eMemory Technology Inc. (3529) Earnings Call Transcript & Summary

Good afternoon, and welcome to eMemory's Second Quarter 2024 Webcast Investor Conference. Joining us today is our Chairman, Dr. Charles Hsu, President, Mr. Michael Ho, Head of IR, Ms. Li-Jeng Chen; and the Director of the Finance Department, Ms. 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, followed by our business outlook. Next, Dr. Charles Hsu will give a talk titled a must in security, 100x faster, PUF-based true random number generator. Then we will conclude today's conference with a Q&A session, where our management team will 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 after the conference is finished. For more information, please visit the company's website 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. And thank you for attending our conference call today. As mentioned in the previous quarters, our company has entered a multiyear growth cycle. Last quarter, we shared that we have licensed our technologies to almost every foundry worldwide, with more than 600 process platforms. This number of process platforms continues to grow every year. Within this process platforms all technologies developed from OTP to more sophisticated security IPs and the MTP expanding to -- extending various kind of emerging memories. The royalties received per wafer for this technology will increase as the process nodes, the memory functions and the functions of the chip security IPs all become more advanced. Later, I will introduce one of our security IP PUFtrng, which is PUF-based True Random Number Generator which is 100x faster than the widely used conventional one. It leverages our PUF as the perfect source of randomness. This shows that we can continue to use our existing technology and IPs to develop new IPs with additional functions. Not only does this shows the importance and the competitiveness of our intrinsic technology, but it also forms the most important foundation for our long-term growth allowing us to continue innovation based on our core competencies. Therefore, we are very confident about our future. Next, I will invite our President, Michael Ho to share our second quarter performance and the future outlook.

Good afternoon, everyone. Now let's begin with our 2024 second quarter financial results. The second quarter revenue was TWD 893 million up 11.2% sequentially and up 28.2% year-over-year. Operating expenses were TWD 398 million up 4.1% sequentially and 21.3% year-over-year. Operating income was TWD 495 million with an increase of 17.7% sequentially and an increase of 34.3% year-over-year. Operating margin increased by 3.1 percentage points sequentially and increased by 2.6 percentage points year-over-year to 55.5%. Our net income amounting to TWD 475 million experienced an increase of 10.3% sequentially and 35.1% year-over-year. EPS for the quarter was TWD 6.36 and ROE was 67.3%. Next, let's move on to the revenue contribution by licensing and royalty. Licensing in the second quarter accounted for 33.6% of the total revenue, up 31.3% sequentially and up 20% year-over-year. Royalties in the second quarter contributed 66.4% of the total revenue, increasing 3.3% sequentially and increasing 32.8% year-over-year. Total revenue for the second quarter increased by 11.2% compared to the previous quarter and increased by 28.2% compared to the previous year. For the first half of 2024, the licensing and royalty revenues are as follows: licensing, in the first half accounted for 31.1% of the total revenue, up 34.4% year-over-year. Royalties in the first half contributed 68.9% of the total revenue, increasing 20.2% year-over-year. Total revenue for the first half increased by 24.3% compared to the previous year. With that, I will comment on our revenue contribution by specific IPs. NeoBit accounted for 24.6% of total licensing revenue in the second quarter, increasing 54% sequentially and increasing 30.2% year-over-year. Its royalties accounted for 26.7% of total royalty, up 11.3% sequentially and up 15.9% year-over-year. NeoFuse accounted for 33.3% of the total licensing revenue in the second quarter, down 16.7% sequentially and down 11.9% year-over-year. In terms of total royalty revenue, NeoFuse royalties increased by 0.2% sequentially and increased by 39% year-over-year accounting for 70.6% of total royalties. For PUF-based Security IPs contributed 12.5% of licensing revenue increasing 105.2% sequentially and increasing 36% year-over-year, while its royalty accounted for less than 1% of total royalties. For MTP technology, accounted for 29.6% of total licensing revenue, up 110% sequentially and up 69.9% year-over-year. Royalty from MTP increased 11.6% sequentially and increased 86.8% year-over-year, accounting for 2.7% of total royalties. For the first half of 2024, the revenue by technologies are as follows: NeoBit licensing revenue increased 43.4% year-over-year and royalty increased 1.4%, accounting for 24.9% of the total revenue. For NeoFuse, licensing revenue increased 19.3% and the royalty increased 27.3% year-over-year, contributing to 62.3% of the total revenue. For PUF-based Security IP, licensing revenue increased 26.3% year-over-year, while the royalty contribution was less than 1%, accounting for 3.3% of total revenue. For MTP technology, licensing revenue increased 64.3% year-over-year, and royalty revenue increased 75.7% accounting for 9.5% of total revenue. Now let's look at the royalties for 8-inch and 12-inch wafers. For 8-inch wafers accounted for 42.5% of royalties, up 2.6% sequentially and up 26.7% year-over-year. For 12-inch wafers contributed 57.5% of royalties, increasing 3.8% sequentially and up 37.6% year-over-year. In total, 171 product tape-outs were completed in the second quarter. We will provide more information in the management report. In the next section, I will address our future outlook. For licensing revenues, licensing revenue will continue its growth momentum due to the strong demand from both foundries and chip companies. For royalty revenues, we will expect the royalty sequential growth in the second half due to the new products ramping up. Now moving on to the IP -- new IP technology and business development for new IP technologies, One, NeoFuse is developing in the FinFET HV process to meet customers' next-generation OLED DDI plans; two, RRAM is expanding into more processes with the increased customer demand; three, NeoFlash continues progress in specialty process, replacing embedded flash and external NOR flash. Four, developing 2nm technologies with leading foundries. For the business development platform, one, new CPU architecture for security IP will start to contribute to revenue; two, successfully integrate NeoFuse for SRAM repair with EDA company. This concludes my comments. Next, I will pass the time to Charles. Thank you.

Okay. I will take time to introduce our very fast TRNG, which we developed in these couple of years, which is must in securities. And it is 100x faster TRNG than the conventional TRNG. The role of True Random Number Generator, we call it TRNG is crucial in fortified secure system against increasing the sophisticated attacks. Today, I will discuss why having just a TRNG is not enough. You need high-speed and high-quality TRNG. Before understanding the role of the randomness in protecting a secure system, it is essential to know the methods used to attack these systems. This slide will explain why True Random Number Generator is necessary for hardware security. There are 2 ways to break in a secure system. The first is by cryptographic analysis, we call it a crypto analysis. A technique existing for centuries where attackers us uncover weaknesses in cryptographic algorithms. And over time, such attack have driven the evolution of cryptographic standards with older algorithms like the data encryption standard we call it the DES replaced by newer ones such as Advanced Encryption Standard, we call it AES. The second method is implementation attacks, which target specific weaknesses within system's implementation, exploiting the weaknesses to discover secrets within the secure system. Now assuming that we have a well-designed secure system that is resistant to crypto analysis and the implementation attacks. How will an attacker be able to break in the system? In this case, they will attempt to guess the secret key. As obtaining it would allow them to access a lot of confidential information from the system. To prevent an attacker from correctly guessing the secret key, we must minimize its probability. The first factor is key length which is predefined by the crypto algorithms. For example, if the key length is 128-bit it means that there are 2^128, two of the power of 128 numbers of possible key combinations that can be generated. The second factor is ensuring random generation of the key so that each combination is equally probable. If you refer to this diagram, you can see that after the user encrypt their secret with randomly generated key, the probability of an attacker guessing the correct combination is reduced to 1/2^128, therefore, to achieve this result and reduce the likelihood of an attacker guessing the correct secret key we must use TRNG. And in the next slide, I will explain why it is important to have a high-speed TRNG. By high speed, we are referring to a high throughput capability of generating a large number of random numbers in short time. The speed and throughput of TRNG matters because large systems often have numerous applications and hardware components that require and consume random numbers. Many of these components cannot generate their own random numbers and rely on a hardware-based TRNG. Given the extensive number of components, the throughput of the TRNG is vital to ensure numbers are generated quickly and efficiently. Besides generating large volume of rendered numbers, a high throughput and high-quality TRNG can help prevent side-channel attack. Side-channel attacks are one of the most popular form of attacks, where attacker exploits leaked information from source such as the power consumption or electromagnetic emissions to discover keys. To counter this attack the new random numbers must be generated continuously to mask the side-channel information. The middle figure shows a simplified example of crypto operation protected by masking. In a masked operation, the random bit transformed the input data into 2 independent sets and are sent to 2 separated mask operations compared to a single operation, separating them ensures there is no leakage. Since the operation continuously consume random numbers, the throughput of TRNG is also important in this use case. Beyond security, there are other applications that are highly reliant on random numbers. One example is banking where many keys or pin numbers require high-quality random numbers to ensure the security of accounts. Because the number of accounts and the transaction are enormous, a highly throughput TRNG is essential as well, as shown on the third figure at the bottom. Having described the importance of high-throughput and the high-quality TRNG, it is clear that our solution, PUFtrng meets this criteria. A unique feature of our TRNG is that it has 2 entropy sources unlike conventional TRNG that relies on a single source. Our PUFtrng combines dynamic entropy from ring oscillators with static entropy from our NeoPUF chip fingerprint. Normally, dynamic entropy operates slowly due to the time-consuming process of collecting natural fluctuation, leading to low throughput in conventional TRNGs, as illustrated using the dashed lines. The throughput in this case is usually less than 1 megabit per second. To overcome this limitation, PUFtrng uses high-quality random bit generated by PUF to refine the output of the dynamic entropies. Typically, it takes the dynamic entropy source a long time to gather natural fluctuations, however, by using PUF for refinement, we can achieve a high throughput by reducing the collection time of the dynamic entropy, yet still achieving high-quality output. As shown in the figure, PUFtrng delivers throughput approximately 100x greater than conventional TRNG while maintaining similar power consumption. We also create innovations to better show the difference between PUFtrng and the conventional TRNG. In Figure 1 to 4, we invented a simple to topology optimization inside the circle to determine the randomness of the random number generated. The uniformity of the lines inside the circle indicates the randomness of the number. If the numbers are truly random, the circle begins to feel uniformly as small random number are generated. On the other hand, if the randomness of the generated number is poor, the line inside the circle will not be uniformly distributed. So for the dynamic entropy generated by ring oscillator, as shown in Figure 1 within the blue circles, a single blue line represents a 10-bit random number. The dynamic entropy constantly generates random data. However, its output quality is not enough for secret keys because the line tends to appear at the same positions indicate that the randomness of the random number is poor. And for the figure 2, it is a conventional TRNG usage post-processing. Here, the random numbers are generated by gradually accumulating more random bits from the dynamic entropy over time and after going through post-processing, the output will slowly become uniform and complete as shown in these animations. And by looking at the Figure 3, which is a static entropy generated by PUF. Static entropy is instantly ready and does not change over time. The lines are very uniform, indicating the result of very high quality from the start. And by looking at the Figure 4, which is a PUFtrng using PUF refinement. In the PUFtrng, the generated numbers from the uniform lines, indicating high-quality output. Additionally, the speed of the generation is much faster than the conventional TRNG. This is evident from the fact that the output generated by the lines of PUFtrng appears much faster than in conventional TRNG, highlighting the difference in the generation speed. So another way to understand this to imagine conventional TRNG as a classic cars. They are functionable but not efficient, they are either slow or consume a lot of gas when running fast. In comparison, PUFtrng is like a new engine car, it runs fast by consuming low power. Lastly, another advantage of PUFtrng is both the dynamic entropy and the static entropy are implemented in a macro and prequalified across various technology nodes. This ensures superior quality compared to RTL-based TRNGs provided by other IP vendors or customers. So in conclusion, the integration of high-speed, high-throughput TRNG is vital for maintaining system security and the functionality. Our PUFtrng solutions stands out with its ability to generate high-quality random numbers efficiently. This not only protects against sophisticated attacks, but also supports large-volume random number generation. With PUFtrng, we are setting a standard for the high-quality TRNGs. This concludes my remarks. Thank you very much for your time. And next, we will enter the Q&A section.

[Operator Instructions] 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.

[Foreign Language]

[Foreign Language]

So the first question was price cutting competition among Chinese foundries is very serious. Have you been affected? Michael?

In recent years, China has been actively expanding its production capacity by building the new foundries and the new fab areas for mature process, many of which are using our logic NVM technologies. In terms of the royalties from China, the capacity expansion and the increasing penetration rate will drive overall growth even if foundry wafer price declines.

[Foreign Language]

[Foreign Language]

The second question was, recently Trump said that TSMC should pay the United States for protection? Will this have an impact on eMemory? Michael.

Our IP is licensed to foundries around the world, including the U.S., such as Intel and Global Foundry. With our platforms established globally we have a widespread range of technologies and in various process, making us less susceptible to political influence.

[Foreign Language]

[Foreign Language]

The question was, are you seeing a trend of customers incorporating AI capabilities into their designs? How does this affects your company? Michael?

As far as we know, we already have customers who have incorporated AI functions into their designs and are moving to tape out in more advanced processes for functions such as recognition and image processing. As mentioned in the last earnings, AI application systems involve data input, data/model storage and computing accelerators. Currently, our IPs are adopted for applications related to data input through various sensors. For data storage in NAND and DRAM, customers are using our IPs for SSD controller and CXL memory interface. In terms of computing, which is mainly in advanced processes, we have customers adopting our root of trust IPs and SRAM repair IPs. These developments are expected to drive our future licensing and royalty growth.

[Foreign Language]

[Foreign Language]

So the question was about the lecture just now. Chairman Hsu just mentioned that eMemory's true random number generator is the fastest in the world. If this is the case, why haven't major high-performance computing vendors used it. Charles?

Our true random number generator is based on our OTP and PUF technologies. This is a hard IP and must be qualified in each process. The number of our qualified process continue to grow and progressing towards the most advanced nodes. As more customers adopt our technology and accumulate production records. And as hackers use faster computing to launch attack, the market will demand faster random number generator. We are working with our CPU partners to promote this technology and are very confident about its future development.

[Foreign Language]

[Foreign Language]

The question was, what is the reason why PUFs royalties have not increased significantly. Can Charles please answer this question?

Licensing for our top technology has been increasing. The first customer to adopt our PUF was China's largest chip company, however, due to U.S. government sanctions we were unable to continue receiving royalties from their production. Despite this, we have more than 60 tape-outs gradually entering the mass production stage and will soon contribute to our royalties. Thank you.

[Foreign Language]

[Foreign Language]

The question was, what is the progress in 2 nanometers and 3 nanometers. Is it possible that after eMemory's IP is ready for 3 nanometers, customers will already start to migrate to 2 nanometers. Can Michael, please answer this question.

Our development of 3-nanometer and 2-nanometer has always aligned with the requirements of the most advanced foundry process and the customers. Currently, some customers have already started design in for 3-nanometer, while 2-nanometer is still in the early development stage, progressing much faster than the previous years. Even if our technology is not adopted by the first generation, there will still be future opportunities due to the process migrations. For example, even in the mature nodes, we still have hundreds of new products using our technologies every year.

[Foreign Language]

[Foreign Language]

So this next question is about MTP. MTP licensing fees and royalties have experienced the most significant growth. In which application is it mainly used? How do we expect MTP to contribute to the company's future? Can Michael, please answer this question?

Our MTP technologies include comprehensive product lines that meet different specifications ranging from like a high endurance NeoEE and medium-density NeoMTP, which require no additional masks to high-density embedded fresh such as NeoFlash and RRAM. The recent growth in MTP is driven by the adoption of NeoEE in the PMIC and SPD ICs on DDR5 as well as the NeoMTP in the new 4-color e-paper and electronic shelf label, ESL, driver ICs. Currently, technology licensing to foundries and the adoption across different applications are accelerating. The royalty rate of MTP is higher than the OTP so its contribution to our revenue will become increasingly significant.

[Foreign Language]

[Foreign Language]

The question is about DRAM. So they're asking what is our progress in DRAM. Can Michael, please answer this question.

Our OTP is mainly used for repair function in the DRAM. It has already been implemented on the multiple process platforms of several customers and is continuously being developed for more advanced process recently with the mass production of a new process of DRAM customers. The royalties have increased significantly.

[Foreign Language]

[Foreign Language]

So this next question is, Michael just mentioned that you guys have 3-nanometer customers? And what is the application of your first 3-nanometer customer? And when will we start to see contribution. Can Michael please answer this question.

Our first 3-nanometer customers are primarily in the data center server applications. They require not only OTP, but also the PUF-based security to protect the data. They will contribute to our revenue in the second half of this year.

[Foreign Language]

[Foreign Language]

So the Chairman mentioned in previous quarters that SRAM repair density is increasing, providing more opportunities for our IPs. What is our progress in this area? And when can we expect to see more contributions, Michael?

As the demand for AI and high-performance computing increased, the density of embedded SRAM is also growing. Therefore, the demand for using OTP for SRAM repair continues to rise. We have seen an increase in the adoption of our solutions by more customers and products, resulting in more royalty, which we foresee will be a trend and the contribution will increase significantly in the future.

[Foreign Language]

[Foreign Language]

This next question is for Michael. When will OLED DDI 28-nanometer mass production begin?

As early as 4 years ago, our OTP was adopted in the 28-nanometer OLED DDI products by many customers. These products have a gradually entered the production and continue to contribute to royalties. The penetration rate of OLED DDI in mobile, tablet and notebooks keep on rising and moving towards more advanced process, driving the royalty growth. Currently, some customers are moving to 22-nanometer OLED DDI accumulating good production records and are developing products for the next generation of FinFET HV process. We are optimistic about our continued growth in this market.

[Foreign Language]

[Foreign Language]

So the next question is although the recent blue screen crash event at CrowdStrike seems unrelated to security, CrowdStrike Endpoint Security is also touted as Zero Trust. How can eMemory compete with software companies using a hardware approach? Can Charles, please answer this question.

With the Zero Trust framework, eMemory's hardware security provides hardware, root of trust and security computation capabilities. For cybersecurity software companies, hardware security can enhance the security of their application. eMemory's security IP aims to boost the competitiveness of these software companies. Not to compete with them.

[Foreign Language]

[Foreign Language]

The next question is what role does eMemory play in Google's OpenTitan framework, Charles?

The OpenTitan platform sets the standard for securing data transmission between IoT device and the cloud through hardware security. Our security IP, being a hardware security solution, can be directly applied to Google OpenTitan platform for encryption and decryption.

[Foreign Language]

[Foreign Language]

So given the talk that the Chairman gave about TRNG, there's still someone online who is curious about True Random Number Generators. So the question is, it looks like that TRNG can provide random source for generating secret keys and obfuscating to help resist against attacks. Does it imply that having high-speed TRNG and security design is more important than having a PUF? Can Charles answer this question?

I think in the security, very high security systems the high-speed TRNG is very important as well as PUF because in our invention, the high-speed TRNG need to rely on the PUF as an input to the True Random Number Generators. In addition to that, PUF also played a very important role as a unique ID and this unique ID, you have to find the number, which is very unique and come from the natural randomness. And by using this natural randomness the unique ID can also create the mother security key for the system. So in the very high secure systems, the path and the high-speed TRNG are very important. Thank you.

As we have answered most of the questions. And in the interest of time, we will now begin the closing comments. Charles, can you please conclude the presentation.

Okay. Thank you once again for your patience and the support for eMemory. We will continue to work hard on the technology and IP innovation and also the PUF-based hardware security solutions for our customers and bringing higher return for our shareholders. And for more information about our PUF-based security IP and technology, we encourage you to visit our PUF security website on pufsecurity.com and check out our articles and other materials.

Thank you, ladies and gentlemen. Please be advised that the conference recording will be accessible within the next 3 hours. Thank you, everyone, for joining us today. We hope you will join us again next quarter. You may now disconnect. Goodbye, and have a good day.