Good afternoon, and welcome to eMemory's Q2 2023 webcast investor conference. Today, we have our Chairman, Dr. Charles Hsu, President, Mr. Michael Ho, Head of IR, Miss Li-Ching Chen, and the Director of the Finance Department, Miss Teresa Kuo. The format of today's event will be as follows. First, eMemory's Chairman, Dr. Charles Hsu, will give an opening remark. Afterwards, President Mr. Michael Ho will summarize our operations in the Q2 of 2023, followed by our business outlook. Next, Dr. Charles Hsu will give a talk about securities forefront, confidential computing, which is something you guys might be familiar with. We will go into the Q&A session, where our management team will be ready to answer your questions. Please feel free to submit your questions in the input box on the webcast window throughout the conference. As a reminder, this conference is being recorded, and a webcast replay will be available within 3 hours after the conference is finished. Please visit the website at www.ememory.com.tw under the Investor Relations section. As usual, before we begin, we would like to remind everyone that today's presentation may contain forward-looking statements subject to the risk factors associated with the semiconductor and IP business. Please refer to the cautionary statement on page 3 of today's presentation. Now, I would like to give the floor over to eMemory's Chairman, Dr. Charles Hsu. Good afternoon, everyone. Thank you for attending our conference call today. We expect our revenue to grow sequentially in the second half of the year due to strong licensing and accumulated new tape out, entering into the production stage. There's a very strong demand for our 5 nanometer and the 3 nanometer solutions, especially from mobile, data center, AI, and autonomous driving. We are very confident in our future growth for many years to come. Next, I'd like to invite our President, Michael Ho, to share our Q2 performance and the future outlook. Michael? Okay, thank you. Good afternoon, everyone. Now, let's begin with our 2023 Q2 financial result. The Q2 revenue was NT dollars 697 million, up 4.3% sequentially, but down 12.5% year-over-year. Operating expenses were NT dollars 328 million, up 9% sequentially, mainly attributable to the increase in bonuses and the rewards. Operating income was NT dollars 369 million, with an increase of 0.5% sequentially, but decreasing 19.9% year-over-year. The operating margin decreased by 2.1 percentage points sequentially, and decreased by 4.9 percent points year-over-year to 52.9%. EPS for the quarter was NT dollars 4.71, and ROE was 53.5%. Now, let's move on to the revenue contribution by licensing and the royalty. Licensing in the Q2 accounted for 35.8% of the total revenue, up 74.6% sequentially, and 24.4% year-over-year. Royalties in the Q2 contributed 64.2% of the total revenue, decreasing 14.8% sequentially, and 25% year-over-year. Total revenue for the Q2 increased by 4.3% compared to the previous quarter, decreased by 12.5% compared to the previous year. With that, I will comment on our revenue contribution by specific IPs. NeoBit accounted for 22.7% of total licensing revenue in the Q2, increasing 101% sequentially and up 27.6% year-over-year. Its royalties accounted for 30.5% of total royalty, down 14.6% sequentially and down 42.8% year-over-year. NeoFuse accounted for 45.4% of the total licensing revenue in the Q2, up 60% sequentially and up 20.5% year-over-year. In terms of total revenue, royalty revenue, NeoFuse royalties decreased by 15.3% sequentially and 10.2% year-over-year, accounting for 67.5% of total royalties. PUF-based security IPs contributed to 11% of the licensing revenue, increasing 66.3% sequentially, but decreasing 41.4% year-over-year. Its royalties accounted for less than 1% of total royalties, up 123.3% compared to the previous quarter, but down 18.8% compared to the previous year. For MTP technology, accounted for 20.9% of total licensing revenue, up 89.9% sequentially, and up 239% year-over-year. Royalty from MTP decreased 1.6% sequentially, and 58.4% year-over-year, accounting for 1.9% of total royalties. Let's look at the royalties for 8-inch and 12-inch wafers. 8-inch wafers accounted for 44.5% of royalties, down 19.2% sequentially, and 34.5% year-over-year. 12-inch wafers contribute 55.5% of royalties, decreasing 10.9% sequentially, and decreasing 15% year-over-year. In total, 147 product tape outs were completed in the Q2. We will provide more information in the management report. In the next section, I will address our future outlook. We expect our revenue to grow sequentially in the year's second half. Licensing revenues, we expect the licensing revenue to increase significantly compared to the first half of the year. For the royalty revenues, as new applications gradually enter the mass production stage, we expect the royalties to grow quarter after quarter for the rest of the year. Moving on to the new IP technology and business development. One, the demand for 5 nanometer design licensing is very strong. We have customers' adoption for autonomous driving, data centers, and AI-related this year. Two, we are developing 3 nanometer in several foundries with many customer requests, and continue cooperating with CPU partners for 3 nanometer confidential computing. Three, 22 nanometer emerging memory, MRAM and RRAM, completed verification with customers design-ins. This concludes my comments. Next, I will pass the time to Charles. Thank you. Okay, in this meeting, I'd like to again to introduce about the confidential computing, particularly on the GPU. Since last time I have presented the confidential computing in CPU, because the GPU now is so hot for the AI applications. We do see that the NVIDIA has promote their GPU as a confidential computing. I would like to talk about that. Let me turn to the page 13, okay? If you remember my previous talk, I mentioned confidential computing, which has become even more relevant than 2 years ago. Today, I will share with you a more detailed version of how it works, what NVIDIA is using confidential computing for, and most importantly, how exactly a PUF-based confidential computing method difference is different from others and why it is more secure. In many applications, computers deal with sensitive data or face certain data regulations, such as health and financial services. The integrity and the confidentiality of this data can be compromised by an unauthorized parties. This is why we need to establish a confidential computing. Otherwise, known as protection of data in use by performing the computation in a hardware-based trusted execution environment. Okay, you can see from the page 13, I have illustrate the three protection of the data, which is including the data at the rest, and the data in transit, and the data in use. The protection of data in use is mainly by use by the confidential computing. In the confidential computings, a trusted execution environment is deployed to provide an isolated environment, separated from the main operating system, where data can be processed without exposure to the rest of the system. Okay, please turn to page 14. To understand how confidential computing works, please look at the left graph.... For example, for CPU, has come in the CPU, the figure of the CPU has a secure and a secure zone. Application and OS on regular CPU cannot access the secure ROM. However, with the confidential computing, sensitive data from the apps and OS can be stored in the secure zone, under individually isolated TEE on the CPU. Now, please turn to page 15. Another example of confidential computing is in a graphic processing unit, GPU. GPU has its own hardware root of trust to enable an isolated TEE for protecting the integrity and the confidentiality of data and the code during the transfer from CPU to GPU, also processing in the GPU. Based on the NVIDIA's H100 data sheet, it shows that NVIDIA use confidential computing for their AI training. As you can see in this, view graph, during a call, and the data transfers from CPU to GPU, the multiple individual TEEs are created through the GPU's hardware root of trust. That the unauthorized entities are prevented from accessing the AI application data and the code. The TEE, trusted execution environment, okay, TEE, needs to have a robust and hardware-based security, which we call it the hardware root of trust, to ensure the following actions. One is to do the device authentications. For example, in this, view graph, there's, there are 2 parts. One is on the top, is the CPU and many users use a CPU, and on the bottom is a GPU. Okay? When the communication begin or the processing begin, then the CPU has to start to talk, communicate with the GPU. First of all, the device authentication need to be done between CPU and the GPU. Okay, that is, when the user and the device are communicating with the GPU, they have to confirm if the firmware and the software of the GPU are secure. The GPU also needs to authenticate the firmware and the software from OS or apps to check that they have not been tampered with. The second function is to guarantee the data integrity and the confidentiality between the CPU and the GPU. Okay? A better way to understand this is that the data being transferred will be encrypted and decrypt with encryption key exclusively to the CPU and the GPU. The data transmit from the CPU and the GPU will be protected. This view graph demonstrate the security confidential computing between the CPU and the GPU. Please turn to the next page. We know that in order to set up the In order to set up the confidential computing environments, we have to we have to know that how to use, identify... Since we have to set up the extent, trusted execution environment, we have to identify that the trusted execution environment is basically need to be accomplished by the hardware root of trust. From the page 16, the hardware root of trust in a TEE can help protect the integrity and the confidentiality of the data and the costs. It consists of a key storage, key generation, and the key protections, and the cryptographic engines, all combined for the comprehensive securities. Let's look at this view graph. Our PUF-based crypto processor, we call it the PUFcc, can fulfill the role of all functions mentioned above. Our secure OTP, consisting of both our anti-fuse technologies and PUF-based technologies. creates secure key storage that hides the key and there, way, way above from the potential attackers. Furthermore, the PUF, our PUF-based TRNG can efficiently generate truly random keys to protect the data and the code. From this view graph, you can see that for the hardware, Root of Trust, basically consists of two parts. One is the key managements, and one is cryptography. In the key management, key management also consists of key storage and key generations. For key storage, we, we have OTP, and we have also PUF plus OTP, to create the Secure OTP and also the ultra-secure key storages. For the key generations, and we, we use the PUF-based TRNG to create a high speed and high entropy key generations. I will focus on talking about the key generations and the key storage, since our fundamental technology have achieved a very high secure key storage and the key generations by compared to the conventional methods. Our PUF-based OTP helps the industry migrate to a very high level secure security regime by improving the security of key, key storage and the key generations. In the conventional key storage, by using eFuse, which is basically has not very secure because of the following two reasons. First of all, the burning of the... Because the eFuse is by burning of the fuse. eFuse require the fuse burning to determine the key whether it is one or zero. Please refer to page 17, and you can see the low, low, low, left side. You can see that the eFuse, conventional eFuse need to when the data is original 0, and they need to be, the fuse need to be burned and become 1. When the fuse is burned, it was broken. This method will cause the eFuse can be easy to reverse engineer by looking at which fuse was burned. There's also the risk of the fuse going back, because the thin piece of metal can sometimes reattach itself, which causes the key to disappear. The second problem by using eFuse is because eFuse the size of eFuse is very large, so it cannot not overcome the density over 8 Kbits. Due to... Because if the density become larger, and the yield will become lower. As the technology migrates to more advanced node, eFuse is in efficient, insufficient to provide the key storage and also other security functions. With our anti-fuse OTP, we call it NeoFuse, we will not face any issues with reverse engineering, and also, we will not face the size limitation and even key disappearing. We use an electron, because we use electron tunneling method at the transistor level, which can shrink with the process node, and it cannot be seen under a microscope. Combining our OTP with our PUF-based technologies, makes key storage even more secure. When used with PUF's unique value, different from chip to chip, the key's address can be scrambled and stored in a unique, randomized location on the secure OTP to make the key storage ultra secure. Let's turn to page 18. After I talk about the key storage, the next I'm going to talk about the PUF-based key generation. Key generation, the conventional random, true random number generator used for hardware root of trust is not PUF-based. By comparing PUF-based true random number generation generator versus conventional true random number generation in the 3 categories, PUF-based is superior for the following 3 reasons. First of all, the PUF-based, it has a higher entropy. Okay, PUF-based TRNG has much better and a much higher entropy, because it is derived from the PUF seeds. Which is a perfect random sources, make it truly random. Conventional TRNG uses digital circuit to generate random number, which has, have a low entropy. For example, our entropy is actually about 100 times than the conventional TRNG. The second advantage of our PUF-based TRNG is high speed. Another critical feature of TRNG is speed. For PUF-based TRNG, the speed can generate a high volume of keys, which is about 100 times faster than conventional TRNG. Because the conventional TRNG require a lot of more post-processing, which is slow the speed. The third advantage is about the power consumption. In our PUF-based, the TRNG, we have less power consumption. The power consumption with a PUF-based TRNG is also around 100 times lower. Again, this has to do with the post-processing that conventional TRNG requires, so that will leading to more cost and energy use. After compared to the advantage of our PUF-based key generation and the PUF-based TRNG, next, now, I'd like to make the summaries. For the confidential computing, in the future, it is a must for GPU and the CPU application because it, it need to, the confidential computing, need to provide the integrity and the confidentiality of the data and the code. As compared to conventional root of trust, eMemory's root of PUF, PUF-based root -- hardware root of trust, provide the best quality of a unique identity and a secure key storage, and a much higher speed key generation for CPU, GPU, and DPU to facilitate their confidential computing. All this, the differences mentioned above are why confidential computing architectures are moving away from conventional solutions toward PUF-based. For example, Arm, for example, is putting our solution as a reference design for its Armv9, confidential computing. With our developments, reaching the most advanced process nodes, we expect increase in adoption of our security solution in the future. This conclude my remarks. Next, we will enter the Q&A section. Thank you, Charles. This concludes our prepared statement. We will now begin the Q&A session. Please submit your questions in the input box on the webcast window, and you will enter a queue. Should you wish to raise your question in Chinese, we will translate it to English before our management team answers it. All of our questions will follow the format of answering the Chinese version first, followed by the English version. We will now collect the questions and begin our Q&A session. I will hand the mic over to Li-Ching. [Foreign Language] Speaking in Mandarin. Okay. The question that was asked was: What's your view on the operating margin for the second half of the year? Michael? It will be a few% higher than the first half of the year. Thank you. speaking in Mandarin. ... [Foreign Language] 后 量 子 的 加 密 演 算 法 的 主 要 问 题 是 在 精 密 的 ， 密 钥 的 大 小 非 常 巨 大 ， 可 能 比 现 行 使 用 的 RSA 和 ECC 的 密 钥 大 上 几 千 ， 几 千 倍 。 我 们 uh，PathBase 的 Root of Trust 的 强 ， 强 项 在 于 产 生 密 钥 的 长 ， 长 度 与 速 度 ， 是 目 前 全 世 界 最 容 易 、 最 快 速 ， 也 却 最 安 全 的 方 式 。 对 于 客 户 采 用 我 们 的 solution 来 因 应 这 样 ， 这 种 后 量 子 加 密 演 算 法 将 变 得 呃 ， 绝 对 必 要 。 上 一 次 呃 ， 大 规 模 更 换 加 密 演 算 法 是 在 2000 年 左 右 ， 当 时 美 国 决 定 采 用 AES 算 法 ， 各 大 相 关 企 业 光 是 将 部 分 加 密 方 式 采 用 新 的 方 法 就 花 了 十 年 。 所 以 这 一 次 的 加 密 演 算 法 转 换 也 应 该 差 ， 也 差 不 多 需 要 有 十 年 。 对 公 司 来 说 ， 呃 ， 这 是 非 常 正 面 ， 会 加 速 客 户 转 换 成 我 们 的 solution。 The question was: how will the post-quantum encryption algorithm affect the company? Charles. Encryption uses crypto algorithms to protect sensitive information, including secure website or even emails. The widely use of public key encryption system relies on algorithms difficult for even the fastest confidential, conventional computers to solve in limited time. However, quantum computers will threaten existing public key encryption within a decade. The National Institute of Standards and Technology, and NIST is expected to announce new algorithms for encryption in 2024. The main problem with the potential post-quantum encryption algorithm is that the key size is very long, which may be thousands of times longer than current RSA and ECC keys. The length of the eMemory PUF-based Root of Trust lies in its flexible and high speed key generation, which is currently the world's easiest, fastest, and safest method. It will be necessary for customers to adopt PUF-based hardware, hardware Root of Trust to face post-quantum crypto algorithm. The last large scale replacement of encryption algorithm was around year 2000. At that time, the United States decided to adopt the AES algorithm. It took 10 years for major related companies to switch to the new encryption methods, so we expect that changing to a new encryption algorithm will also take around 10 years. This is very positive for speeding up customers adoption of our solutions. [Foreign Language] 这 个 电 源 管 理 IC PMIC 市 场 ， 现 在 很 糟 ， 相 当 低 迷 。 再 加 上 TI 杀 价 竞 争 。 请 问 一 下 ，TI 是 不 是 公 司 的 客 户 ？ 如 果 不 是 的 话 ， 那 TI 杀 价 竞 争 会 不 会 抢 到 既 有 PMIC 公 司 的 单 ， 进 而 影 响 公 司 未 来 权 力 金 的 收 入 ？Michael. [Foreign Language] 嗯 ， 确 实 景 气 不 是 太 好 哦 ， 那 所 以 常 常 听 到 有 一 些 杀 价 竞 争 的 新 闻 。 那 TI 目 前 不 是 我 们 的 客 户 ， 我 们 PMIC 的 客 户 事 实 上 很 多 ， 但 是 主 要 的 权 力 金 的 贡 献 是 来 自 于 像 是 uh，Dialog 或 者 其 他 那 种 非 传 统 PMIC 的 公 司 ， 那 像 是 以 processor 的 公 司 自 制 的 PMIC 来 搭 配 销 售 ， 或 是 有 像 手 机 厂 自 制 的 PMIC 这 些 为 主 。 那 像 这 一 类 的 客 户 的 话 ， 他 们 的 年 产 的 wafer 量 事 实 上 还 蛮 大 的 ， 大 概 有 七 、 八 十 万 片 ， 而 且 他 们 导 入 的 PMIC 的 数 量 是 逐 年 在 增 加 。 这 些 客 户 取 代 了 原 本 呃 ， 部 分 的 一 些 IDM 大 厂 ， 像 是 呃 ， 提 到 的 这 个 TI， 那 所 以 像 TI 的 这 种 杀 价 竞 争 ， 我 们 认 为 是 不 会 呃 ， 影 响 我 们 的 ， 我 们 这 一 类 呃 ，PMIC 的 公 司 ， 因 为 这 些 公 司 大 概 都 是 代 工 厂 最 top tier 的 客 户 ， 那 除 了 PMIC 之 外 ， 他 们 还 有 很 多 大 量 的 先 进 制 成 的 产 品 在 代 工 厂 生 产 ， 所 以 他 们 拿 到 的 这 个 wafer 价 格 大 概 也 是 最 优 惠 的 。 所 以 即 使 前 几 年 代 工 厂 涨 价 的 时 候 ， 这 些 客 户 的 代 工 价 格 并 没 有 什 么 太 多 的 变 化 ， 而 且 他 们 的 销 售 都 是 以 搭 配 销 售 或 者 是 自 用 为 主 ， 不 会 ， 比 较 不 会 受 到 市 场 竞 争 的 影 响 。 那 他 们 权 力 金 的 起 伏 主 要 会 受 到 呃 ， 像 是 终 端 需 求 或 者 是 库 存 调 整 的 这 个 影 响 。 每 一 家 客 户 的 状 况 不 会... [Foreign Language] 不 太 相 同. 我 们 看 到 的 是 最 先 调 整 库 存 的 客 户 已 经 开 始 在 投 片, 明 年 陆 续 也 会 恢 复 正 常 的 一 个 投 片 水 准. 我 们 认 为 在 PMIC 的 权 力 金 在 明 年 会 有 比 较 强 劲 的 成 长, 除 了 因 为 库 存 调 整 的... [Foreign Language] 低 基 期 外, 过 去 我 们 2 年 我 们 有 陆 续 导 入 了 像, 韩 国 手 机 大 厂 的 PMIC, 像 是 DDR5 的 PMIC, 以 及 更 多 的 客 户, 以 及 应 用 端 采 用 像, 像 是, 车 用, data center, wearable. 加 上 我 们 一 些 U.S. 的 大 客 户 的 PMIC, 也 都 从 8-inch 的 像 是 0.18, 0.13 的 制 程 转 到 12-inch, 像 是 90/55nm 的 制 程. 这 些 都 会 带 动 每 一 颗 权力金 的 上 涨. 我 们 觉 得 说 在 PMIC 的 部 分, 这 个 部 分 带 来 的 权力金 还 是 会 持 续 增 加 的. Thank you. For power management IC, the market is quite sluggish, coupled with TI's price competition. Is the TI customer of the company, and if not, will TI's price competition grab orders from existing PMIC company customers and affect the company's future royalty income? Michael. TI is not currently our customers. We have many PMIC customers. The main royalty contribution comes from like a Dialog and the non-traditional PMIC companies, many from processors or system companies. This type of customers annual wafer production volume is around 700K to 800K pieces. They have replaced some of the chips of the traditional IDM, like TI. TI's pricing competition will not affect such companies, because they are all top tier customers of foundries with the lowest foundry cost. Even when foundry raised the wafer price in the past years, it didn't affect theirs. Their PMIC products are mainly used in-house or bundle sales with processors. Their royalty was affected primarily by the end market demand and the inventory adjustments, and as such, customers' situation differs. The first customers to adjust inventory have already started placing wafer orders. The rest will gradually return to normal production level next year. We believe that PMIC's royalties will grow strongly next year. In addition to the low year-over-year base comparison, new customers gains like a Korean smartphone penetrating into EV cars, data center, wearable through the key customers, as well as the ASP enhancements through the customer migrating into more leading process node. Our U.S. customer PMIC migration, migrating from 8 inch to 12 inch, will all make the PMIC royalty growth strongly next year. Thank you. [Foreign Language] 请 问 ， 请 问 公 司 目 前 有 没 有 AI， 运 算 晶 片 ， 就 是 process 方 面 的 客 户 ？Charles. [Foreign Language] AI 的 定 义 非 常 广 泛 ， 我 想 目 前 市 场 上 最 care 的 是 指 的 是 NVIDIA 跟 AMD 这 两 家 公 司 做 的 大 型 语 言 训 练 模 型 ， 这 两 家 过 去 都 曾 经 跟 我 们 接 触 过 ， NVIDIA 是 从 2020 年 就 开 始 ， 主 要 是 希 望 能 够 导 入 这 个 PUF 跟 OTP 到 他 们 的 这 个 security 的 架 构 ， 后 来 没 有 放 进 去 ， 是 因 为 那 时 候 我 们 4nm 跟 5nm 的 这 个 验 证 还 没 有 完 成 。 在 这 个 硬 体 信 任 跟 hardware Root of Trust 的 key storage， 他 们 目 前 用 的 是 eFuse， 如 同 我 前 面 提 到 的 ， eFuse 在 key storage 上 面 是 不 安 全 的 ， 必 须 要 用 这 个 OTP 或 是 说 PUF， 加 上 OTP 来 提 供 非 常 安 全 的 这 个 key storage. 在 这 个 硬 体 信 任 跟 就 是 hardware Root of Trust 的 这 个 key generation， 他 们 目 前 用 的 也 是 这 个 conventional 的 TRNG， 并 没 有 加 上 PUF， 所 以 使 得 他 们 这 个 产 生 key 的 这 个 乱 度 跟 速 度 都 会 远 远 不 及 我 们 所 提 供 的 这 个 PUF-based 的 这 个 TRNG. 因 为 我 们 的 这 个 产 生 的 乱 度 几 乎 是 这 个 conventional 的 TRNG 的 100 times. 他 们 同 样 的 ， 他 们 传 统 的 这 个 TRNG， 它 消 耗 的 这 个 电 量 也 是 我 们 的 100 times， 因 为 它 的 速 度 比 较 慢. 特 别 在 未 来 在 这 个 量 子 远 算 法 所 需 要 的 这 个 钥 匙 长 度 大 幅 加 长 ， 会 大 幅 增 加 运 算 时 间 ， 所 以 我 们 高 速 的 key generation 的 特 性 是 非 常 有 优 势. 我 们 很 有 信 心 ， 随 着 我 们 已 经 到 开 发 到 这 个 3nm 的 进 展 顺 利 的 话 ， 目 前 造 成 这 更 多 的 量 产 的 记 录 ， 未 来 还 是 这 些 NVIDIA 或 是 AMD， 未 来 还 是 很 有 机 会 能 够 用 这 个 PUF-based 的 solution. 当 然 ， 除 了 这 两 家 主 要 的 晶 片 供 应 商...... [Foreign Language] 我 们 在 United States ， 在 China 及 United States 云 端 业 者 的 自 制 晶 片 上 面 ， 今 年 我 们 会 有 客 户 导 入 。 广 义 的 AI 如 果 涵 盖 edge computing 的 话 ， 我 们 已 经 有 不 少 的 客 户 ， 我 们 的 solution 面 积 小 ， 而 且 用 很 简 单 的 方 法 就 能 产 生 基 本 这 个 security 的 功 能 ， 对 于 这 个 edge computing 是 非 常 有 竞 争 力 的 。 The question was: Do you have any AI processor customers? Charles. The definition of AI is very broad. I think that most people care about now is the large-scale language training models made by NVIDIA and AMD, both of which has made contact with us in the past. NVIDIA started reaching out in 2020, and we have been discussing how to implement our OTP and PUF into their security architecture. The reason have not been, the reason not for not being adopted was that our 4 and 5 nanometer, at that time, did not complete the qualification yet. For the key storage and the hardware Root of Trust, they currently use eFuse. As I mentioned earlier, eFuse is not safe for key storage, OTP and PUF, and OTP or PUF plus OTP must be used to provide key for provide the secure key storage. For key generation of hardware Root of Trust, they currently use the traditional or conventional to random number generator without using PUF, so that the entropy and the speed of key generation is inferior to what we provide in PUF-based TRNG. In fact, our random, randomness is about 100 times than theirs. They have also consumed the 100 times of power for the key generations, especially in the future, the length of the key required by post-quantum encryption algorithm will be greatly increased, which will also increase the calculation time. Having high speed key generation such as ours, is very advantageous. We are very confident that, with the progress of our 3nm and more production records, we have great opportunities in the future. In addition to these two major chip supplier, we also have cloud customers doing in-house AI processor in China and the U.S. this year. Furthermore, if AI, in a broad sense, also covers the edge computing, then we already have many customers. Our solution is, only a small area and can generate a basic security function in a simple way, which is very competitive in edge computing. [Foreign Language] OK, 另外一题也是跟这个有点关，但是又有点不大一样，就是在 AI server 里面，它不是只有 processor，它还是有其他晶片。请问公司的 IP 在里面，有没有 CPU、GPU 或是其他晶片？ Michael 请回答。 [Foreign Language] 最 近 的 AI 话 题 很 热 。 目 前 所 谓 的 这 个 AI server 可 能 就 是 指 像 是 H100 系 列 的 这 样 的 server， 在 里 面 会 有 机 会 用 到 我 们 的 IP 的 应 用 。 已 经 tape out 或 有 量 产 的 ， 有 像 是 SSD controller， 像 CXL memory controller， retimer， 还 有 在 像 是 那 个 DDR5 里 面 ， DIMM 里 面 的 PMIC 跟 SPD Hub， 这 部 分 都 会 用 到 我 们 的 IP. 目 前 我 们 在 G，CPU 跟 GPU 的 部 分 还 没 有 导 入 ， 但 是 等 我 们 的 3nm 的 OTP 验 证 完 成 ， 应 该 会 有 机 会 。 特 别 是 像 晶 片 公 司 如 果 有 用 到 Arm 的 solution 的 话 ， 都 应 该 会 用 到 我 们 的 IP。 谢 谢 。 This is also a question related to AI. Which various AI server related applications have adopted the company's IPs? Does this include CPU and GPU? Michael. The application inside the AI server adopting our IPs with tape outs or in production are SSD controllers, CXL memory controllers, retimers, PMIC and SPD hub for DDR5 DIMM card. The CPU and the GPU has not yet adopt our IP, but we have a good, good chance when we complete the 3 nanometer OTP verification, especially for those which adopt Arm solution. Thank you. [Foreign Language] 还 是 跟 AI 有 关 ， 就 是 我 们 以 前 有 提 到 ， 就 是 说 ， 第 一 个 用 我 们 PUF 的 客 户 是 AI。 好 ， 那 为 什 么 他 们 先 用 呢 ？ Charles. [Foreign Language] 哦 ， 呃 ，OK， 呃 ， 因 为 新 ， 新 ， 先 进 的 AI model 都 一 般 都 会 存 在 这 个 ， 呃 ，commodity 的 flash 当 中 ， 而 这 列 ， 类 的 标 准 储 存 装 置 ， 因 为 本 身 没 有 security 的 功 能 ， 所 以 就 制 造 了 被 偷 窃 或 是 改 ， 修 改 的 可 能 性 。... [Foreign Language] 利用 PUF 作为 基础 的 硬体 安全 信任 Root of Trust IP and the Crypto Coprocessor IP, 如 PUFcc, 可以 用来 加密 认证 这些 AI models, 保护 其 避免 被 修改 或 偷窃, 这 也是 PUFsecurity 跟, yeah, eMemory 在 这 IP, Secure IP 方面 的 利基. 此外呢, 这个 AI 的 processor 里面 基本上 都 会 有 大量 的 SRAM. 那 这个 SRAM 的 density 越来越 高 的话, 那 这个 里面 有 不好 的 broken bit 的 机会 就 更多. 一般 像 目前 我们 用 在 DRAM 的 做 修补, 这个 DRAM 的 broken bit, 这个 OTP 呢 一样 也 可以 用来 作为 这个 修补, 这个 AI 里面 这个 large density SRAM 的 broken bit 的 修补. 因为 假如 是 用 eFuse 的话, 因为 eFuse 的 density 超过 8K bit, 就 慢慢 就 开始 变得 density 很大, 而且 eFuse 变大 之后, 因为 area 变大 之后, 它 也 有 它 自己 本身 的 也 有 issue. 所以 当 这个 SRAM 的 density 变大 的 时候 呢, 那 用 eFuse 来 修补 就 不 feasible. 我们 的 OTP 比 eFuse 的 density 呢, 在 同样 面积 上面, 我们 有 大 到 500 到 1,000 倍, 所以 对于 AI 内含 大量 density SRAM 呢, 就 变成 非常 必要. 基于 这个 SRAM 修补 跟 security 的 requirement 呢, AI 晶片 有 相当 大 的 比例, 会 把 eFuse 转成 我们 的 这个 OTP. This question is also related to AI. You mentioned that AI-related applications are amongst the few early adopters of PUF, which is the case if you attended our meetings. Why are they the first to adopt PUF, Charles? Okay, in advanced AI model are typically store in conventional commodity flash memory. This standard storage device has no security function. It can be stolen or modified. Our PUF-based hardware security Root of Trust IP and Crypto Coprocessor can encrypt and authenticate these AI models to protect them from being stolen or modified, which is why customers need the PUFsecurity and the eMemory's solutions. In addition, all processors have a SRAM. One type of... SRAM is one type of volatile memory, just like a DRAM. When SRAM is manufactured, it will have a bad bit, bad bit that need OTP for memory repair. In the past, eFuse was used to do SRAM repair functions. However, the SRAM density in AI chips is much bigger, which a eFuse cannot fulfill due to the density limitation. Our OTP offers a 500 to 1,000 times bigger density than eFuse. Therefore, for AI with a large density SRAM, our IP become a necessity based on SRAM repair and the security requirements, most AI chips will convert from eFuse to our OTP. [Foreign Language] 成 熟 制 程 很 多 都 在 China，过 去 联 案 都 面 临 China 客 户 权 利 金 收 取 的 问 题，未 来 如 果 成 熟 制 程 的 客 户 都 转 往 China，公 司 是 不 是 也 会 收 不 到 权 利 金？Michael. [Foreign Language] 好 的 问 题 哦 。 呃 ， 我 们 目 前 授 权 给 中 国 有 九 大 晶 圆 厂 ， 那 来 自 中 国 的 这 些 终 端 客 户 也 高 达 九 百 多 个 ， 那 整 体 的 这 个 占 公 司 的 营 收 大 概 是 十 到 十 五 个 percent。 我 们 在 过 去 并 没 有 发 生 过 像 积 欠 权 利 金 或 者 是 应 收 账 款 这 些 问 题 ， 这 都 是 来 自 于 说 ， 呃 ， 这 跟 IP 的 本 质 有 很 大 的 关 系 。 像 ARM 的 话 ， 它 是 提 供 soft IP， 它 是 给 客 户 这 个 RTL， 就 会 很 类 似 像 在 晶 片 设 计 的 一 个 软 体 ， 那 授 权 金 跟 权 利 金 的 收 取 对 象 都 是 终 端 的 晶 片 公 司 。 那 如 果 他 们 在 下 一 个 晶 片 没 有 授 权 而 再 度 使 用 的 时 候 ， 那 IP 公 司 就 只 能 靠 像 市 场 调 查 或 者 是 查 账 audit 这 样 的 方 式 ， 才 能 够 确 保 ， 呃 ， 权 利 金 的 正 确 。 但 我 们 是 把 这 个 IP 的 技 术 ， 呃 ， 最 底 层 的 电 晶 体 架 构 授 权 给 代 工 厂 ， 然 后 由 代 工 厂 按 照 这 个 royalty rate 跟 wafer 的 这 个 数 量 来 支 付 我 们 权 利 金 。 那 对 ， 呃 ， 晶 片 公 司 来 说 ， 我 们 只 收 这 个 基 本 的 设 计 的 授 权 金 。 那 由 于 ， 呃 ， 中 国 的 晶 圆 代 工 厂 只 有 九 家 ， 然 后 ， 呃 ， 晶 片 客 户 有 九 百 多 家 ， 所 以 相 对 之 下 ， 我 们 要 去 收 取 跟 查 账 就 容 易 很 多 了 。... [Foreign Language] 加上，我们持续在代工厂都有新的技术开发，以及新的客户的订单导入. 如果代工厂不付权利金的话，我们这些已经设计好的设计订单，然后没有量产的客户的产品，就会有这个侵权的问题而不能生产，这 对 代 工 厂 来 说 损 失 会 更 大. 在权利金收取相对来讲就比较不会有这样的问题. 另外，在 embedded NVM 的风险主要是来自于说生产之后，如果记忆体的储存或读取有问题，那么整个晶片或者是甚至是整个系统都不会，都会无法运作，产生的问题跟损失会更大. 一般来讲，这客户不敢贸然去使用这种非法取得的这个 IP, 谢谢. Many mature processes are located in China, and in the past, even Arm faced problems of collecting royalties from Chinese customers. If customers all move their mature products to Chinese foundries, will eMemory face trouble receiving royalties? Michael. Foundries, customers, and over 900 chip customers in China, accounting for about 10%-15% of our total revenue. There have been no problems with collecting royalties and other payments. This is related to the nature of IPs. Arm offers soft IP, providing customers RTL, is similar to software for chip design. They collect the license and the royalties from chip companies. They must audit their customers if customers use the same RTL to other chips without being licensed. We license the fundamental transistor to the foundries. The foundry pay the royalties based on the royalty rate and the wafer volume, while the chip companies pay design license fee only. Since there are only nine foundries, but more than 900 chip customers, customers in China, it's much easier to collect and audit the counts of foundries than the chip companies. In addition, we have continuously licensed our technology to foundries for new process nodes and new applications, and continuing tape outs through the end customers adoption. The contract will be breached if the foundry fails to pay the royalties. As a result, all customers tape out in the pipeline cannot go for production. The loss to the foundry will be much bigger than not paying royalties, not to mention the infringement of lawsuit penalty. The risk of embedded non-volatile memory is that there is a problem with the memory reading after production, the entire chip or even the system will shut down, and the resulting penalty for damage will be much bigger. Customers don't want to risk illegally obtaining the IPs. Thank you. [Foreign Language] 请 问 Q2, 请 问 MTP 的 授 权 今 在 Q2 为 什 么 会 成 长 那 么 多? 那 主 要 大 概 是 哪 些 应 用? Michael. [Foreign Language] 工 厂 的 产 能 利 用 率 在 这 一 两 季 普 遍 来 讲 都 蛮 低 的 。 开 发 这 些 特 殊 的 制 程 ， 特 别 是 的 这 个 embedded non-volatile memory 来 取 代 外 挂 的 记 忆 体 ， 这 是 可 以 提 供 代 工 厂 的 这 个 晶 片 的 价 格 。 加 上 很 多 应 用 ， 因 为 速 度 啊 ， 像 功 耗 或 者 是 空 间 的 问 题 ， 逐 项 会 导 向 这 个 embedded NVM 的 应 用 。 我 们 的 MTP 是 逻 辑 制 程 相 容 的 ， 所 以 代 工 厂 可 以 在 不 用 额 外 多 投 资 设 备 ， 就 可 以 轻 易 地 将 现 有 的 逻 辑 制 程 capacity 扩 充 给 NVM 的 产 品 使 用 。 这 对 代 工 厂 来 讲 是 一 个 很 划 算 的 技 术 投 资 。 此 外 ， 像 这 一 类 的 授 权 也 需 要 代 工 厂 投 入 R&D 的 资 源 。 当 景 气 好 的 时 候 ， 代 工 厂 会 忙 着 做 生 产 ， 但 现 在 因 为 生 产 线 有 空 ， 技 术 的 开 发 就 可 以 加 速 进 展 。 因 此 ， 我 们 认 为 我 们 的 MTP 的 相 关 技 术 也 开 始 进 入 了 这 个 多 年 的 成 长 循 环 ， 预 计 在 明 年 就 可 以 看 到 明 显 的 权 利 金 的 成 长 。 谢 谢 ！ Why has MTP's license fee increased in Quarter 2? What applications have adopted our MTP technology? Michael. The utilization rate of foundries is very low. There are incentives for foundries to develop the specialty technologies, especially embedded non-volatile memories, to increase wafer value. With the requirement of high speed, low power, and small form factors, replacing external memory with embedded memory is accelerating. Our MTP technology is compatible with the logic process, so the foundry can easily expand the existing logic process for embedded non-volatile memory without investing in additional equipments. With fab's low utilization rate, foundries are allocating more resources to speed up the technology development. We believe our MTP-related technology have started to enter a multi-year growth cycle. We'll have a significant royalty growth next year. Thank you. [Foreign Language] 听 说 你 的 竞 争 对 手 Synopsys 把 OTP 跟 PUF 跟 EDA tool 绑 在 一 起 销 售 ， 甚 至 不 用 钱 。 请 问 这 对 公 司 造 成 什 么 样 的 影 响 ？Michael。 [Foreign Language] Synopsys 大 概 在 2018 年 并 购 了 我 们 一 些 竞 争 对 手 ， 像 是 Sidense 跟 KiloPass， 所 以 大 概 已 经 有 5 年 了 。 过 去 确 实 我 们 也 常 常 听 到 客 户 端 有 提 到 说 ， 哦 ， 他 们 用 bundle EDA 的 tool 一 起 卖 ， OTP 的 价 格 用 很 低 ， 甚 至 是 不 用 钱 的 ， 大 家 都 大 概 都 有 听 过 ， 这 种 不 用 钱 的 会 最 贵 哦 。 所 以 我 们 听 到 客 户 这 边 并 没 有 因 为 这 样 而 转 用 到 Synopsys 的 solution， 我 们 也 没 有 因 此 而 改 变 了 我 们 的 定 价 策 略 。 反 而 我 们 听 到 的 是 有 越 来 越 多 使 用 Synopsys 的 客 户 开 始 转 用 我 们 的 IP。 最 近 听 到 的 反 而 是 说 ， 啊 ， 因 为 他 们 的 过 去 的 杀 价 策 略 没 有 效 果 ， 所 以 反 而 提 高 了 售 价 。 PUF 的 话 ， 是 一 个 更 fundamental 的 一 个 不 同 ， 像 Synopsys 用 的 是 SRAM PUF， 它 的 performance 是 完 全 没 有 办 法 跟 我 们 的 NeoPUF 来 相 比 的 。 Synopsys 的 security 部 门 是 并 购 了 很 多 公 司 组 成 的 ， 他 们 的 策 略 ， 啊 ， 主 要 是 focus 在 系 统 整 合 这 个 security 的 方 面 ， 跟 我 们 在 主 攻 全 世 界 最 好 的 Root of Trust 的 是 有 所 不 同 的 ， 谢 谢 。 This person asked, "I heard that Synopsys, your competitor, bundles OTP, PUF, and EDA together and even offered it as a free IP. Will this affect eMemory?" Michael. has acquired our previous main competitors, Sidense and KiloPass, for around 5 years. In the past, we often heard our clients say that they bundle EDA tools to sell together, and the OTP price was very low or even free to get our customers to switch. However, no customers have ever switched because of this, and we haven't changed our pricing strategy either. Accurately, more and more customers using Synopsys has switched back to our IPs. The latest news is that Synopsys raised their price because previous price killing strategy was ineffective. PUF is even more fundamental different. Synopsys use the SRAM PUF, and the performance is incomparable to our NeoPUF IP. Synopsys' security department is composed many of many companies acquired through the M&A. Their strategy focus on the security aspect of system integration, which is differs from our focus on offering the best root of trust in the world. Thank you. [Foreign Language] 请 问 跟 Arm 的 合 作 什 么 时 候 才 会 有 贡 献 ？Michael。 [Foreign Language] Arm 的 合 作 事 实 上 是 一 直 有 持 续 在 进 行 的。 我 们 已 经 有 客 户 采 用 我 们 的 security IP, 跟 Arm 的 CPU 同 时 使 用, 也 做 了 tape out。 其 中 还 有 其 他 的 客 户 也 正 在 design in 中。 在 今 年 开 始 就 已 经 有 授 权 金 的 贡 献。 谢 谢。 I see someone asked this question, and the question is, "When will you start seeing contributions from your collaboration with Arm?" Michael. With Arm has been ongoing, some customers have already tape out our security solution through Arm's platform and contribute a license fee already. Thank you. [Foreign Language] 做 到 12nm 的 ISP。 另 外 ，ISP 的 OTP 的 density 如 果 从 2K 上 升 到 16K 的 话 ， 那 对 权 利 金 的 产 值 是 不 是 倍 数 成 长 呢 ？Michael。 [Foreign Language] ISP 的 部 分 呢 ， 我 们 确 实 已 经 有 客 户 已 经 导 入 这 个 12 nm， 也 做 tape out， 那 目 前 已 经 有 做 小 量 的 一 个 量 产 。 那 由 于 12-inch 的 wafer 的 价 格 大 概 比 22/28 nm 的 贵 上 大 概 40% 左 右 ， 如 果 既 有 的 22/28 nm 的 客 户 转 到 12 nm 生 产 的 话 ， 那 晶 片 大 小 如 果 不 变 ， 那 之 间 的 差 异 最 主 要 就 是 在 于 这 个 wafer 价 格 的 上漲 ， 那 这 跟 使 用 多 少 density 的 OTP 是 没 有 相 关 的 。 谢 谢 。 Does eMemory do 12 nanometer ISP? In addition, if the OTP of the ISP rises from 2K to 16K, will the output value of royalty be multiplied? Must have adopted our IP in 12 nanometer ISPs with small volume production. Since the price of a 12 nanometer wafer is about 40% higher than that of a 22 and 28 nanometer. If existing 22, 28 nanometer customers switch to 12 nanometer, chip size remains the same, then the royalty difference will be the increase of a wafer pricing, which has nothing to do with how much density is used. Thank you. [Foreign Language] CXL 架 构 下 ， 力旺 的 IP 扮 演 什 么 角 色 ？Charles. [Foreign Language] CXL 是 AI server 里 面 对 所 有 的 processor，CPU、GPU、DPU 之 间 的 这 个 data 传 输 协 定 ， 对 于 安 全 要 求 等 级 很 高 ， 我 们 的 CE 的 CXL memory expander controller 的 客 户 导 入 我 们 的 这 个 Root of Trust， 用 来 保 护 这 个 资 料 传 输 的 安 全 。 Under the CXL framework, what role does in-memory IPs play? CXL is the protocol to enhance the memory capacity sharing between CPU and the other devices, such as the GPU and the DPU, all CXL memory expander controller customers adopt our Root of Trust to protect the security of the data confidentiality and the integrity during the data transfer. [Foreign Language] OK, DDR5 除了 PMIC 之外, 其他的 part 像是 RCD 跟 DB、SPD Hub, 有没有用到另外的 IP? 哦, Michael. [Foreign Language] RCD and DB currently do not use our IP. In the SPD Hub part, we will use our NeoEE to replace external E2PROM and integrate features like Temperature Sensing. Currently, we see many customers developing this type of memory with 100,000 read/write cycles. These cases are mostly for SPD Hub use. In a DDR5 DIMM module, there will be two chips. One is PMIC, and the other is SPD Hub. Both will need to use our MTP. Thank you. Other than PMIC, do other parts of DDR5 also need to use your IPs? For example, register clock driver, data buffer, and SPD Hub, etc. Michael? RC, DB are not currently used, but SPD Hub have adopt our NeoEE for the 100K endurance. Inside the DDR5 DIMM card, there are two chips, one is the for the SPD Hub and the other for the PMIC, we'll use our MTP solution. Thank you. [Foreign Language] eFuse 在 28nm 以 下 会 在 微 缩 上 出 现 问题. 现 在 先 进 技 术 ， 先 进 制 程 而 言 ， 如 果 客 户 eFuse 会 有 意 外 烧 断 或 是 不 烧 断 ， 它 会 长 回 去 ， 哦 ， 产 生 yield issue， 还 没 有 采 用 eMemory 的 solution 的 时 候 ， 大 部 分 是 采 用 什 么 的 方 案 ？Michael。 [Foreign Language] 有 听 到 客 户 的 一 个 回 馈 ， 就 是 说 客 户 如 果 用 eFuse 而 出 现 这 个 意 外 烧 断 的 情 况 的 时 候 ， 一 般 就 是 会 用 倍 数 的 容 量 来 做 修 补 或 ECC. 这 样 的 部 分 的 话 ， 就 会 占 掉 太 多 的 面 积 ， 尤 其 是 对 于 这 个 先 进 制 程 来 讲 ， 这 个 wafer 非 常 的 贵 ， 所 付 出 的 代 价 就 很 高 。 另 外 的 话 ，eFuse 需 要 大 电 流 来 烧 ， 这 对 于 线 路 上 的 设 计 就 有 一 定 的 难 度 跟 风 险 。 因 为 说 eFuse 目 前 都 是 晶 圆 厂 提 供 的 ， 如 果 有 问 题 的 话 ， 晶 圆 厂 就 会 负 责 。 在 先 进 制 程 上 ， 目 前 都 还 是 以 eFuse 为 主 。 另 外 一 种 技 术 ， 就 是 所 谓 的 anti-fuse 以 我 们 的 这 个 NeoFuse 跟 竞 争 对 手 来 为 主 。 竞 争 对 手 的 IP 在 以 前 ， 在 过 去 的 这 个 先 进 制 程 上 有 一 些 issue， 目 前 没 有 听 说 有 什 么 客 户 在 用 ， 这 也 影 响 了 这 个 客 户 对 于 这 一 类 技 术 的 一 个 信 心 ， 这 也 使 得 我 们 在 过 去 这 几 年 的 导 入 比 较 缓 慢 。 随 着 我 们 的 量 产 技 术 逐 渐 地 累 积 ， 加 上 我 们 的 IP 都 已 经 完 成 了 验 证 ， 在 5nm 已 经 跨 过 ， 而 且 现 在 也 在 导 入 3nm 的 这 个 验 证 。 我 们 以 过 去 的 经 验 来 看 ， 我 们 取 代 这 个 eFuse 的 这 个 部 分 只 是 时 间 早 晚 的 一 个 问 题 。 谢 谢 。 eFuse cannot shrink below 28 nanometers. However, in terms of advanced manufacturing processes, if the customer's eFuse is actually accidentally blown out, what kind of solution will be adopted if the eMemory solution has not yet been implemented? Michael. ... use the eFuse, and it blows up accidentally, it will usually increase the eFuse area. However, it will take up too much space and would be very expensive for advanced process chips. In addition, eFuse cannot accommodate a circuit design that needs to burn in a large current. However, since the eFuse is provided by the foundry, if there is any problem, the foundry will be responsible. eFuse is still the dominant in the advanced manufacturing process. Another type of technology is the anti-fuse. Many are NeoFuse and the competitors solution. However, the IP of our competitors have experienced the problem in the advanced manufacturing process. At present, no heard about any customer adopt their solution in leading process, but it also affect the customer's confidence and slow down the adoption of this technology. With more production records and successfully moving into the five nanometer and the three nanometer, we are confident to replacing eFuse. It's only a matter of time. Thank you. Due to time constraints, this will be our last question. Lijing? Michael? The last question is: why is your royalty ratio of revenue much higher than other IP companies, and why has it been like this for such a long time? Michael. The IP companies mainly collect, license fees, but without or a little, royalties. Our strategy is to charge less on the licensing in the early stage, but insist on collecting the royalties. Thank you. Thank you. We will now begin the closing comment. Charles, can you please proceed? Of conference today. For more information about our PUF-based security IP and the technology, we encourage you to visit our PUFsecurity website at www.pufsecurity.com, and check out our articles and other materials. Thank you again for your patience and support for eMemory. We will continue to work hard on our technology and IP innovation, and also PUF-based hardware security solutions for our customer and bring higher returns to our shareholders. Thank you. Thank you everyone for joining us today. You may now disconnect. Goodbye, and have a good day.