Sandisk Corporation · NASDAQ:SNDK
Sandisk — Investor Day (2026)
98 slides from Sandisk's investor day presentation, 2026
Slide 1: Sandisk Investor Day: In Focus 2026
SANDISK" dye DAY N FOCUS 2026
Slide 2: Disclaimers: Forward-Looking Statements
DISCLAIMERS FORWARD-LOOKING STATEMENTS This presentation contains forward-looking statements within the meaning of U.S. federal securities laws, including, but not limited to, statements regarding expectations for: the business outlook, financial condition and financial and operational performance o
Slide 3: Agenda
AGENDA SNDK IN FOCUS NAND FLASH TECHNOLOGY LEADERSHIP NAND MARKET UPDATE AI INFRASTRUCTURE OUTLOOK SUSTAINABLE VALUE CREATION NEXT WAVE OF TECHNOLOGY INNOVATION ©2026 SANDISK CORPORATION OR ITS AFFILIATES ALL RIGHTS RESERVED
Slide 4: In Focus
IN FOCUS DAVID GOECKELER CHAIRMAN & CHIEF EXECUTIVE OFFICER
Slide 5: Welcome
WELCOME Delivering on Our Invest in the Three Imperatives for Commitments Business Sustained Value Creation ©2026 SANDISK CORPORATION OR ITS AFFILIATES ALL RIGHTS RESERVED
Slide 6: Invest in the Business
INVEST IN THE BUSINESS MULTI-HORIZON INNOVATION CASH POSITIVE BALANCE SHEET BiCS 9, 10, 11, 12, 13 Systems Capabilities Across Consumer, No Debt Edge, and Datacenter Portfolios High Bandwidth Flash (HBF™) Significant Cash Reserves 3D Matrix Memory ©2026 SANDISK CORPORATION OR ITS AFFILIATES ALL RIGH
Slide 7: Invest in the Business: Nanya Investment
INVEST IN THE BUSINESS NANYA INVESTMENT JV Extended through 2034 Secure DRAM for eSSD Portfolio R&D Scale (BiCS) Delivers ~4% Equity Stake Technology & Cost Leadership JV is the Largest Producer of Foundational for NBMs NAND Wafers in the World (33%) Source: TechInsights NAND Market Report Q2 2026 ©
Slide 8: Three Imperatives for Sustained Value Creation
THREE IMPERATIVES FOR SUSTAINED VALUE CREATION Increase Reduce Consistent Profitability Cyclicality Revenue Growth MAXIMIZE RETURN ACROSS ALL TIME HORIZONS ©2026 SANDISK CORPORATION OR ITS AFFILIATES ALL RIGHTS RESERVED
Slide 9: Increase Profitability
THREE IMPERATIVES FOR SUSTAINED VALUE CREATION INCREASE PROFITABILITY We have been Investing in NAND Innovation (BiCS), Systems High-Value Technical Expertise (Controllers), and Manufacturing for Decades Expertise and Scale WE HOLD THE WHOLE STACK, WITH MINIMAL PROFITABILITY LEAKAGE NAND IP → FRONT-
Slide 10: Reduce Cyclicality: Consumer
THREE IMPERATIVES FOR SUSTAINED VALUE CREATION REDUCE CYCLICALITY (→) CONSUMER Consumer Business as the Foundation — Inherently Less Cyclical GM% 351K Points of Sale with a Global Customer Base 0 Global Brand Equity TechInsights TAM Sandisk Consumer 2.1B Sandisk data from internal sources and Indust
Slide 11: Reduce Cyclicality: New Business Models
THREE IMPERATIVES FOR SUSTAINED VALUE CREATION REDUCE CYCLICALITY (→) NEW BUSINESS MODELS Longer Term Commitments Month-by-Month Mix / Volume / by Both Parties Price Commitments • 1 to 5 years, with average weighted length of 4+ years • Bigger customers have longer commitments Financial Guarantees t
Slide 12: Reduce Cyclicality: Lower Volume, High-Value Business
THREE IMPERATIVES FOR SUSTAINED VALUE CREATION 00.01 REDUCE CYCLICALITY (→) LOWER VOLUME, 2 HIGH-VALUE BUSINESS Auto IoT Physical AI Emerging Technologies ©2026 SANDISK CORPORATION OR ITS AFFILIATES ALL RIGHTS RESERVED
Slide 13: Consistent Revenue Growth
THREE IMPERATIVES FOR SUSTAINED VALUE CREATION CONSISTENT REVENUE GROWTH INVESTING TO GROW VOLUME MID-TO-HIGH TEENS PER YEAR ©2026 SANDISK CORPORATION OR ITS AFFILIATES ALL RIGHTS RESERVED
Slide 14: Consistent Revenue Growth: BiCS Roadmap
THREE IMPERATIVES FOR SUSTAINED VALUE CREATION CONSISTENT REVENUE GROWTH BIT-COST SCALABLE (BiCS) ROADMAP 27% INVESTING TO CAGR GROW VOLUME Yearly Average 27% MID-TO- Productivity CUMULATIVE BIT GROWTH HIGH 54% BiCS Gen-to-Gen Average Bit Growth Per Wafer TEENS 5 1.0x PER9 YEAR Nodes Over Years BiCS
Slide 15: In Focus Today
IN FOCUS TODAY NAND Innovation NAND Market Era of The Business HBF™ Roadmap Update AI Inference Model Updates ALPER ILKBAHAR ERIC CHERRSTROM KHURRAM ISMAIL LUIS VISOSO ALPER ILKBAHAR Chief Technology Officer VP, Market Intelligence Chief Product Officer Chief Financial Officer Chief Technology Offic
Slide 16: NAND Flash Technology Leadership
NAND FLASH TECHNOLOGY LEADERSHIP ALPER ILKBAHAR CHIEF TECHNOLOGY OFFICER
Slide 17: Sandisk Technology Pillars
SANDISK TECHNOLOGY PILLARS Keep the Exponential Scaling Deliver Performance, Power, Density, Innovate to Amaze! Engine Running and Capital Efficiency Leadership (↓) (↓) (↓) A 19-Generation Scaling Engine Category-Leading Performance, Beyond Scaling: Breakthrough with a Clear BiCS Roadmap for Power E
Slide 18: The Most Scalable Semiconductor Technology
THE MOST SCALABLE SEMICONDUCTOR TECHNOLOGY TODAY 1 Gb 10 Gb 100 Gb 1 Tb 2 Tb PROGRESS OVER LAST 18 MONTHS 2 NEW NAND GENERATIONS 1Gb X2-160nm 2Gb X2-130nm 4Gb X2-90nm BiCS9 BiCS10 COST (LOG SCALE) 8Gb X2-70nm 16Gb X3-56nm 32Gb X3-43nm 19 64Gb X3-32nm 64Gb X3-24nm 256Gb 17 128Gb X3-19nm 128Gb X3-BiCS
Slide 19: Four Vectors of 3D NAND Scaling
FOUR FOUR VECTORS VECTORS OF 3DOF 3DSCALING NAND NAND SCALING Cost Reduction % / Bit Growth % 1 Higher is Better Lateral Overhead Overhead Memory Hole Staggering and Scaling Density Overhead Reduction ~1 Logical Scaling ER 1 à ER 1 2 3 à ER 1 2 3 4 5 6 7 à ER 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 à ER
Slide 20: Nodes Designed to Optimize Sandisk Capital Intensity
NODES DESIGNED TO OPTIMIZE SANDISK CAPITAL INTENSITY 0.45 FLASH CAPEX / INCREMENTAL PETABYTE OUTPUT 3.00 Rolling 3-Year Average from 2018 - 2025 2.66 CAPEX / INCREMENTAL OUTPUT [au] 0.40 INDUSTRY / ( SANDISK + KIOXIA) + 2.41 2.50 INDUSTRY 0.35 INDUSTRY / ( + ) 0.30 1.83 2.00 1.55 1.72 1.68 1.44 1.38
Slide 21: Nodes Designed to Optimize Sandisk Capital Intensity: Capex vs Output
NODES DESIGNED TO OPTIMIZE SANDISK CAPITAL INTENSITY 2021 – 2025 % OF TOTAL NAND CAPEX VS 2021 – 2025 % OF TOTAL NAND CAPEX & % OF TOTAL NAND OUTPUT & % OF TOTAL NAND OUTPUT BY COMPANY % PRODUCTION / % CAPEX BY COMPANY 100% 50% 250% PEER 4 8% % PRODUCTION / % CAPEX BY COMPANY 90% 11% 45% 223% 80% 40
Slide 22: A Technology Roadmap Engineered for AI's Pace
A TECHNOLOGY ROADMAP ENGINEERED FOR AI’S PACE Future BiCS Technology PERFORMANCE / POWER EFFICIENCY BiCS9 BiCS11 BiCS10 BiCS8 CBA-enabled leap BiCS6 Traditional roadmap BiCS GENERATION ROADMAP — MORE BITS PER WAFER à Note: Images are not to scale. ©2026 SANDISK CORPORATION OR ITS AFFILIATES ALL RIGH
Slide 23: BiCS8 – The Gold Standard in NAND
BiCS8 – THE GOLD STANDARD IN NAND BiCS8 ARRA YW AFER 218-Layer TLC & QLC ACE – – CBA Technology G IN TERF DIN – – BON Engineered around Performance, AFER Power Efficiency, and Density W CMOS IMPROVEMENTS OVER PREVIOUS GENERATION (BiCS8 TLC VS BiCS6 TLC) >50% +60% +35% +42% MEMORY DENSITY INTERFACE B
Slide 24: BiCS8 – Engineered for the Inference Age
BiCS8 – ENGINEERED FOR THE INFERENCE AGE BiCS8 ARRA YW AFER 218-Layer TLC ACE – – CBA Technology G IN TERF DIN – – BON Engineered around Performance, AFER Power Efficiency, and Density W CMOS POWER EFFICIENCY WRITE READ +62% PERFORMANCE PER WATT +21% PERFORMANCE PER WATT PERFORMANCE AGAINST PEER AVE
Slide 25: BiCS8 – Engineered for the Inference Age: Power Efficiency
BiCS8 – ENGINEERED FOR THE INFERENCE AGE WRITE PERFORMANCE PER WATT [GBps/W] READ PERFORMANCE PER WATT [GBps/W] BETTER POWER EFFICIENCY à BiCS10 +23.4% BETTER POWER EFFICIENCY à BiCS10 +18.6% 62.7% 64.1% 21.6% 29.1% PEER 2XX GEN* PEER 2YY/3XX* PEER 2XX GEN* PEER 2YY/3XX* AVG AVG AVG AVG *GEN 2XX ref
Slide 26: Introducing BiCS9 2Tb QLC
INTRODUCING BiCS9 2Tb QLC ENABLING NEXT-GENERATION DATACENTER STORAGE INTERFACES PERFORMANCE — Bandwidth Mature BiCS8 Cell Array Write Proven Fab Tools, Yield, and Quality +150% Minimal CAPEX Read +75% BiCS10-like CMOS POWER EFFICIENCY — Bandwidth/Watt BiCS10 Feature Set, Write BiCS9 2Tb QLC Higher
Slide 27: Next-Gen BiCS – BiCS10 332-Layer 1Tb TLC
NEXT-GEN BiCS – BiCS10 BiCS10 332-Layer 1Tb TLC CBA Technology Sampling in August 2026 Development Moving Ahead of Expectations IMPROVEMENTS OVER BiCS8 +59% +33% +23% +38% INTERFACE MEMORY DENSITY INTERFACE BANDWIDTH WRITE & READ BANDWIDTH POWER EFFICIENCY Source: Sandisk testing. ©2026 SANDISK CORP
Slide 28: Next-Gen BiCS – BiCS10 332-Layer 2Tb QLC
NEXT-GEN BiCS – BiCS10 BiCS10 332-Layer 2Tb QLC CBA Technology World’s Highest Density Memory Chip1 IMPROVEMENTS OVER BiCS8 +60% +100% +33% +75% WRITE & READ MEMORY DENSITY WRITE & READ BANDWIDTH INTERFACE SPEED POWER EFFICIENCY Source: Sandisk testing. 1. As of August 13, 2026; based on QLC NAND. ©
Slide 29: Next-Gen BiCS – Scaling to Match the Pace of AI
NEXT-GEN BiCS – SCALING TO MATCH THE PACE OF AI +65% MORE DIE PER WAFER BiCS10 BiCS8 332-Layer 2Tb QLC 218-Layer 2Tb QLC ©2026 SANDISK CORPORATION OR ITS AFFILIATES ALL RIGHTS RESERVED
Slide 30: Technology Transitions Delivering Bit Growth Ahead of Demand
TECHNOLOGY TRANSITIONS DELIVERING BIT GROWTH AHEAD OF DEMAND 54% AVERAGE GEN-TO-GEN, BIT PER WAFER GROWTH OVER 5 GENERATIONS BIT-PER-WAFER GROWTH RELATIVE TO BiCS5 CAGR 27% 1.00x BiCS5 BiCS6 BiCS8 BiCS10 BiCS11 Source: Sandisk internal data. ©2026 SANDISK CORPORATION OR ITS AFFILIATES ALL RIGHTS RES
Slide 31: NAND Market Update
NAND MARKET UPDATE ERIC CHERRSTROM VICE PRESIDENT, MARKET INTELLIGENCE
Slide 32: Flash Market to Reach 1.2ZB in 2026
FLASH MARKET TO REACH 1.2ZB IN 2026 IN 2026 DATA CENTER OVERTAKES EDGE AS THE LARGEST FLASH SWIMLANE Growth in Consumer & Edge Growth Driven by AI/DC Buildout (30%+ CAGR) (High-Teens CAGR) 2500 Key Drivers: Key Drivers: • Phone volumes scale to 1.5Bu by mid 2010s • AI buildout drives DC to majority
Slide 33: Flash Demand Mixes to Premium Swimlanes to Reset the Flash TAM Baseline
FLASH DEMAND MIXES TO PREMIUM SWIMLANES TO RESET THE FLASH TAM BASELINE HISTORICAL ~$60B AVERAGE +/- $20B CYCLES WITH CY26-27 DRIVEN BY DATA CENTER GROWTH $600 Approaching $500 $500B PRIOR OUTLOOK $400 Flash was priced as a commodity: falling Flash is a critical component of a multi- Industry Revenu
Slide 34: Fab Wafer Capacity Reduced Due to 2022-2023 Cycle
FAB WAFER CAPACITY REDUCED DUE TO 2022-2023 CYCLE CAPACITY PEAKED IN CY22, TROUGHED TO ~70% UTILIZATION IN THE CORRECTION, RE-ESTABLISHING CAPACITY ~30% BELOW PEAK 2 . DEMA N D EX P ECTA TION S DID N OT 1 . B UILD OUT | UP TO CY2 2 MA TERIA LIZ E | CY2 3 -CY2 5 3 . S TRUCTURA L RES ET | CY2 6 + Wafe
Slide 35: Cloud Capex Revisions Signal Accelerating Investments
CLOUD CAPEX REVISIONS SIGNAL ACCELERATING INVESTMENTS U.S. DATA CENTER CAPEX KEEPS CLIMBING: 2026 TO $842B AND 2027 TO $1,097B Cloud CAPEX Projections ($B) 2023 Regime Change $1,200 $1,097 Data center investments gives way to AI-led: The permeation of AI has led to 15 consecutive quarters of upward
Slide 36: Edge: Premium Mix Drives Revenue Growth as the Low-End Exits
EDGE: PREMIUM MIX DRIVES REVENUE GROWTH AS THE LOW-END EXITS OEMs ARE SELLING FEWER UNITS AT HIGHER ASPs, SUSTAINING REVENUE GROWTH ACROSS PCs AND SMARTPHONES Smartphone Units and Rev$ Mobile TAM 1,500 $650 $600 Units (Mu) 1,000 Rev $B $550 500 • +13% to +24% range of revenue growth for Dell CSG, $5
Slide 37: The New Flash Reality
THE NEW FLASH REALITY DEMAND EDGE SUPPLY Datacenter Takes the Majority of Premium Mix Carries the Growth Nodal Migration Enabled Bit Growth TAM PC and smartphone revenue sustains as sales Bit growth can be achieved via nodal transitions Hyperscale capex — revised up 15 straight shift to premium and
Slide 38: AI Infrastructure Outlook
AI INFRASTRUCTURE OUTLOOK KHURRAM ISMAIL CHIEF PRODUCT OFFICER
Slide 39: Unprecedented Pace of Innovation: The AI Infrastructure Revolution
UNPRECEDENTED PACE OF INNOVATION THE AI INFRASTRUCTURE REVOLUTION ©2026 SANDISK CORPORATION OR ITS AFFILIATES ALL RIGHTS RESERVED
Slide 40: Human Collective Memory vs AI Collective Memory / Storage
Human Collective Memory AI Collective Memory / Storage ~2.5 PB1 × 8B ≈ Still at Zettabyte level ~20 Yottabytes global install base • Includes short- and long-term memory • Includes ephemeral and persistent memory • Source of understanding • Access layer to stored knowledge The scale of intelligence
Slide 41: AI Data Center TAM: 2030 Shipment Estimate
Workloads2 2030 SHIPMENT EST. AI DC TAM1 25% Fast Data Lakes 40% Staging KV Cache 35% 1.2ZB NAND Technology2 34% TLC 66% QLC 2030 1. Tech Insights NAND Market Report Q2 2026. 2. Sandisk internal splits. ©2026 SANDISK CORPORATION OR ITS AFFILIATES ALL RIGHTS RESERVED
Slide 42: Infrastructure Decision Making: TCO to TVO
INFRASTRUCTURE DECISION MAKING TCO (→) TVO Total Cost of Ownership Total Value of Ownership Cost-Efficiency Focus Value & Output Focus • VMs / transactions • Tokenomics • Utilization, $/unit • Tokens/$, Tokens/W • Lifetime Infrastructure cost • Power utilization and TTM • Lowest cost at scale • High
Slide 43: Infrastructure Decision Making: TVO
INFRASTRUCTURE DECISION MAKING TVO Total Value of Ownership The Race to Scale Intelligence Value & Output Focus Tokenomics and User Capture • Tokenomics • Tokens/$, Tokens/W • Power utilization and TTM Massive Capex and Key Component Timing • Highest value per watt • Specialized architectures Power
Slide 44: The Era of Inference
THE ERA 4 Inference OF INFERENCE 3 Model Training 5 New Content Generation 2 Model Data Preparation 1 Raw Data Archives, Content Storage ©2026 SANDISK CORPORATION OR ITS AFFILIATES ALL RIGHTS RESERVED
Slide 45: 5-Stage AI Data Cycle – The Era of Inference
5-STAGE AI DATA CYCLE – THE ERA OF INFERENCE SNOWBALLING KV CACHE STAGE 1 STAGE 2 STAGE 3 STAGE 4 STAGE 5 Raw Data Archives, Model Data Model Inference New Content Content Storage Preparation Training Generation Foundation Models B. Activation • Prefill Staging A. Augmentation • Decode STAGING (KV •
Slide 46: KV Cache: The Working Memory of AI Inference
KV CACHE THE WORKING MEMORY OF AI INFERENCE STORES PRIOR ATTENTION SO EACH NEW TOKEN CAN REUSE CONTEXT Shared/Prompt-Prefix Reuse Less Recompute Ephemeral User Prompt/ Prefill: KV Decode: Response Cache: Context compute K/V reusable generate tokens attention context one at a time state memory Reused
Slide 47: KV Cache Inference at Scale
KV CACHE INFERENCE AT SCALE UPDATED MEMORY HIERARCHY 2026 Higher Token XPU Generation SRAM 2024 Tier 0 GPU HBM Tier 1 Context & State Preservation Agentic System DRAM Tier 2 AI Longer Context / Flash Reasoning Tier 3 Higher Data Archive Storage HDD Tier 4 Generation ©2026 SANDISK CORPORATION OR ITS
Slide 48: KV Cache Inference at Scale: Updated Memory Hierarchy
KV CACHE INFERENCE AT SCALE UPDATED MEMORY HIERARCHY XPU SRAM 2026 Tier 0 GPU HBM XPU Tier 1 SRAM 2024 Tier 0 System DRAM GPU HBM Tier 2 Tier 1 Persistent Direct Attached SSD System DRAM Tier 3 Tier 2 KV Cache Network Attached Compute SSD Flash (for inference context storage) Tier 3 Tier 3.5 Network
Slide 49: KV Cache Equation of the Future
KV CACHE EQUATION OF THE FUTURE PERSISTENT KV STORE eSSD Number of Average input KV size per Persistent KV concurrent Cache miss tokens per token store eSSD sessions ratio session (EB) during retention time Variables change based on provider, use-case, and architecture ©2026 SANDISK CORPORATION OR I
Slide 50: KV Cache Equation of the Future: 2030 Install Base
KV CACHE EQUATION OF THE FUTURE PERSISTENT KV STORE eSSD Metrics 2030 INSTALL BASE 1 Concurrent Sessions During Retention Time Cache Miss Ratio + Average Input Tokens per Session KV Size per Token Persistent KV Store Install Base – eSSD ZB1 Variables change based on provider, use-case, and architect
Slide 51: AI Data Center TAM: 2030 Shipment Estimate
Workloads2 2030 SHIPMENT EST. AI DC TAM1 25% Fast Data Lakes 40% Staging KV Cache 35% 1.2ZB NAND Technology2 34% TLC 66% QLC 2030 1. Tech Insights NAND Market Report Q2 2026. 2. Sandisk internal splits. ©2026 SANDISK CORPORATION OR ITS AFFILIATES ALL RIGHTS RESERVED
Slide 52: AI Data Center eSSD Placements
AI DATA CENTER eSSD PLACEMENTS Fast Data Lakes QLC - Capacity KV Cache QLC - Capacity DIRECT Attached TLC - Performance Staging REMOTE Network TLC - Performance KV Cache TLC - Performance Staging TLC - Performance KV Cache TLC - Endurance LOCAL Network ©2026 SANDISK CORPORATION OR ITS AFFILIATES ALL
Slide 53: Sandisk SSDs Serve the AI Data Cycle
SANDISK SSDs SERVE THE AI DATA CYCLE COMPUTE TLC eSSD eSSD STORAGE QLC eSSD eSSD SPECIFICATIONS SPECIFICATIONS PCIe® Gen 5.0, NVMe™ 2.0 PCIe® Gen 5.0, NVMe™ 2.0 Compliant Compliant BiCS8 TLC NAND BiCS8 QLC NAND High-Density NAND Packaging FORM-FACTORS FORM-FACTORS E1.S, U.2, E3.S U.2, E3 CAPACITIES1
Slide 54: Inference in Action: Sandisk in the Lab
INFERENCE IN ACTION: SANDISK IN THE LAB REAL-WORLD BENEFITS FOR INFERENCE ~75% Less Energy with SSD ~3× Higher Throughput with SSD Note: Based on internal testing, performance will vary ©2026 SANDISK CORPORATION OR ITS AFFILIATES ALL RIGHTS RESERVED
Slide 55: SSD is a Token Battery
SSD is a Token Battery Tokens = Work Already Done Reuse Stored Tokens Many generated tokens Fetch pre-computed tokens represent prior computation from eSSD in microseconds, that’s often recomputed drawing on stored energy to repeatedly, wasting energy save time and power Storing expended work (token
Slide 56: The Race to Scale Intelligence: Robust Growth Outlook
The Race to Scale Intelligence Robust Growth Outlook • Strong Qualified Product Portfolio with Aligned Roadmap • Efficiencies Drive More Utilization • Industry is Innovating to Flash ©2026 SANDISK CORPORATION OR ITS AFFILIATES ALL RIGHTS RESERVED
Slide 57: Sustainable Value Creation
SUSTAINABLE VALUE CREATION LUIS VISOSO CHIEF FINANCIAL OFFICER
Slide 58: FY'26 A Year of Transformation
FY’26 A YEAR OF TRANSFORMATION OBJECTIVE: SUSTAINABLE VALUE CREATION FOR CUSTOMERS AND SHAREHOLDERS $300B • TAM >3X Y/Y in CY’26, approaching $500B in CY’27 • Datacenter becomes half of the market in CY’26, driven by AI $20B • Sandisk full-year revenue of $20B, up 175% Y/Y • Annualized Q4 revenue of
Slide 59: NBMs: Strong Multi-Year Customer Partnerships
NBMs: STRONG MULTI-YEAR CUSTOMER PARTNERSHIPS CUSTOMERS ACROSS EDGE AND DATACENTER Custom-made Framework Metrics • Strategic partnerships • Multi-year commitment • Average contract duration • Meeting customer and • Contractually growing volumes • NBM % of total bits Sandisk needs • Price fixed/varia
Slide 60: Selective Partnerships
SELECTIVE PARTNERSHIPS TAKING COLLABORATION TO A DIFFERENT LEVEL Engagement with Customers at the Highest Level Strategic Win-Win Conversations 8 Technical Integration and Commercial Alignment 8 Individual Customers DATACENTER-CENTRIC 2 Customers Already Expanded Their NBMs ©2026 SANDISK CORPORATION
Slide 61: Long-Term Relationships
LONG-TERM RELATIONSHIPS FROM QUARTERLY PRICE NEGOTIATIONS TO MULTI-YEAR ENGAGEMENTS Average Weighted Length of Contract is 4+ Years 4 Longest Engagement is 5 Years + Details by Quarter Exploring Opportunities to Go Further Out in Time ©2026 SANDISK CORPORATION OR ITS AFFILIATES ALL RIGHTS RESERVED
Slide 62: Becoming the Predominant Model
BECOMING THE PREDOMINANT MODEL COMPLEMENTED WITH NON-NBM BUSINESS FY 2027 FY 2028 ~1/2 ~2/3 BITS BITS Pricing NBM Non-NBM Combination Fixed and Variable Fluctuate with the Market (Floor and Ceiling Pricing) ©2026 SANDISK CORPORATION OR ITS AFFILIATES ALL RIGHTS RESERVED
Slide 63: Eight Customers: Total Contract Value and RPO
EIGHT CUSTOMERS MEANINGFUL RELATIONSHIPS $93.9B 1 $91.1B 2 Total Contract Value Remaining Performance (TCV) for the 8 NBM Obligation (RPO) for the 8 NBM Customers at Floor Pricing Customers at Floor Pricing Expect Upside Pricing and Revenue 1. Total revenue expected, over the life of the agreements,
Slide 64: Supported by Financial Guarantees
SUPPORTED BY FINANCIAL GUARANTEES IF CUSTOMERS FAIL TO MEET PERFORMANCE OBLIGATIONS $16.5B Total Financial Guarantees of $16.5B Primarily Held by or Provided Through Third-Party Financial Institutions, with $2.5B in Our Cash Balance Unless They Have Prepaid in Cash, Customers Pay for Products in the
Slide 65: Ratio of Financial Guarantees Increases Over Time
RATIO OF FINANCIAL GUARANTEES INCREASES OVER TIME ILLUSTRATIVE 3-YEAR CUSTOMER AT SIGNING TWO YEARS LATER BASE RATIO 2X BASE RATIO Base Ratio Equals the Financial Guarantees Divided by the Total Contract Value In Each Multi-Year Contract, the Ratio Between the Financial Guarantees and the RPO Increa
Slide 66: NBM Priorities Going Forward
NBM PRIORITIES GOING FORWARD 1 (→) Execute NBMs with Excellence 2 (→) Remain Highly Selective and Patient in Evaluating Additional NBMs ©2026 SANDISK CORPORATION OR ITS AFFILIATES ALL RIGHTS RESERVED
Slide 67: NBM and Non-NBM: Long-Term Sustainable Model Through 2030
LONG-TERM SUSTAINABLE MODEL THROUGH 2030 • Predominant model NBM • Attractive financials, even at floor pricing • Reliable volume NON- • Continue to support non-NBM customers • Prioritize strategic customers that value our products NBM • Volume and price will fluctuate with the market ©2026 SANDISK
Slide 68: Long-Term Sustainable Model
LONG-TERM SUSTAINABLE MODEL FY 2028 Through FY 2030 Average Performance Revenue Growth Rate Mid-to-High Teens Non-GAAP Gross Margin ~80% Non-GAAP Operating Margin ~75% Adjusted Free Cash Flow Margin ~50% Capital Intensity (% of Revenue) Mid-Single Digits % NOTE: Forward-looking Non-GAAP measures exc
Slide 69: Three Capital Allocation Priorities
THREE CAPITAL ALLOCATION PRIORITIES 1 2 3 Invest in the Business Maintain Strong Return 100% Excess • Fund the business Balance Sheet Cash to Shareholders • Strengthen capabilities • Healthy cash balance • Lead innovation • No debt • Improve credit rating ©2026 SANDISK CORPORATION OR ITS AFFILIATES
Slide 70: In Closing: Sustainable Value Creation
IN CLOSING SUSTAINABLE VALUE CREATION FOR CUSTOMERS AND SHAREHOLDERS UNIQUE • The NAND market is growing at an accelerated pace, supported by AI OPPORTUNITY • Sandisk is very well positioned to capture the opportunity • Our relationship and partnership with customers is evolving, BUSINESS grounded o
Slide 71: Appendix
APPENDIX ©2026 SANDISK CORPORATION OR ITS AFFILIATES ALL RIGHTS RESERVED
Slide 72: GAAP to Non-GAAP Reconciliations
GAAP TO NON-GAAP RECONCILIATIONS In millions; unaudited Q4’25 FY’25 Q1’26 Q2’26 Q3’26 Q4’26 FY’26 Revenue $ 1,901 $ 7,355 $ 2,308 $ 3,025 $ 5,950 $ 8,965 $ 20,248 GAAP Gross Profit 498 2,212 687 1,541 4,662 7,582 14,472 Stock-based compensation expense 4 16 4 5 4 6 19 Non-GAAP Gross Profit $ 502 $ 2
Slide 73: GAAP to Non-GAAP Reconciliations (Cont'd)
GAAP TO NON-GAAP RECONCILIATIONS (CONT’D) In millions; unaudited Q4’25 FY’25 Q1’26 Q2’26 Q3’26 Q4’26 FY’26 GAAP Operating Income (Loss) $ 18 $ (1,377) $ 176 $ 1,065 $ 4,111 $ 7,037 $ 12,389 Gross profit adjustments 4 16 4 5 4 6 19 Operating expense adjustments 78 2,050 65 63 103 61 292 Non-GAAP Oper
Slide 74: GAAP to Non-GAAP Reconciliations (Cont'd)
GAAP TO NON-GAAP RECONCILIATIONS (CONT’D) In millions, except per share amount; unaudited Q4’25 FY’25 Q1’26 Q2’26 Q3’26 Q4’26 FY’26 GAAP net income (loss) $ (23) $ (1,641) $ 112 $ 803 $ 3,615 $ 6,903 $ 11,433 Goodwill impairment - 1,830 - - - - - Stock-based compensation expense 49 182 53 58 54 67 2
Slide 75: GAAP to Non-GAAP Reconciliations (Cont'd)
GAAP TO NON-GAAP RECONCILIATIONS (CONT’D) In millions; unaudited Q4’25 FY’25 Q1’26 Q2’26 Q3’26 Q4’26 FY’26 Cash Flows Cash flow provided by operating activities $ 94 $ 84 $ 488 $ 1,019 $ 3,038 $ 7,126 $ 11,671 Purchases of property, plant and equipment, net (45) (204) (50) (39) (45) (43) (177) Free
Slide 76: GAAP to Non-GAAP Reconciliations: Footnotes
GAAP TO NON-GAAP RECONCILIATIONS (CONT’D) FOOTNOTES This presentation contains the following financial measures that are not in accordance with U.S. generally accepted accounting principles (“GAAP”): Non-GAAP gross profit; Non-GAAP operating expenses; Non- GAAP operating income; Non-GAAP interest an
Slide 77: GAAP to Non-GAAP Reconciliations (Cont'd)
GAAP TO NON-GAAP RECONCILIATIONS (CONT’D) Employee termination and other. From time to time, in order to realign the Company’s operations with anticipated market demand, the Company may terminate employees and/or restructure its operations. From time to time, the Company may also incur charges from
Slide 78: Next Wave of Technology Innovation
NEXT WAVE OF TECHNOLOGY INNOVATION ALPER ILKBAHAR CHIEF TECHNOLOGY OFFICER
Slide 79: Sandisk Breaking Down the Memory Wall
SANDISK BREAKING DOWN THE MEMORY WALL 3D Matrix Memory Next-Generation, High-Density Memory HBF™ High Bandwidth Flash ©2026 SANDISK CORPORATION OR ITS AFFILIATES ALL RIGHTS RESERVED
Slide 80: Sandisk 3D Matrix Memory Development Update
SANDISK 3D MATRIX MEMORY DEVELOPMENT UPDATE SOLID PROGRESS TOWARDS TECHNOLOGY QUALIFICATION Last Year’s Update Engaged With IMEC on Technology Development → 2017 2018 2021 2024 2026 Isolated Passive CMOS Dev CMOS Dev Devices Arrays Vehicle Vehicle 100b-1kbit 10-100kbit for IMEC ©2026 SANDISK CORPORA
Slide 81: Sandisk 3D Matrix Memory Development Update
SANDISK 3D MATRIX MEMORY DEVELOPMENT UPDATE SOLID PROGRESS TOWARDS TECHNOLOGY QUALIFICATION Today Delivered 300mm Wafers and → Packaged Parts 2024 2026 Demonstrated Multi-Gb Functional Memory Arrays CMOS Dev Vehicle for IMEC Achieved Performance Approaching Product Specs ©2026 SANDISK CORPORATION OR
Slide 82: Sandisk Breaking Down the Memory Wall
SANDISK BREAKING DOWN THE MEMORY WALL 3D Matrix Memory Next-Generation, High-Density Memory HBF™ High Bandwidth Flash ©2026 SANDISK CORPORATION OR ITS AFFILIATES ALL RIGHTS RESERVED
Slide 83: HBF: Sandisk's Solution to the Memory Wall Problem
HBF™: SANDISK’S SOLUTION TO THE MEMORY WALL PROBLEM HBF • Same read bandwidth as HBM with up to 8- 16x capacity • Invented for AI inference workloads with large MoE based LLMs, long context lengths, and large KV Caches • Architecture developed with inputs from major cloud/AI customers NOTE: Image a
Slide 84: Recent AI Model Trends
RECENT AI MODEL TRENDS THE NEW AI PARADIGM: MEMORY-CENTRIC INFERENCE Total Mixture Active Context Models Released Parameters of Experts Parameters Length Grok 4 Jul 9, 2025 ~1.7T † Dense (Est.) † ~1.7T † 256K GLM 5.2 Jun 13, 2026 744B 256 (8 Active) † 40B 1M GPT 5.6 Jul 9, 2026 ~1.7-1.8T † No Public
Slide 85: Recent AI Model Trends
RECENT AI MODEL TRENDS THE NEW AI PARADIGM: MEMORY-CENTRIC INFERENCE Higher Memory Capacity (Model Size, Context Lengths) + Lower Compute Requirements (MoE Architecture) = Memory-Centric AI ©2026 SANDISK CORPORATION OR ITS AFFILIATES ALL RIGHTS RESERVED
Slide 86: HBF as Innovation Platform
HBF™ AS INNOVATION PLATFORM HBF ARCHITECTURE PROPOSALS FROM ECOSYSTEM HBF Only HBM/HBF Cached HBF Can Replace HBM Stacks to Add More Read- HBM is Used for Tiered Cache and HBF Optimized Memory in a Similar Footprint Can Store Weights and KV Cache HBF (+HBM) Disaggregated (N) stack positions with (k)
Slide 87: HBF vs HBM Simulation Configurations
HBF™ vs HBM SIMULATION CONFIGURATIONS VS Architecture HBF Only HBM Only GPU Perf 4.5 PFLOPS HBM or HBF Total 12.8TB/s Bandwidth/GPU GPU to GPU Bandwidth 900GB/s HBM or HBF Total 4TB 192GB Capacity/GPU Per GPU Connection to KV 128GB/s to LPDDR DRAM 128GB/s to LPDDR DRAM Spillover Memory Number of GPU
Slide 88: HBF vs HBM Performance Comparison for AI Inference
HBF™ vs HBM PERFORMANCE COMPARISON FOR AI INFERENCE HBF Only HBM Only 1 GPU 8 GPU 4 GPU 8× 2× CAPEX EFFICIENCY GPU EFFICIENCY for minimum configuration Half the number of HBF GPUs required to run model : deliver the same token output : 1 HBF GPU vs 8 HBM GPUs1 4 HBF GPUs = 8 HBM GPUs1 1. Based on in
Slide 89: HBF Roadmap
HBF™ ROADMAP 2ND-GEN HBF TECHNOLOGY FOR THE EDGE “AI THAT NEVER FORGETS” Enabling 100B+ Models on Your Edge Device *BASED ON INTERNAL ESTIMATES Actively being co-developed with multiple customers ©2026 SANDISK CORPORATION OR ITS AFFILIATES ALL RIGHTS RESERVED
Slide 90: HBF Consortium Milestones
HBF™ CONSORTIUM MILESTONES AUG 2025 FEB 2026 AUG 2026 2027 Sandisk & SK hynix HBF Technology First HBF Generational Spec Announcement Consortium Formed Spec Released Updates Joint Agreement on Dedicated Workstream Inaugural High Bandwidth Future HBF HBF Standardization, Established Under OCP Flash S
Slide 91: A Board Built for the AI Memory Era
A BOARD BUILT FOR THE AI MEMORY ERA HBF™ TECHNICAL ADVISORY BOARD N EW MEMB ER DAVID PATTERSON RAJA KODURI JIM KELLER Pardee Professor Emeritus, UC Berkeley • Founder & CEO, Oxmiq Labs • CEO, Tenstorrent Fellow, Google Co-Founder, Mihira Visual Studios Legendary chip architect bringing AI-inference
Slide 92: A Board Built for the AI Memory Era: Jim Keller
A BOARD BUILT FOR THE AI MEMORY ERA HBF™ TECHNICAL ADVISORY BOARD NEW MEMBER JIM KELLER CEO, Tenstorrent DEC ALPHA / AMD K7 & K8 / APPLE A4 & A5 / AMD ZEN / TESLA FSD / INTEL / TENSTORRENT Four decades of silicon leadership — lead architect behind AMD Zen, Apple’s A4 and A5, Tesla’s Full Self-Drivin
Slide 93: Tenstorrent Builds AI Computers and RISC-V CPUs
TENSTORRENT BUILDS AI COMPUTERS AND RISC-V CPUs Source: Tenstorrent. ©2026 SANDISK CORPORATION OR ITS AFFILIATES ALL RIGHTS RESERVED
Slide 94: Tenstorrent Galaxy Blackhole Is in Production and Built for Scale
TENSTORRENT GALAXY BLACKHOLE IS IN PRODUCTION AND BUILT FOR SCALE Source: Tenstorrent. ©2026 SANDISK CORPORATION OR ITS AFFILIATES ALL RIGHTS RESERVED
Slide 95: Industry Leading Performance at Scale
INDUSTRY LEADING PERFORMANCE AT SCALE Source: Tenstorrent. ©2026 SANDISK CORPORATION OR ITS AFFILIATES ALL RIGHTS RESERVED
Slide 96: HBF Roadmap: Coming Soon
HBF™ ROADMAP FIRST HBF MEMORY FIRST HBF INFERENCE DIE TAPED OUT PRODUCT SAMPLES COMING SOON ACTUAL DIE PICTURE 2027 ©2026 SANDISK CORPORATION OR ITS AFFILIATES ALL RIGHTS RESERVED
Slide 97: Bringing Everything into Focus
BRINGING EVERYTHING INTO FOCUS We delivered on our commitments Technology and innovation leadership is the foundation to delivering best-in-class products, durable long-term customer relationships, and industry-leading capital efficiency at scale AI Inference is the inflection point; customers are c
Slide 98: Sandisk
LANDISK"