Slide 1: Q1 2023 Investor Conference
Q1 2023 Investor Conference May 10th, 2023 Embedded Wisely, Embedded Widely Embedded Wisely, Embedded Widely
Slide 2: IPR Notice
IPR Notice All rights, titles and interests contained in this information, texts, images, figures, tables or other files herein, including, but not limited to, its ownership and the intellectual property rights, are reserved to eMemory. This information may contain privileged and confidential inform
Slide 3: Cautionary Statement
Cautionary Statement This presentation contains forward-looking statements, which are subject to risk factors associated with semiconductor and intellectual property business. It is believed that the expectations reflected in these statements are reasonable. But they may be affected by a variety of
Slide 4: Contents
Contents 1 Review of Operations 2 Future Outlook How PUF-based Solutions Secure 3 ChatGPT and AI 4 Q&A 5 Appendix Embedded EmbeddedWisely, Wisely,Embedded EmbeddedWidely Widely 4
Slide 5: Review of Operations
Review of Operations Embedded Wisely, Embedded Widely 5
Slide 6: Q1 2023 Financial Results
Q1 2023 Financial Results (thousands of NT dollars) Q1 2023 Q4 2022 QoQ Q1 2022 YoY Revenue 667,751 902,704 -26.0% 727,107 -8.2% Gross Margin 100% 100% - 100% - Operating Expenses 300,657 368,090 -18.3% 313,846 -4.2% Operating Income 367,094 534,614 -31.3% 413,261 -11.2% Operating Margin 55.0% 59.2%
Slide 7: Revenue across Different Streams
Revenue across Different Streams 23Q1 in comparison to 22Q4 Revenue Breakdown $1,000,000 22Q4 23Q1 Royalty 78.6% $500,000 $0 Licensing Licensing Royalty 21.4% Revenue NT$ Thousands Q1 2023 Q4 2022 Change (QoQ) Q1 2022 Change (YoY) Licensing 143,049 205,104 -30.3% 191,608 -25.3% Royalty 524,702 697,6
Slide 8: Revenue by Technology
Revenue by Technology Q1 2023 Total Revenue Licensing Revenue Royalty Revenue Technology % of Q1 Change Change % of Q1 Change Change % of Q1 Change Change Revenue (QoQ) (YoY) Licensing (QoQ) (YoY) Royalty (QoQ) (YoY) NeoBit 28.2% -30.0% -22.7% 19.7% -22.6% -19.8% 30.5% -31.1% -23.2% NeoFuse 63.9% -2
Slide 9: Royalty Revenue by Wafer Size
Royalty Revenue by Wafer Size Q1 Royalty Breakdown • 8-inch wafers contributed 47% of royalty, down 25.5% sequentially and down 9.2% yearly. 12-inch 53.0% • 12-inch wafers contributed 53% of royalty, down 24.1% QoQ but up 5.3% YoY. 8-inch 47.0% Q1 2023 Wafer Size % of Q1 Change (QoQ) Change (YoY) 8-
Slide 10: Future Outlook
Future Outlook Embedded Wisely, Embedded Widely 10
Slide 11: Future Outlook
Future Outlook Licensing & Royalty: New IP Technology & Business Development: ◼ Licensing will significantly grow during ◼ This year, PUF-based solutions will be the rest of the year, driven by PUF- adopted in 5/6/7nm CPU, DPU, AI and based security solutions. Automotive-related applications. ◼ With
Slide 12: How PUF-based Solutions Secure ChatGPT and AI
How PUF-based Solutions Secure ChatGPT and AI Embedded Wisely, Embedded Widely 12
Slide 13: Asking ChatGPT about Security
Asking ChatGPT about Security. Use PUF-based Security 1 Key generation for encryption Unique Identity (UID) and Secret 2 Key generation for authentication 3 ID generation for users Secure over-the-air (OTA) 4 firmware/software updates through signature verification Mitigate human errors by 5 hardwar
Slide 14: Major Attacks in AI
Major Attacks in AI . Poisoning Attack Backdoor Attack Poisoned training data A replaced model leads to a poisoned model gives incorrect output Poison Data Data Set Malicious Backdoor Neurons Training Merging Layer Model Result Embedded Wisely, Embedded Widely 14
Slide 15: Major Attacks in AI cont.
Major Attacks in AI cont. . Evasion Attack Stealing Attack AI model mis-performs due to modified Hacker steals AI know-how including data, Input Data or Inference Results model parameters, or sensitive information Data Set Sensitive User Data Training Training Method + = Model Model Extraction Girl
Slide 16: How to Prevent Attacks on AI
How to Prevent Attacks on AI . ◼ Each of these types of attacks requires different protection methods ◼ PUF-based security co-processor, PUFcc, can well protect AI systems Types of Attacks Problems Solution Poisoning Attack Training data may be corrupted • Signing the training data Model may be modi
Slide 17: How to Prevent Attacks on AI cont.
How to Prevent Attacks on AI cont. . ◼ Each of these types of attacks requires different protection methods ◼ PUF-based security co-processor, PUFcc, can well protect AI systems Types of Attacks Problems Security Solution Attacker can tamper input data to • Authenticate and provision users (UID) Eva
Slide 18: Securing AI with PUF-based Solutions
Securing AI with PUF-based Solutions . ◼ PUF-based IPs provide comprehensive protection to safeguard AI applications Best in-class Root of Trust Complete set of Cryptographic Engines ▪ Secure OTP Qualified Worldwide ▪ Symmetric ciphers (AES) ▪ High performance TRNG ▪ Public Key cryptography (RSA/ECC
Slide 19: Q&A
Q&A Embedded Wisely, Embedded Widely 19
Slide 20: Appendix
Appendix Embedded Wisely, Embedded Widely 20
Slide 21: Company Overview
Company Overview ▪ eMemory is the global leader of embedded non-volatile memory IP Revenue Trend (Unit: NT$ 1,000) 3,500,000 Royalty License/Design Revenue 3,000,000 2,500,000 2,000,000 1,500,000 1,000,000 500,000 0 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021
Slide 22: Quarterly Revenue Pattern
Quarterly Revenue Pattern ▪ 1st month: Receive License Fees of the month and Royalty from most foundries on previous quarter’s wafer shipments. ▪ 2nd month: Receive License Fees of the month and Royalty from other foundries. ▪ 3rd month: License Fees Only. (Unit: NT$ 1,000) 2018 2019 2020 2021 2022
Slide 23: Quarterly Number of New Tape-outs
Quarterly Number of New Tape-outs 2018 2019 2020 2021 2022 2023 (Unit: IP Numbers) 200 184 180 149 154 150 160 143 150 138 143 137 134 136 140 120 111 107 109 109 103 100 97 100 91 89 86 80 60 40 20 0 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Embedded Wisely, Embedded Widely 23
Slide 24: Worldwide Customers
Worldwide Customers ▪ Our IP solutions are adopted by leading foundries, IDMs and fabless worldwide Country Foundry IDM Fabless Taiwan 4 1 323 China 9 0 1105 Korea 4 0 96 Japan 4 7 80 North 1 1 365 America Europe 2 1 211 Others 1 0 103 Embedded Wisely, Embedded Widely 24
Slide 25: Business Model
Business Model ▪ Recurring royalty is the backbone of our business ▪ 70-75% revenue are from royalty based on wafer production ▪ More adoption = more volume shipment ▪ More advanced node wafers = higher ASP per wafer Revenue Breakdown 1-4 years 1-4 years License Fee Design License Fee Royalty Foundr
Slide 26: Technology Licenses
Technology Licenses Number of Licenses Year 2016 2017 2018 2019 2020 2021 2022 2023 Q1 License 43 20 26 18 28 32 29 9 Note: Terms (including number of process platforms and licensing fees) for each technology license are set contractually. Payments are made according to set milestones, and there are
Slide 27: New Technology Under Development
New Technology Under Development ▪ New technologies are being developed for 140 platforms by Q1 2023. ▪ 9 licensing contracts were signed. 0.11~ 0.15~ Technology 3nm 4/5nm 6/7nm 12/16nm 22/28nm 40nm 55/65nm 80/90nm >0.25um 0.13um 0.18um NeoBit - - - - - - 2 1 14 11 1 NeoFuse 2 2 1 7 11 6 17 7 2 3 -
Slide 28: Technology Development
Technology Development ▪ Developments by process nodes 12" Fabs Production Development IP Type Process Type 3nm 0 2 OTP FF 4/5nm 2 4 OTP, PUF FF 6/7nm 4 1 OTP, PUF FF, FF+ 12/16nm 8 7 OTP, PUF FF, FF+, FFC. FFC+, LPP, DRAM LP/ULP/ULL, HPC/HPC+, HV-OLED, DRAM, SOI, ReRAM, 22/28nm 44 14 OTP, PUF ,MTP
Slide 29: Embedded Wisely, Embedded Widely
Embedded Wisely, Embedded Widely For more information, please visit: eMemory Website: https://www.ememory.com.tw/ PUFsecurity Website: https://www.pufsecurity.com/ Embedded Wisely, Embedded Widely 29