revenue split* Company C Company B Company A Memory-first Revenue split* Company C Company B Company A Connectivity Compute Memory Marvell Connectivity Compute Revenue split* Connectivity-first 19 © 2026 Marvell. All rights reserved.
Slide 20 — Marvell is a different kind of AI company
© 2026 Marvell. All rights reserved. 20 Marvell is a different kind of AI company 20 FY26 total revenue $8.2B 70% connectivity 30% compute Comms and other Interconnect Switch and storage Custom © 2026 Marvell. All rights reserved. *Approximate revenue split Marvell Connectivity Compute Revenue split* Connectivity-first
Slide 21 — Marvell connects AI infrastructure
© 2026 Marvell. All rights reserved. 21 21 Marvell connects AI infrastructure across every distance Fiber optics Up to 1,000km Distributed DCs Scale-across Up to 2km Data center Scale-out Scale-up Scale-in Copper © 2026 Marvell. All rights reserved. Switch and storage Custom Interconnect Rack | Multi-rack Up to 50m Up to 100mm Package
Slide 22 — Marvell connects AI infrastructure
© 2026 Marvell. All rights reserved. 22 22 Marvell connects AI infrastructure across every distance Up to 1,000km Distributed DCs Scale-across Up to 2km Data center Scale-out Scale-up Scale-in Copper © 2026 Marvell. All rights reserved. Switch and storage Custom Interconnect Rack | Multi-rack Up to 50m Fiber optics Up to 100mm Package
Slide 23 — Marvell connects AI infrastructure
© 2026 Marvell. All rights reserved. 23 23 Marvell connects AI infrastructure across every distance Up to 1,000km Distributed DCs Scale-across Up to 2km Data center Scale-out Scale-up Scale-in © 2026 Marvell. All rights reserved. Switch and storage Custom Interconnect Copper Fiber optics Up to 100mm Package Rack | Multi-rack Up to 50m
Slide 24 — One switching foundation for scale-out and up
© 2026 Marvell. All rights reserved. 24 One switching foundation for scale-out and up Scale-up switch New opportunity ▪Connecting accelerators within pods ▪Supporting every major protocol ▪Scales from copper to integrated optics 24 Scale-out switch Proven at hyperscale ▪Connecting AI clusters across pods ▪High-radix ethernet at cloud scale ▪Multiple generations deployed © 2026 Marvell. All rights reserved. Switch and storage Custom Interconnect
Slide 25 — Unmatched leadership in storage and memory
© 2026 Marvell. All rights reserved. 25 Custom solutions Proven IP + decades of expertise Unmatched leadership in storage and memory DRAM / memory expansion Dense SRAM Flash Hard drives Merchant / standard products SRAM technology Memory controllers/CXL NAND controllers HDD controllers 25 © 2026 Marvell. All rights reserved. Switch and storage Custom Interconnect Custom AI infrastructure Merchant technologies increasingly customized for hyperscaler-specific designs
Slide 26 — Dramatically expanding custom opportunity
© 2026 Marvell. All rights reserved. 26 Dramatically expanding custom opportunity 26 Multiple strategic agreements Custom silicon now shipping to the top four hyperscalers XPU attach is semi-custom Security Memory expansion Near-memory compute Storage NIC Management co-processor Inference accelerator © 2026 Marvell. All rights reserved. Switch and storage Custom Interconnect
Slide 27 — $10B
© 2026 Marvell. All rights reserved. 27 Custom revenue is accelerating FY29 custom target >2x FY29 vs. FY28 $10B
Slide 28 — >3x
© 2026 Marvell. All rights reserved. 28 Custom revenue is accelerating FY29 custom target >2x FY29 vs. FY28 $10B Updated FY29 custom target >3x FY29 vs. FY28 $12B+
Slide 29 — Data center CAPEX drives long-term growth
© 2026 Marvell. All rights reserved. 29 Data center CAPEX drives long-term growth ~$3T ~35% CAGR CY25-CY30 Data center CAPEX CY30 Marvell data center TAM CY30 ~$385B ~50% CAGR CY25-CY30 ~$1.6T Accelerated compute TAM CY30 ~48% CAGR CY25-CY30 Source: Industry estimates, public information
Slide 30 — $400B
© 2026 Marvell. All rights reserved. 30 Data center CAPEX drives long-term growth Marvell data center TAM CY30 ~$385B ~50% CAGR CY25-CY30 $400B ~45% CAGR CY25-CY30 Marvell total TAM CY30 $15B Comms and other TAM CY30 ~3% CAGR CY25-CY30 Source: Industry estimates, public information
Slide 31 — Growth across the stack
© 2026 Marvell. All rights reserved. 31 Growth across the stack 31 TAM CAGR CY25 – CY30 Industry growth drivers TAM CY30 Switching and storage ~40% Growth of scale-out New scale-up ~$85B Interconnect ~65% Growth of scale-out New scale-up and across ~$65B Custom ~55% Increase of XPU New XPU attach ~$235B Comms and other ~3% Modest growth ~$15B ~45% TOTAL ~$400B AI infrastructure scaling drives every layer © 2026 Marvell. All rights reserved. Source: Industry estimates, public information
Slide 32 — Our growth continues to accelerate
© 2026 Marvell. All rights reserved. 32 FY27E FY26 Our growth continues to accelerate FY28E $8.2B $12B 47% YoY $20B Total revenue 50% YoY
Slide 33 — $20B
© 2026 Marvell. All rights reserved. 33 FY27E FY26 Our growth continues to accelerate FY28E $8.2B $12B 47% YoY $20B FY28 updated revenue outlook $20B 67% YoY Total revenue
Slide 34 — 2024 | A bold data center target
© 2026 Marvell. All rights reserved. 34 Data center revenue 2024 | A bold data center target FY24 $2.2B FY29E $15B 2024 AI Era Event
Slide 35 — 2025 | Raising the data center target
© 2026 Marvell. All rights reserved. 35 2025 | Raising the data center target FY24 $2.2B FY29E ~$18B Data center revenue 2025 Custom AI Day
Slide 36 — Today | Data center target pulled in one year early
© 2026 Marvell. All rights reserved. 36 Today | Data center target pulled in one year early FY24 $2.2B FY29E ~$18B $20B FY28E Updated outlook Data center Data center revenue Investor Day 2026 Total revenue
Slide 37 — Today | Data center target pulled in one year early
© 2026 Marvell. All rights reserved. 37 ~$18B Today | Data center target pulled in one year early FY24 $2.2B FY28E Updated outlook ~70% DC CAGR Data center Data center revenue Investor Day 2026
Slide 38 — Setting our long-term outlook
© 2026 Marvell. All rights reserved. 38 $6.1B $70-90B FY31E Setting our long-term outlook 55-70% DC CAGR FY26 $8.2B $67.5- 87.5B ~$18B $20B ~70% DC CAGR $2.5B Total revenue Data center Comms and other FY28E
Slide 39 — Marvell long-term revenue target
© 2026 Marvell. All rights reserved. 39 FY26 $8.2B FY31E Marvell long-term revenue target 55-60% FY26-FY31 revenue CAGR FY31 total revenue $70-90B Total revenue
Slide 40 — What underpins the long-term outlook
© 2026 Marvell. All rights reserved. 40 Interconnect What underpins the long-term outlook Comms and other Switching and storage Custom Revenue CAGR FY26 – FY31 ~$37.5B Implied revenue (midpoint) FY31 ~$2.5B ~$10B ~$30B TAM CAGR CY25 – CY30 Marvell growth drivers 60-70% ~65% Leadership in scale-out / across Growing scale-up >3% ~3% Maintain position 40-50% ~40% Growing scale-out New scale-up opportunity 75-85% ~55% Ramping current XPU wins Rapid growth in XPU attach $70-90B FY31 total revenue target TOTAL 55-60% $70-90B ~45% Expanding AI infrastructure footprint
Slide 41 — Why invest in Marvell?
© 2026 Marvell. All rights reserved. 41 Why invest in Marvell? © 2026 Marvell. All rights reserved. $400B CY30 Marvell TAM forecast $12B+ FY29 updated custom revenue target $20B FY28 updated total revenue outlook $70-90B FY31 total revenue outlook EPS … Trusted by the world’s AI infrastructure leaders 41
Slide 42 — Chris Koopmans
Chris Koopmans President and COO
Slide 43 — Scaling the company to meet the opportunity
© 2026 Marvell. All rights reserved. 43 Scaling the company to meet the opportunity 43 Team | Process | Systems Supply chain Customers © 2026 Marvell. All rights reserved.
Slide 44 — Scaling the company to meet the opportunity
© 2026 Marvell. All rights reserved. 44 Scaling the company to meet the opportunity 44 Customers © 2026 Marvell. All rights reserved.
Slide 45 — AI buildout continues at massive scale and speed
© 2026 Marvell. All rights reserved. 45 $593B $1T AI buildout continues at massive scale and speed 2025 Custom AI Day 20% CAGR CY28 Source: Industry estimates, public information Total DC CAPEX CY25
Slide 46 — AI buildout continues at massive scale and speed
© 2026 Marvell. All rights reserved. 46 AI buildout continues at massive scale and speed Source: Industry estimates, public information $1.2T 73% YoY CY26 Update $711B CY25 Update Investor Day 2026 Total DC CAPEX
Slide 47 — AI buildout continues at massive scale and speed
© 2026 Marvell. All rights reserved. 47 AI buildout continues at massive scale and speed CY30 $711B $3.1T Source: Industry estimates, public information CY25 Update 35% CAGR Total DC CAPEX
Slide 48 — Large hyperscale data centers driving majority
© 2026 Marvell. All rights reserved. 48 Large hyperscale data centers driving majority 40% CAGR $711B Source: Industry estimates, public information All others Top 10 hyperscalers >60% >70% CY30 CY25 Update $3.1T Total DC CAPEX 35% CAGR
Slide 49 — Focus on large customers, deep partnerships
© 2026 Marvell. All rights reserved. 49 Focus on large customers, deep partnerships Data center operations at massive scale System development and deployment Workload development and management Advanced node silicon and packaging Hyperscalers Own the workload, the data center, and the system architecture Marvell The core IP partner that helps realize that vision High speed analog mixed signal connectivity IP Optical connectivity system-level design Extreme co-design
Slide 50 — Marvell customer journey spans full platform
© 2026 Marvell. All rights reserved. 50 Marvell customer journey spans full platform 50 Off-the-shelf system Custom system, standard silicon Semi-custom silicon Full custom silicon Merchant Custom Diverse product portfolio at every customer
Slide 51 — Cementing strategic partnerships
© 2026 Marvell. All rights reserved. 51 Long term, portfolio-wide, multi-generational Cementing strategic partnerships © 2026 Marvell. All rights reserved. 51
Slide 52 — Scaling the company to meet the opportunity
© 2026 Marvell. All rights reserved. 52 Scaling the company to meet the opportunity Team | Process | Systems © 2026 Marvell. All rights reserved. 52
Slide 53 — ~4X
© 2026 Marvell. All rights reserved. 53 Scaling engineering to meet opportunity © 2026 Marvell. All rights reserved. 53 Investing ahead of the data center opportunity Non-GAAP R&D $1.5B Data center revenue FY24 Data center revenue growth* $2.2B ~4X *4X growth rate based on annualized FY27 Q2 data center revenue
Slide 54 — Building the company for the long term
© 2026 Marvell. All rights reserved. 54 Building the company for the long term Research funding Curriculum Scholarships Internships 5 countries 14 universities Analog design Physical design Design for test 200+ scholarships 50+ universities >500 annually Majority convert Marvell has become a premier destination for talent
Slide 55 — Embracing AI to accelerate silicon engineering
© 2026 Marvell. All rights reserved. 55 Embracing AI to accelerate silicon engineering Faster chip design Accelerating product development cycle from concept to production Faster ramp to impact Accelerating new engineer development from day one to full speed Built on decades of Marvell expertise, engineered to scale © 2026 Marvell. All rights reserved. 55 Marvell Design Assistant Proprietary in-house design tool
Slide 56 — ~80%
© 2026 Marvell. All rights reserved. 56 Cloud-first design engineering mindset drives scale 2023 First announcement 2024-2025 Optimized E2E flow ~80% EDA infrastructure in the cloud 2026 Broad adoption © 2026 Marvell. All rights reserved. 56
Slide 57 — Scaling the company to meet the opportunity
© 2026 Marvell. All rights reserved. 57 Scaling the company to meet the opportunity Supply chain © 2026 Marvell. All rights reserved. 57
Slide 58 — Supply chain excellence at scale
© 2026 Marvell. All rights reserved. 58 Supply chain excellence at scale Relationship Track record Investment Results Business Intelligence Platform
Slide 59 — Scaling for growth
© 2026 Marvell. All rights reserved. 59 Scaling for growth Talent pipeline Supply chain relationships built on trust Deep customer relationships Building for the long term World-class team ready to deliver © 2026 Marvell. All rights reserved. 59
Slide 60 — Sandeep Bharathi
Sandeep Bharathi President, Data Center Group
Slide 61 — 2020 vision | Data infrastructure silicon provider
© 2026 Marvell. All rights reserved. 61 Design 28G SerDes for Wi-Fi, storage Process Fast follower Architecture Medium-size chips 2020 vision | Data infrastructure silicon provider Best IP on first shuttle with leading foundry Pivot to SerDes building blocks for data centers © 2026 Marvell. All rights reserved. 61 Large reticle-size chip 16nm 12nm ~200mm2 28G Source: 2020 Marvell Investor Day technology overview
Slide 62 — 2026 | Delivered and proven at every step
© 2026 Marvell. All rights reserved. 62 2026 | Delivered and proven at every step © 2026 Marvell. All rights reserved. Process First-to-node leadership 5nm 3nm 2nm 14Å (High volume, taped out, IP-ready) 100% A0-to-production 62 16nm 12nm 2020 2026 5nm 14Å Architecture Optimized compute for AI Industry’s largest reticle-size chips 10+ products >600mm2 (5/3/2nm) ~200mm2 2020 T100 Switch IC 2026 >850mm2 Design Best-in-class IP portfolio Leadership SerDes on every advanced node 28G 480 Gbps 2026 400G at 2nm 2020 Achieved Skipped 7nm
Slide 63 — AI tokenomics is a connectivity challenge
© 2026 Marvell. All rights reserved. 63 AI tokenomics is a connectivity challenge Cost per million output token Performance Tokenomics drivers Bandwidth Latency Power Reach/Distance © 2026 Marvell. All rights reserved. Performance refers to number of tokens allocated to a given user every second AI no longer only a silicon problem Data center connectivity challenge 63
Slide 64 — Memory wall stalls compute
© 2026 Marvell. All rights reserved. 64 Memory wall stalls compute © 2026 Marvell. All rights reserved. 64 Ideal world Memory Infinite memory Uniform latency High bandwidth Compute Memory wall is a bottleneck for performance as AI infrastructure scales Real-world memory wall XPU HBM HBM HBM Memory wall DRAM
Slide 65 — Crossing the memory wall
© 2026 Marvell. All rights reserved. 65 © 2026 Marvell. All rights reserved. 65 Crossing the memory wall Memory controllers Memory access CXL technologies Memory expander Photonic interconnects Memory fabric Power-efficient silicon Memory on a chip Memory as shared resource Rearchitecting systems to bring memory closer over scale-up fabric
Slide 66 — Optics enables multi-rack, scale-up domains
© 2026 Marvell. All rights reserved. 66 © 2026 Marvell. All rights reserved. 66 Optics enables multi-rack, scale-up domains Copper The limit Optical Tomorrow Across racks 1000+ XPUs No limit Today 1 rack Up to 144 XPUs Copper limit Scale-up represents >85%* of AI data center traffic Solving the scale-up bottleneck is the biggest gain *Internal company estimates
Slide 67 — Breaking the copper wall with photonics
© 2026 Marvell. All rights reserved. 67 © 2026 Marvell. All rights reserved. 67 Breaking the copper wall with photonics Board-level, laser-free µLED Short-reach, high-density µVCSEL Ultra-low power, high density MRM High-speed, long-reach MZM, TFLN Ultra-fast, ultra-compact Plasmonics, SiPho All-optical scale-up fabric EAM Pursuing every optical path, across every distance 3D integrated silicon photonics light engine Integration on reverse side Electronics Photonics Substrate More data movement drives more optical content at scale
Slide 68 — Common technology foundation
© 2026 Marvell. All rights reserved. 68 Common technology foundation Connectivity across every reach Advanced packaging Photonics Advanced node and process Analog mixed signal
Slide 69 — Going beyond CMOS
© 2026 Marvell. All rights reserved. 69 69 Going beyond CMOS Old world: Silicon CMOS All ICs New world: Heterogeneous integration CMOS + SiPho + SiGe + TFLN in Light Engine Silicon Germanium SiGe Plasmonics Indium Phosphide InP CMOS Lithium Niobate TFLN Silicon Photonics SiPho Co-integrated in system and in package © 2026 Marvell. All rights reserved. Fully integrated, process agnostic, foundry ready and future proof
Slide 70 — Analog mixed signal powers data movement engine
© 2026 Marvell. All rights reserved. 70 70 Analog mixed signal powers data movement engine Chip I/O Move data in / out of a chip Die-to-Die Connect multiple chiplets on one package Complex System Multiple chips and packages combine into full system 400G per lane on leading node 8x growth in data rate in 6 years IP + silicon Industry-leading asset End-to-end electrical and optical I/O Security Memory expansion Near-memory compute Storage NIC Management co-processor Inference accelerator © 2026 Marvell. All rights reserved. Marvell SerDes leadership for every reach
Slide 71 — Packaging innovation for photons and electrons
© 2026 Marvell. All rights reserved. 71 Packaging innovation for photons and electrons © 2026 Marvell. All rights reserved. 71 Full integration and co-design to scale in volume CPO 2D package 1 die Substrate Lid Silicon 2D package 2 die Silicon Substrate Lid 2.5D package >8 die Silicon chiplets Interposers 3.5D package >8 die Stacked silicon chiplets
Slide 72 — Industry-leading core technology
© 2026 Marvell. All rights reserved. 72 Industry-leading core technology Process and packaging Proven execution, deployed at scale IP and architecture System-level solutions Security Memory expansion Near-memory compute Storage NIC Management co-processor Inference accelerator © 2026 Marvell. All rights reserved. 72
Slide 73 — Tokenomics is physics
© 2026 Marvell. All rights reserved. 73 Marvell is the trusted and undisputed leading solution provider Tokenomics is physics © 2026 Marvell. All rights reserved. 73
Slide 74 — Xi Wang
Xi Wang SVP and GM, Connectivity Business
Slide 75 — The broadest end-to-end AI connectivity portfolio
© 2026 Marvell. All rights reserved. 75 75 Marvell is a connectivity-first company The broadest end-to-end AI connectivity portfolio Up to 2km Data center Rack | Multi-rack Up to 50m Scale-out Scale-up Scale-in Up to 1,000km Distributed DCs Scale-across LED, VCSEL Optical I/O chiplet PAM DSP ACC, NPO, CPO, VCSEL TIA | Driver PAM DSP Coherent-lite DSP TIA | Driver AEC DSP Coherent modules Coherent DSP TIA | Driver © 2026 Marvell. All rights reserved. Up to 100mm Package
Slide 76 — Industry-first technology from the ground up
© 2026 Marvell. All rights reserved. 76 Broadband SiGe/GaAs/InP IPs around optics TIA Driver High-speed analog Industry-first technology from the ground up Broadest portfolio of optical technologies MZM, MRM, EAM, silicon photonics, μLED, μVCSEL, plasmonics Photonics First to advanced node First to A0 production PAM4, PAM6, PAM8 coherent, coherent-lite DSP, EIC, SerDes Advanced node DSP Built on decades of investment across every key connectivity technology
Slide 77 — TIAs and Drivers required for every optical interconnect
© 2026 Marvell. All rights reserved. 77 TIAs and Drivers required for every optical interconnect Performance rules Lowest noise, high gain, high linearity Dozens of specialized products every year Pairs with wide variety of optics Expanding into new form factors Pluggables transceivers → NPO, CPO 77 © 2026 Marvell. All rights reserved. DSP Driver Modulator TIA PDs Analog SerDes and Digital Photonics Laser What’s inside an optical module?
Slide 78 — Marvell high-speed analog foundation
© 2026 Marvell. All rights reserved. 78 Marvell high-speed analog foundation ▪Multi-foundry process leadership ▪Broadband GaAs, SiGe, and InP ▪30-40 tape-outs per year ▪Optimized for every platform Technology innovation ▪First to every generation ▪Both PAM and coherent pluggable platforms ▪Enabling next-gen NPO ▪At every speed, every distance Product leadership © 2026 Marvell. All rights reserved. 78 25+ years of leadership, powering every generation
Slide 79 — Multi-generational DSP platform leadership
© 2026 Marvell. All rights reserved. 79 DSP, TIA, Driver Pioneering connectivity solutions, first to every generation Multi-generational DSP platform leadership 400G 7nm PAM DSP, 4x100G 100G/Lane TIA, Driver 800G 5nm PAM DSP, 8x100G 100G/Lane TIA, Driver 1.6T 3nm PAM DSP, 8x200G 200G/Lane TIA, Driver 200G 16nm PAM DSP, 4x50G 50G/Lane TIA, Driver
Slide 80 — Early co-definition with hyperscalers
© 2026 Marvell. All rights reserved. 80 Early co-definition with hyperscalers 80 Addressing the scaling challenges of tomorrow with customers Solution defined years ahead Build-in unique features for key hyperscalers Collaboration across the optics ecosystem Production ramp ready from design phase
Slide 81 — Build optimized solution for diversifying AI workload
© 2026 Marvell. All rights reserved. 81 Ara 1.6T 3nm PAM DSP Foundation Shipping now Build optimized solution for diversifying AI workload Optimized 1.6T solutions, specialized for AI, executed in parallel 1.6T 3nm PAM DSP Ara X Optimized for Robustness Shipping now Ara T 1.6T 3nm PAM TRO DSP Optimized for Power Shipping now 1.6T 3nm PAM MACsec DSP Ara M Optimized for Security Coming soon 1.6T 3nm PAM Gearbox DSP Petra Optimized for Flexibility Shipping now 81
Slide 82 — Scaling beyond 1.6T
© 2026 Marvell. All rights reserved. 82 Scaling beyond 1.6T 3.2T solution Be more power efficient Link robustness for AI Scalable for volume deployment Maintain pluggable form factor
Slide 83 — Addressing the complexities of 3.2T
© 2026 Marvell. All rights reserved. 83 © 2026 Marvell. All rights reserved. Marvell® 3.2T optical DSP solution Marvell developing both technologies for complete infrastructure needs Addressing the complexities of 3.2T 2nm Coherent-lite DSP 2nm PAM DSP 83
Slide 84 — Marvell industry-first 2nm live demos at ECOC
© 2026 Marvell. All rights reserved. 84 8x400G Optical Module Using PAM 448G SerDes Marvell industry-first 2nm live demos at ECOC 84 © 2026 Marvell. All rights reserved. Path to 3.2T and beyond 3.2T PAM4 Refreshing the definition of bandwidth 3.2T Coherent-lite Data center and campus reach, balanced power, and cost 2x1.6T Coherent-lite Module 1600GE Coherent-lite DSP 84
Slide 85 — 6.4T
© 2026 Marvell. All rights reserved. 85 The path to 3.2T and beyond 200G 400G 800G 1.6T DSP, TIA, Driver 2nm 3.2T Coherent-lite 2nm 3.2T PAM 3.2T 6.4T Marvell continues to lead the speed and bandwidth for every transition
Slide 86 — Next-gen chip-to-chip interconnect for scale-in
© 2026 Marvell. All rights reserved. 86 86 Next-gen chip-to-chip interconnect for scale-in Up to 2km Data center Rack | Multi-rack Up to 50m Scale-out Scale-up Up to 1,000km Distributed DCs Scale-across LED, VCSEL Optical I/O chiplet PAM DSP ACC, NPO, CPO, VCSEL TIA | Driver PAM DSP Coherent-lite DSP TIA | Driver AEC DSP Coherent modules Coherent DSP TIA | Driver © 2026 Marvell. All rights reserved. Scale-in Up to 100mm Package
Slide 87 — Extending optics native to compute for scale-in
© 2026 Marvell. All rights reserved. 87 Extending optics native to compute for scale-in Today: Copper I/O ASIC Electrical I/O SerDes I/O by CMOS SerDes Reach: ~10mm Future: Optical I/O New light sources, μLED, μVCSEL Reach: 1.5m+ ASIC Optical I/O chiplet Enabling scale-in architecture with optical chiplet integration
Slide 88 — Marvell is a connectivity-first company
© 2026 Marvell. All rights reserved. 88 Marvell is a connectivity-first company 88 Most advanced technology Leading every hop, every distance, every generation Deep customer engagements Architected to scale for AI volume © 2026 Marvell. All rights reserved.
Slide 89 — Russ Esmacher
Russ Esmacher SVP and GM, Data Center Interconnect Business
Slide 90 — Marvell connects AI infrastructure
© 2026 Marvell. All rights reserved. 90 90 Leadership across the full connectivity stack Marvell connects AI infrastructure across every distance Up to 1,000km Distributed DCs Up to 2km Data center Up to 100mm Package Rack | Multi-rack Up to 50m Scale-across Scale-out Scale-up Scale-in © 2026 Marvell. All rights reserved.
Slide 91 — DCI market tailwinds
© 2026 Marvell. All rights reserved. 91 DCI market tailwinds More fiber capacity Line system density Expanding AI clusters accelerating demand for coherent pluggables Transition from boxes to pluggables Power efficiency 15x bandwidth vs. traditional DCI Scale-across Emerging Up to 1,000km AI cluster to AI cluster Data center densification Space and power constraints drives more data centers © 2026 Marvell. All rights reserved. 91
Slide 92 — Marvell led the pluggable transition a decade ago
© 2026 Marvell. All rights reserved. 92 2026 COLORZ® 1600 2024 COLORZ® 800 Marvell led the pluggable transition a decade ago Transport box DCI box 92 2016 COLORZ® 100 Marvell coherent pluggables Previous © 2026 Marvell. All rights reserved. First at every generation, and the pace is accelerating 2020 COLORZ® 400
Slide 93 — 15x
© 2026 Marvell. All rights reserved. 93 15x Bandwidth vs. DCI Scale-across connects AI clusters Accelerating the adoption of 1.6T and 3.2T AI cluster AI cluster Up to 1,000km
Slide 94 — Two paths to market
© 2026 Marvell. All rights reserved. 94 Two paths to market Coherent modules Enabling hyperscalers with Marvell® COLORZ® modules Coherent DSP Enabling broader market with Marvell DSPs powering partner modules Marvell fulfills accelerating demand with modules and DSPs © 2026 Marvell. All rights reserved. 94
Slide 95 — Scaling coherent optics for high volume
© 2026 Marvell. All rights reserved. 95 Scaling coherent optics for high volume Anatomy of a Marvell coherent optics module Coherent DSP DAC ADC Driver TIA PDs Optical frontend Modulator Electronics ASIC Photonics Laser Marvell core competencies developed over 20+ years Deep photonics expertise MZM, SiPho, TFLN, plasmonics High-volume manufacturability >1M units >15B field hours Largest coherent DSP team First to every node Simultaneous chips Highest speed analog Leveraging Marvell scale Advanced SerDes Key IP and depth First to every node TIA DRIVER
Slide 96 — 3.2T
© 2026 Marvell. All rights reserved. 96 ZR 14Å Plasmonics ZR+ 14Å Plasmonics Coherent-lite 14Å 3.2T ZR 2nm TFLN ZR+ 2nm TFLN Coherent-lite 2nm 1.6T Multi-generational scale-across platform leadership 400G ZR 7nm SiPho 100G ZR 28nm SiPho Multiple DSPs, CMOS node leader, more advanced photonics ZR 5nm SiPho ZR+/MACsec 2nm SiPho 800G
Slide 97 — Marvell leading next gen at 2nm
© 2026 Marvell. All rights reserved. 97 Marvell leading next gen at 2nm Industry-first 2nm 1.6T ZR/ZR+ extending links to 1,000km+ COLORZ 1600 Powered by Electra DSP 97 © 2026 Marvell. All rights reserved. Industry-first 2nm 800G ZR/ZR+ with built-in MACsec encryption COLORZ 800 Powered by Libra DSP
Slide 98 — Marvell industry-first 2nm live demos at ECOC
© 2026 Marvell. All rights reserved. 98 Marvell industry-first 2nm live demos at ECOC 98 © 2026 Marvell. All rights reserved. 1.6T ZR/ZR+ More bandwidth, lower power, longer reach 2nm 1.6T Coherent DSP for ZR/ZR+ Pluggable Optics 800ZR/ZR+ with MACsec Secure AI traffic, lower cost, simpler operations 800G ZR/ZR+ Pluggable Optics with MACSec
Slide 99 — Leading scale-across with coherent pluggables
© 2026 Marvell. All rights reserved. 99 Leading scale-across with coherent pluggables © 2026 Marvell. All rights reserved. 99 Market leader Supplying to top 5 hyperscalers Leadership through delivering innovation at scale First to market Leveraging core technologies Delivering at scale >1M shipped
Slide 100 — Dave Lazovsky
Dave Lazovsky EVP and GM, Data Center Networking Business
Slide 101 — Marvell Data Center Networking
© 2026 Marvell. All rights reserved. 101 Cloud Optics Pluggable transceivers and coherent DCI 101 © 2026 Marvell. All rights reserved. Marvell Data Center Networking Integrated Optics NPO and CPO Scale-out and scale-up Switching
Slide 102 — Established switching portfolio, accelerating momentum
© 2026 Marvell. All rights reserved. 102 Ethernet ESUN 102 Established switching portfolio, accelerating momentum Proven multi-generational execution Silicon die Line card and package Liquid-cooled chassis 100T BGA/CPC/NPO/CPO Scale-out and Scale-up 12.8T 400T Switch bandwidth 51T 200T 100T
Slide 103 — Established switching portfolio, accelerating momentum
© 2026 Marvell. All rights reserved. 103 Ethernet ESUN UALink 103 Established switching portfolio, accelerating momentum Proven multi-generational execution 12.8T 400T Switch bandwidth 51T 200T 100T Customer deployment at scale | Comprehensive switch portfolio extending to 200T, 400T, and UAL
Slide 104 — The industry’s only portfolio of all switch protocols
© 2026 Marvell. All rights reserved. 104 NVLink UALink Ethernet / ESUN © 2026 Marvell. All rights reserved. 104 Logical layer: Protocol The industry’s only portfolio of all switch protocols Marvell network solutions Protocol agnostic | Tailored to each customer Marvell: One partner across every protocol
Slide 105 — Marvell connects AI infrastructure
© 2026 Marvell. All rights reserved. 105 105 Marvell connects AI infrastructure across every distance Up to 1,000km Distributed DCs Up to 2km Data center Up to 100mm Package © 2026 Marvell. All rights reserved. Rack | Multi-rack Up to 50m Leadership across the full connectivity stack Scale-across Scale-out Scale-up Scale-in
Slide 106 — AI workloads drive new infrastructure demands
© 2026 Marvell. All rights reserved. 106 AI workloads drive new infrastructure demands ▪Larger models ▪Compute and network intensity ▪Higher memory capacity and bandwidth ▪More compute per query ▪Low-latency scale-up networks ▪Higher memory capacity ▪More concurrent interactions ▪Larger KV cache ▪Greater scale-up domains ▪Dynamic all-to-all traffic ▪Higher connectivity bandwidth ▪More tokens per dollar Foundation models at scale Reasoning + inference-time compute MoE + long context Agentic AI + AI adoption at scale 106 © 2026 Marvell. All rights reserved. Explosion of technology needs: Compute | Memory | Scale-up networking | Connectivity reach
Slide 107 — Reasoning at scale creates a memory challenge
© 2026 Marvell. All rights reserved. 107 Reasoning at scale creates a memory challenge © 2026 Marvell. All rights reserved. 107 KV cache grows with context Capacity increases Bandwidth increases Latency sensitivity increases Reload or Recompute (and evict to make room) or Reload Recompute Evict Impact on system efficiency Higher latency More memory traffic or recompute Lower efficiency Required KV entry is not in fast memory Required KV entry is not in fast memory Longer, more complex context Prompt Reasoning Response Continue reasoning…
Slide 108 — Memory scaling beyond the server and rack
© 2026 Marvell. All rights reserved. 108 XPU / server Direct-attach memory (SRAM / HBM) Fastest access Capacity constrained within package and server Rack Memory-semantic network More capacity Scale-up network expands memory capacity Pod / multi-rack Larger scale-up domains Pod-scale access Memory capacity expands across multiple racks Memory scaling beyond the server and rack © 2026 Marvell. All rights reserved. 108 Memory architectures scale with inference demand Scaling memory beyond the server requires a high-bandwidth, low-latency scale-up network
Slide 109 — Scaling beyond the rack requires optics
© 2026 Marvell. All rights reserved. 109 Scaling beyond the rack requires optics © 2026 Marvell. All rights reserved. 109 Beyond the XPU package and the server Inference-time compute / reasoning models require access to more memory Beyond the rack The network must extend beyond the rack to access more memory Copper Reach limit Optical Memory can only scale if the network scales with it Longer reach and lower power Lower latency Higher bandwidth Multi-rack scale-up networks require advanced optics and high-radix, single-hop switching
Slide 110 — Scale-up requirements drive optics closer to compute
© 2026 Marvell. All rights reserved. 110 10x bandwidth density Supporting HBM bandwidth Low latency Approach HBM load latency Low power Minimize data movement power Reach Extend across multiple racks Scale-up requirements drive optics closer to compute XPU 1 accesses memory in XPU 2 through the scale-up network XPU 1 + HBM XPU 2 + HBM Equivalent to roughly 10 × 1.6T links Scale-up network requirements Scale-up bandwidth per XPU ≥16 Tbps Traditional pluggables cannot efficiently support scale-up bandwidth density, latency and power Pluggable Front-panel optics NPO Optics adjacent to the package CPO Optics co-packaged with XPU
Slide 111 — Hyperscale data center optical scale-up networks
© 2026 Marvell. All rights reserved. 111 Hyperscale data center optical scale-up networks 200G Light Engine NPO / CPO MRM / OCI MSA NPO / CPO Photonic Fabric TM NPO / CPO 400G Light Engine NPO / CPO 200G NPO SiGe Drivers / TIAs Physical layer: Copper Optics © 2026 Marvell. All rights reserved. 111 Bandwidth 1.6T - 6.4T 6.4T 25.6T 12.8T 16T - 32T BW / fiber 200G DR 200G DR 400G DR/1.6T (FR4) 400G (OCI MSA) 900G Timing Now Now 2027 2027 2027 Module Providers Marvell ”Fast Pipe” Light Engines Marvell “Flat Pipe” Light Engines A comprehensive optical portfolio for scale-up architectures
Slide 112 — 200G / 400G
© 2026 Marvell. All rights reserved. 112 200G / 400G Light Engines © 2026 Marvell. All rights reserved. 112
Slide 113 — 200G / 400G
© 2026 Marvell. All rights reserved. 113 200G / 400G Light Engines 113
Slide 114 — 200G / 400G
© 2026 Marvell. All rights reserved. 114 © 2026 Marvell. All rights reserved. 114 200G / 400G Light Engines
Slide 115 — Photonic FabricTM / OCI MSA
© 2026 Marvell. All rights reserved. 115 Photonic FabricTM / OCI MSA 115 © 2026 Marvell. All rights reserved.
Slide 116 — Photonic FabricTM / OCI MSA
© 2026 Marvell. All rights reserved. 116 Photonic FabricTM / OCI MSA © 2026 Marvell. All rights reserved. 116
Slide 117 — Photonic FabricTM / OCI MSA
© 2026 Marvell. All rights reserved. 117 Photonic FabricTM / OCI MSA © 2026 Marvell. All rights reserved. 117
Slide 118 — Scale-up networks co-engineered end to end
© 2026 Marvell. All rights reserved. 118 Scale-up networks co-engineered end to end © 2026 Marvell. All rights reserved. 118 Signal integrity, low power and latency Design for link performance Accelerating time to token generation Architect for rapid bring-up Link management and network security Operate with confidence XPU Compute side I/O chiplet XPU Compute side I/O chiplet Physical layer Protocol Physical layer Protocol Scale-up switch I/O chiplet I/O chiplet Faster time to market with lower risk through deployment, operations, and support
Slide 119 — Customer operational priorities shape architecture
© 2026 Marvell. All rights reserved. 119 Lower failure blast radius Field-replaceable optics Flexible and configurable © 2026 Marvell. All rights reserved. 119 Different priorities Different architectures One goal: optimize the system for each customer Customer operational priorities shape architecture NPO: Near-package optics Physical layer: Copper Optics
Slide 120 — Customer operational priorities shape architecture
© 2026 Marvell. All rights reserved. 120 © 2026 Marvell. All rights reserved. 120 CPO: Co-packaged optics Customer operational priorities shape architecture Different priorities Different architectures One goal: optimize the system for each customer Lowest power and latency Greater integration and reduced system complexity Optimized for maximum integration at scale Physical layer: Copper Optics
Slide 121 — Shared memory beyond the XPU
© 2026 Marvell. All rights reserved. 121 © 2026 Marvell. All rights reserved. Shared memory beyond the XPU High-bandwidth, low-latency access to shared memory across the scale-up domain 121 Marvell® Photonic Fabric TM Memory Module Photonic Fabric TM Memory Appliance
Slide 122 — The trusted partner for
© 2026 Marvell. All rights reserved. 122 © 2026 Marvell. All rights reserved. 122 Scale-out Switching Growing to >$1B in FY28 Multi-billion $ annual trajectory The trusted partner for scale-out, scale-up switching and optics solutions Scale-up Switching Strong customer traction Additional multi-billion $ annually Scale-up Optics Significant Marvell content Revenue opportunity substantially greater than scale-up switching
Slide 123 — Will Chu
Will Chu EVP and GM, Custom Cloud Solutions Business
Slide 124 — Multiple multi-generational wins at top 4 hyperscalers
© 2026 Marvell. All rights reserved. 124 Multiple multi-generational wins at top 4 hyperscalers © 2026 Marvell. All rights reserved. 124 Custom Cloud Solutions Marvell core IP XPU XPU attach
Slide 125 — XPUs growing in complexity to scale AI performance
© 2026 Marvell. All rights reserved. 125 XPUs growing in complexity to scale AI performance 1-2 die | 4-6 HBM N N+1 4-8 die | 12 HBM | 2-4 copper I/O N+2 8+ die | 12+ HBM | 4+ optical I/O © 2026 Marvell. All rights reserved. 125 Connectivity IP is the catalyst for the next generation of XPUs Marvell critical technologies SerDes Die-to-die Advanced packaging Custom SRAM Custom HBM PIVR CPO NPO 3D DRAM NVLink Fusion
Slide 126 — Custom opportunity expanding rapidly
© 2026 Marvell. All rights reserved. 126 Custom opportunity expanding rapidly Standard product IP leadership enabling semi-custom opportunities Traditional data centers Standard / merchant products NIC Security Memory Management CPU AI data centers Custom / semi-custom products Security Memory expansion Near-memory compute Storage NIC Inference accelerator Management co-processor
Slide 127 — AI workloads outgrow off-the-shelf NICs
© 2026 Marvell. All rights reserved. 127 AI workloads outgrow off-the-shelf NICs ▪Foundational NICs are prevalent ▪SmartNICs with integrated CPUs ▪I/O speed <100G Marvell® OCTEON® SmartNICs Traditional data center ▪Specific I/O speed, BW and config ▪Optimized data path, accelerators ▪Custom network protocols ▪Optimized integrated CPU AI challenges Marvell® Custom NIC Marvell NIC complexity increasing due to higher performance CPUs and XPUs
Slide 128 — Memory expansion and near-memory compute
© 2026 Marvell. All rights reserved. 128 ▪CPU-bound workloads ▪Direct-attach memory sufficient ▪Marvell early standard CXL effort Traditional data center Memory expansion and near-memory compute ▪Inference and KV cache workloads ▪Memory-bound, latency-sensitive ▪CPU and XPU capacity and BW ▪Memory shortage AI challenges Marvell Marvell® Structera Custom Memory expansion extending beyond CPUs to XPUs
Slide 129 — AI storage demands semi-custom silicon
© 2026 Marvell. All rights reserved. 129 AI storage demands semi-custom silicon ▪Merchant controllers for SSDs ▪Merchant accelerators for offload Traditional data center ▪Multi-gen FW / SW control and multi-vendor NAND flexibility ▪Storage offload performance required for AI responsiveness AI challenges Marvell® Bravera SSD controllers Marvell ▪Flash-agnostic storage controllers ▪AI-optimized storage accelerators Storage capacity and offload requirements are rising rapidly
Slide 130 — A new class of AI infrastructure management silicon
© 2026 Marvell. All rights reserved. 130 A new class of AI infrastructure management silicon ▪BMC manages standard servers ▪~150 sensors, 100s KB/day ▪Minimal processor and I/O perf Traditional data center ▪Extremely complex AI servers ▪>1500 sensors, 100s MB/day ▪~1000X more performance AI challenges Marvell AI infrastructure management AI system management needs high-performance multi-core processors
Slide 131 — Inference offload opens a new XPU attach market
© 2026 Marvell. All rights reserved. 131 Inference offload opens a new XPU attach market ▪Inference architectures evolving ▪Need to offload XPUs for inference ▪Specialized co-processors attached to XPUs Next-gen AI challenges Marvell Custom SRAM SerDes Die-to-die Adv. package Well-positioned to win in exciting new XPU attach market
Slide 132 — Significant custom opportunity
© 2026 Marvell. All rights reserved. 132 Significant custom opportunity 132 More customers More attach categories More XPUs Higher attach rate Marvell opportunity Building a balanced portfolio © 2026 Marvell. All rights reserved. Higher ASPs
Slide 133 — XPU and XPU attach driving growth
© 2026 Marvell. All rights reserved. 133 XPU and XPU attach driving growth © 2026 Marvell. All rights reserved. 133 Adding XPU attach to growing XPU business for greater diversity Customers, team, and platform to deliver on the opportunity $12B+ in FY29 Updated forecast $30B+ in FY31 at midpoint target, based on design win diversity
Slide 134 — Dan Durn
Dan Durn EVP and CFO
Slide 135 — Long-term model execution FY21 → FY26
© 2026 Marvell. All rights reserved. 135 © 2026 Marvell. All rights reserved. 135 Long-term model execution FY21 → FY26 © 2026 Marvell. All rights reserved. 135 *Non-GAAP R&D; Capital allocation represents FY21-FY26 cumulative investment Revenue accelerated Capital allocation Op income expanded 23% CAGR vs. 15–20% target set in FY21 32% CAGR Delivered earnings leverage Disciplined capital allocation >$8B Organic investment* ~$11B M&A $4B Return to shareholders We exceeded our targets while adding capabilities
Slide 136 — What you have heard today
© 2026 Marvell. All rights reserved. 136 © 2026 Marvell. All rights reserved. 136 What you have heard today Technology leadership Market opportunity ~$400B TAM Revenue opportunity $70-90B Purpose built to deliver significant growth ~45% CAGR CY25-CY30 55-60% CAGR FY26-FY31 DSP and optical Analog mixed signal Silicon and packaging
Slide 137 — What’s in front of us
© 2026 Marvell. All rights reserved. 137 © 2026 Marvell. All rights reserved. 137 What’s in front of us Faster growth, expanding margin, and rising cash generation Revenue Revenue growing faster than market 55-60% 5-year CAGR FY26-FY31 Earnings Faster than revenue Operating leverage at scale Cash flow Funds growth and shareholder returns Strong and rising
Slide 138 — Faster revenue growth, more data center
© 2026 Marvell. All rights reserved. 138 FY26 FY31 Faster revenue growth, more data center $8.2B $70–90B © 2026 Marvell. All rights reserved. 138 55-60% CAGR Total revenue Data center Comms and other Data center ~65% Rev CAGR FY26–FY31 Comms and other >3% Rev CAGR FY26–FY31
Slide 139 — Data center growth is broad-based
© 2026 Marvell. All rights reserved. 139 © 2026 Marvell. All rights reserved. 139 Data center growth is broad-based Three large and diversified high-growth engines Interconnect ~$65B TAM in FY31 ~65% FY26-FY31 revenue CAGR* Switching and storage ~$85B TAM in FY31 Custom ~$235B TAM in FY31 ~45% FY26-FY31 revenue CAGR ~80% FY26-FY31 revenue CAGR *FY26-FY31 midpoint of revenue CAGR
Slide 140 — Significant operating leverage
© 2026 Marvell. All rights reserved. 140 Significant operating leverage Gross Margin Influenced by product mix Growth in custom Operating Margin Increases with scale OpEx to rise at ~half the % rate of revenue growth Earnings compound faster than revenue © 2026 Marvell. All rights reserved. 140
Slide 141 — Capital allocation priorities
© 2026 Marvell. All rights reserved. 141 Capital allocation priorities Organic growth Talent, IP, and supply to fund the roadmap Strategic M&A Proven track record of acquiring technology and teams Return capital >50% of free cash flow, on average over time Funding growth and delivering returns while maintaining a strong balance sheet
Slide 142 — >$30
© 2026 Marvell. All rights reserved. 142 Long-term Non-GAAP financial model >$30 Metric FY31 target model Revenue $70-90B Gross Margin 56-59% Operating Margin 44-46% Tax Rate 15% Free Cash Flow >36% FY31 implied EPS
Slide 143 — Why invest in Marvell?
© 2026 Marvell. All rights reserved. 143 $400B CY30 Marvell TAM forecast $12B+ FY29 updated custom revenue target $20B FY28 updated total revenue target $70-90B FY31 total revenue target >$30 FY31 implied Non-GAAP EPS Why invest in Marvell? © 2026 Marvell. All rights reserved. Trusted by the world’s AI infrastructure leaders 143
Slide 144
Slide 145 — Key Terms
© 2026 Marvell. All rights reserved. 145 © 2026 Marvell. All rights reserved. 145 Key Terms
Slide 146 — Key terms
© 2026 Marvell. All rights reserved. 146 Key terms ▪ ACC: Active copper cable ▪ AEC: Active electrical cable ▪ ADC: Analog-to-digital converter ▪ BMC: Baseboard management controller ▪ CPO: Co-packaged optics ▪ CMOS: Complementary metal-oxide-semiconductor ▪ CXL: Compute Express Link ▪ DCI: Data center interconnect ▪ DAC: Digital-to-analog converter ▪ DSP: Digital signal processor ▪ DRAM: Dynamic random-access memory ▪ EAM: Electro-absorption modulator ▪ EIC: Electric interface chip ▪ ESUN: Ethernet for scale-up networking ▪ G: Gigabits per second ▪ GaAs: Gallium Arsenide ▪ HBM: High-bandwidth memory ▪ InP: Indium Phosphide ▪ KV cache: Key-value cache ▪ MACsec: Media access control security ▪ LED: Micro-light-emitting diode ▪ MZM: Mach-Zehnder modulator ▪ MRM: Micro-ring Modulator ▪ MoE: Mixture-of-experts ▪ MSA: Multi-supplier agreement ▪ VCSEL: Micro vertical-cavity surface-emitting laser ▪ NPO: Near-packaged optics ▪ NIC: Network interface controller ▪ OCI: Optical compute interconnect ▪ PD: Photo-diode ▪ PAM: Pulse-amplitude modulation ▪ SerDes: Serializer-deserializer ▪ SiGe: Silicon Germanium ▪ SiPho: Silicon photonics ▪ SRAM: Static random-access memory ▪ T: Terabits per second ▪ TFLN: Thin-film lithium niobate ▪ TIA Transimpedance amplifier
Slide 147 — Appendices
© 2026 Marvell. All rights reserved. 147 © 2026 Marvell. All rights reserved. 147 Appendices Reconciliations from GAAP to Non-GAAP
Slide 148 — Discussion of Non-GAAP financial measures
© 2026 Marvell. All rights reserved. 148 Non-GAAP financial measures exclude the effect of stock-based compensation expense, amortization of acquired intangible assets, acquisition and divestiture related costs, restructuring and other related charges (gains), (including, but not limited to, changes in contractual obligations, employee severance costs, and facility exit related charges), change in fair value of contingent consideration liability and forward stock purchase contract, resolution of legal matters, and certain expenses and benefits that are driven primarily by discrete events that management does not consider to be directly related to Marvell’s core business. Although Marvell excludes the amortization of all acquired intangible assets from these Non-GAAP financial measures, management believes that it is important for investors to understand that such intangible assets were recorded as part of purchase price accounting arising from acquisitions, and that such amortization of intangible assets that relate to past acquisitions will recur in future periods until such intangible assets have been fully amortized. Investors should note that the use of intangible assets contributed to Marvell’s revenues earned during the periods presented and are expected to contri bute to Marvell’s future period revenues as well. Marvell uses a Non-GAAP tax rate to compute the Non-GAAP tax provision. This Non-GAAP tax rate is based on Marvell’s estimated annual GAAP income tax forecast, adjusted to account for items excluded from Marvell’s Non-GAAP income, as well as the effects of significant non-recurring and period specific tax items which vary in size and frequency, and excludes tax deductions and benefits from acquired tax loss and credit carryforwards and changes in valuation allowance on acquired deferred tax assets. Marvell’s Non-GAAP tax rate is determined on an annual basis and may be adjusted during the year to take into account events that may materially affect the Non-GAAP tax rate such as tax law changes; acquisitions; significant changes in Marvell’s geographic mix of revenue and expenses; or changes to Marvell’s corporate structure. For the second quarter of fiscal 2027, a Non- GAAP tax rate of 11.0% has been applied to the Non-GAAP financial results. Marvell believes that the presentation of Non-GAAP financial measures provides important supplemental information to management and investors regarding financial and business trends relating to Marvell’s financial condition and results of operations. While Marvell uses Non-GAAP financial measures as a tool to enhance its understanding of certain aspects of its financial performance, Marvell does not consider these measures to be a substitute for, or superior to, financial measures calculated in accordance with GAAP. Consistent with this approach, Marvell believes that disclosing Non-GAAP financial measures to the readers of its financial statements provides such readers with useful supplemental data that, while not a substitute for GAAP financial measures, allows for greater transparency in the review of its financial and operational performance. Externally, management believes that investors may find Marvell’s Non-GAAP financial measures useful in their assessment of Marvell’s operating performance and the valuation of Marvell. Internally, Marvell’s Non-GAAP financial measures are used in the following areas: ▪ Management’s evaluation of Marvell’s operating performance; ▪ Management’s establishment of internal operating budgets; ▪ Management’s performance comparisons with internal forecasts and targeted business models; and ▪ Management’s determination of the achievement and measurement of certain types of compensation including Marvell’s annual inc entive plan and certain performance-based equity awards (adjustments may vary from award to award). Non-GAAP financial measures have limitations in that they do not reflect all of the costs associated with the operations of Marvell’s business as determined in accordance with GAAP. As a result, you should not consider these measures in isolation or as a substitute for analysis of Marvell’s results as reported under GAAP. The exclusion of the above items from our GAAP financial metrics does not necessarily mean that these costs are unusual or infrequent. Marvell does not provide a reconciliation of its forward-looking Non-GAAP measures to the most directly comparable GAAP measures for periods after the third quarter of fiscal 2027 because certain items that impact those GAAP measures are uncertain, depend on various factors, could be material to Marvell’s results computed in accordance with GA AP, and cannot be provided without unreasonable effort. These items include, but are not limited to, restructuring and other related charges, asset impairments, stock-based compensation expense and other nonrecurring expenses that cannot reasonably be estimated in advance. Discussion of Non-GAAP financial measures
Slide 149 — Appendices
© 2026 Marvell. All rights reserved. 149 Appendices (Millions) (Unaudited) Q2’26 Q1’27 Q2’27 GAAP gross profit $ 1,010.6 $ 1,260.8 $ 1,455.6 Special items – expenses (income): Stock-based compensation 13.4 14.2 15.9 Amortization of acquired intangible assets 167.4 150.8 142.7 Restructuring related charges (gains) (a) ⎯ (2.0) (0.2) Other cost of goods sold ⎯ ⎯ ⎯ Total special items 180.8 163.0 158.4 Non-GAAP gross profit $ 1,191.4 $ 1,423.8 $ 1,614.0 Reconciliation of GAAP gross profit to Non-GAAP gross profit (a) Restructuring and other related items include gain on sale of property, changes in contractual obligations, employee severance costs, facility exit related charges, and other.
Slide 150 — Appendices
© 2026 Marvell. All rights reserved. 150 Appendices (Unaudited) Q2’26 Q1’27 Q2’27 GAAP gross margin 50.4% 52.1% 53.1% Stock-based compensation 0.7% 0.6% 0.6% Amortization of acquired intangible assets 8.3% 6.3% 5.2% Restructuring related charges (gains) (a) ⎯% (0.1) % ⎯ % Other cost of goods sold ⎯ % ⎯ % ⎯ % Non-GAAP gross margin 59.4% 58.9% 58.9% Reconciliation of GAAP gross margin to Non-GAAP gross margin (a) Restructuring and other related items include gain on sale of property, changes in contractual obligations, employee severance costs, facility exit related charges, and other.
Slide 151 — Appendices
© 2026 Marvell. All rights reserved. 151 Appendices Reconciliation of GAAP operating expenses to Non-GAAP operating expenses (Millions) (Unaudited) Q2’26 Q1’27 Q2’27 GAAP operating expenses $ 720.5 $ 921.4 $ 995.9 Special items – (expenses) income: Stock-based compensation (140.2) (193.4) (310.3) Amortization of acquired intangible assets (76.3) (74.4) (72.2) Restructuring related (charges) gains (a) (8.7) (10.7) 2.8 Other (b) (2.7) (66.0) (5.4) Total special items (227.9) (344.5) (385.1) Non-GAAP operating expenses $ 492.6 $ 576.9 $ 610.8 (a) Restructuring and other related items include gain on sale of property, changes in contractual obligations, employee severance costs, facility exit related charges, and other. (b) Other costs in operating expenses, operating income and interest and other loss, net include acquisition and divestiture related costs, gain or loss on investments, gain on sale of intellectual property, and legal contingency matters.
Slide 152 — Appendices
© 2026 Marvell. All rights reserved. 152 Appendices Reconciliation of GAAP operating income to Non-GAAP operating income (Millions) (Unaudited) Q2’26 Q1’27 Q2’27 GAAP operating income $ 290.1 $ 339.4 $ 459.7 Special items – (expenses) income: Stock-based compensation 153.6 207.6 326.2 Amortization of acquired intangible assets 243.7 225.2 214.9 Restructuring related charges (gains) (a) 8.7 8.7 (3.0) Other cost of goods sold ⎯ ⎯ ⎯ Other (b) 2.7 66.0 5.4 Total special items 408.7 507.5 543.5 Non-GAAP operating expenses $ 698.8 $ 846.9 $ 1,003.2 (a) Restructuring and other related items include gain on sale of property, changes in contractual obligations, employee severance costs, facility exit related charges, and other. (b) Other costs in operating expenses, operating income and interest and other loss, net include acquisition and divestiture rela ted costs, gain or loss on investments, gain on sale of intellectual property, and legal contingency matters.
Slide 153 — Appendices
© 2026 Marvell. All rights reserved. 153 Appendices Reconciliation of GAAP operating margin to Non-GAAP operating margin (Unaudited) Q2’26 Q1’27 Q2’27 GAAP operating margin 14.5 % 14.0 % 16.8 % Stock-based compensation 7.7 % 8.6 % 11.9 % Amortization of acquired intangible assets 12.1 % 9.3 % 7.8 % Restructuring related charges (gains) (a) 0.4 % 0.4 % (0.1) % Other cost of goods sold ⎯ % ⎯ % ⎯ % Other (b) 0.1 % 2.7 % 0.2 % Non-GAAP operating margin 34.8 % 35.0 % 36.6 % (a) Restructuring and other related items include gain on sale of property, changes in contractual obligations, employee severance costs, facility exit related charges, and other. (b) Other costs in operating expenses, operating income and interest and other loss, net include acquisition and divestiture related costs, gain or loss on investments, gain on sale of intellectual property, and legal contingency matters.
Slide 154 — Appendices
© 2026 Marvell. All rights reserved. 154 Appendices Reconciliation of GAAP interest and other loss to Non-GAAP interest and other loss (Millions) (Unaudited) Q2’26 Q1’27 Q2’27 GAAP interest and other loss, net $ (56.4) $ (256.1) $ (81.4) Special items – expenses (income): Change in fair value of contingent consideration liability, net of forward stock purchase contract ⎯ 250.7 52.0 Other (b) 8.2 (34.7) (0.9) Total special items 8.2 216.0 51.1 Non-GAAP interest and other loss, net $ (48.2) $ (40.1) $ (30.3) (b) Other costs in operating expenses, operating income and interest and other loss, net include acquisition and divestiture related costs, gain or loss on investments, gain on sale of intellectual property, and legal contingency matters.
Slide 155 — Appendices
© 2026 Marvell. All rights reserved. 155 Appendices Reconciliation of GAAP net income to Non-GAAP net income (1 of 2) (Millions) (Unaudited) Q2’26 Q1’27 Q2’27 GAAP net income $ 194.8 $ 34.5 $ 308.0 Special items – expenses (income): Stock-based compensation 153.6 207.6 326.2 Amortization of acquired intangible assets 243.7 225.2 214.9 Restructuring related charges (gains) (a) 8.7 8.7 (3.0) Other cost of goods sold ⎯ ⎯ ⎯ Change in fair value of contingent consideration liability net of forward stock purchase contract ⎯ 250.7 52.0 Other (b) 10.9 31.3 4.5 Pre-tax total special items 416.9 723.5 594.6 Other income tax effects and adjustments (c) (26.2) (40.0) (36.7) Non-GAAP net income $ 585.5 $ 718.0 $ 865.9
Slide 156 — Appendices
© 2026 Marvell. All rights reserved. 156 Appendices Reconciliation of GAAP net income (loss) to Non-GAAP net income (2 of 2) (a) Restructuring and other related items include gain on sale of property, changes in contractual obligations, employee severance costs, facility exit related charges, and other. (b) Other costs in operating expenses, operating income and interest and other loss, net include acquisition and divestiture rela ted costs, gain or loss on investments, gain on sale of intellectual property, and legal contingency matters. (c) Other income tax effects and adjustments relate to tax provision based on a Non-GAAP income tax rate of 11.0% for the three and six months ended August 1, 2026 and three months ended May 2, 2026. Other income tax effects and adjustments relate to tax provision based on a Non-GAAP income tax rate of 10.0% for the three and six months ended August 2, 2025.
Slide 157 — Appendices
© 2026 Marvell. All rights reserved. 157 Appendices Reconciliation of GAAP diluted net income per share to Non-GAAP diluted net income per share (Millions, except per share amounts) (Unaudited) Q2’26 Q1’27 Q2’27 GAAP weighted-average shares outstanding ⎯ basic 862.6 882.0 897.4 GAAP weighted-average shares outstanding ⎯ diluted 870.4 893.3 921.2 Non-GAAP weighted-average shares outstanding ⎯ diluted 870.4 893.3 921.2 GAAP diluted net income per share $ 0.22 $ 0.04 $ 0.33 Non-GAAP diluted net income per share $ 0.67 $ 0.80 $ 0.94
Slide 158 — Appendices
© 2026 Marvell. All rights reserved. 158 Appendices Reconciliation of GAAP outlook to Non-GAAP outlook (1 of 3) (Millions, except per share amounts) (Unaudited) Q3’27 GAAP net revenue $3,150 +/- 5% Special items: ⎯ Non-GAAP net revenue $3,150 +/- 5% GAAP gross margin 52.9% - 53.9% Special items: Stock-based compensation ~0.8% Amortization of acquired intangible assets ~3.9% Non-GAAP gross margin 57.5% - 58.5%
Slide 159 — Appendices
© 2026 Marvell. All rights reserved. 159 Appendices Reconciliation of GAAP outlook to Non-GAAP outlook (2 of 3) (Millions, except per share amounts) (Unaudited) Q3’27 Total GAAP operating expenses ~ $1,015 Special items: Stock-based compensation 285 Amortization of acquired intangible assets 72 Restructuring related charges 1 Other 2 Total Non-GAAP operating expenses ~ $655
Slide 160 — Appendices
© 2026 Marvell. All rights reserved. 160 Appendices Reconciliation of GAAP outlook to Non-GAAP outlook (3 of 3) (Millions, except per share amounts) (Unaudited) Q3’27 GAAP diluted net income per share $0.53 +/- $0.05 Special items: Stock-based compensation 0.34 Amortization of acquired intangible assets 0.21 Other income tax effects and adjustments (0.03) Other 0.05 Non-GAAP diluted net income per share $1.10 +/- $0.05