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

Welcome to eMemory's Third Quarter 2020 Webcast Investor Conference. Today with us, we have Chairman, Dr. Charles Hsu; President, Dr. Rick Shen; Vice President of Business Development and Spokesman, Mr. Michael Ho; Director of Investor Relations, Ms. Lijing Chen; and Director of Finance Department, Ms. Teresa Kuo. The format for today's event will be as follows. First, eMemory's Chairman, Dr. Charles Hsu, will give his opening remarks. Afterwards, eMemory's President, Dr. Rick Shen, will summarize our operations in the third quarter of 2020 followed by our future outlook. Next, Chairman, Dr. Charles Hsu, will introduce how PUF works in AI security. Then 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 2 hours after the conference is finished. Please visit the company's 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 that are subject to risk factors associated with the semiconductor and IP business. Please refer to the cautionary statement as shown on Page 3 of today's presentation. Now I would like to turn the floor over to eMemory's Chairman, Dr. Charles Hsu.

Okay. Good afternoon, everyone. I hope you are all healthy and doing very well. As mentioned in our previous conference calls, we are entering a multiyear growth cycle with a strong set of tape-outs in the pipeline. We expect the momentum of growth continue for license activities and also royalty revenues. In addition, our PUF-based solutions have gained many interest as higher security requirements have become indispensable. We are actively engaging with the leading companies for the design of our NeoPUF into their security architectures, and we expect more design wins in the future. Next, I would like to invite Rick to report our third quarter operating results and the future outlook of our business. Rick?

Thank you, Charles. Good afternoon, everyone. I will first begin with our third quarter results. Third quarter revenue was TWD 441 million, a sequential increase of 4.2% and 31.1% year-over-year, or up 5.7% sequentially and 38.2% year-over-year in U.S. dollars. The operating expenses were TWD 244 million, up 7.3% sequentially and 23.6% year-over-year, mainly attributable to expected credit losses of TWD 11.23 million, affecting the operating ratio by 2.54%. The reason is that a Chinese foundry delayed the remittance due to the tax exemption approval flow deferment. It is expected to be deferred to the fourth quarter of this year, and the expense will be reversed then. This brings us to the operating income of TWD 197 million with an increase of 0.7% sequentially and 41.8% year-over-year. Therefore, the operating margin declined by 1.6 percentage points sequentially but increased by 3.3 percentage points year-over-year to 44.7%. Overall, our third quarter EPS was TWD 2.26, and the ROE was 38.8%. For the first 3 quarters of 2020, the revenue was TWD 1,280 million, up 22.1% year-over-year and grew 26.8% in U.S. dollars. The operating expenses increased 18.1%, and the operating margin was 45.9% with an increase of 1.8 percentage points. EPS up 24.5% to TWD 6.92, and ROE gained 6.2 percentage points to 39.5%. Now let's move on to the revenue contribution by licensing and royalty. Licensing in the third quarter accounted for 25.2% of the revenue, down 5.9% sequentially, up 7.2% year-over-year, or down 4.8% sequentially and -- but up 12.9% year-over-year in U.S. dollars. Royalty in the third quarter contributed 74.8% of the total revenue, increased 8.2% sequentially and 41.8% year-over-year, or up 9.8% sequentially and 49.5% year-over-year in U.S. dollars. For the first 3 quarters of 2020, total revenue grew by 22.1% year-over-year in NT dollars or 26.8% in U.S. dollar. Licensing and royalty grew 6.8% and 28.7%, respectively, or 11% and 33.5%, respectively, in U.S. dollars. In terms of the revenue contribution by technologies, the results are as follows. NeoBit accounted for 22.4% of total licensing revenue in the third quarter, increased 22.5% sequentially but decreased 31.4% year-over-year. Its royalty accounted for 64.2% of total royalty, up 5.8% sequentially and 18.9% year-over-year, mainly due to content increase of PMIC, TDDI and other applications. NeoFuse accounted for 61.7% of total licensing revenue in the third quarter, down 18.2% sequentially but up 42.7% year-over-year. Its royalty increased by 11.4% sequentially and 128.2% year-over-year due to the continuous production of existing and new applications. This brings the royalty of NeoFuse to contribute 32.4% of total royalty. Our PUF-based security IP contributed to 5.6% of licensing revenue in the third quarter. Although this technology has not contributed to any royalty yet, engagement with industrial leaders is still actively ongoing. As for MTP technology, licensing revenue increased 3% sequentially but decreased 40.9% year-over-year to accounted -- to account for 10.3% of licensing revenue in the third quarter. Royalty from MTP increased 26.9% sequentially and 45.5% year-over-year to contribute 3.4% of total royalty. Currently, our MTP team is working with partners on developing MRAM, ReRAM and AI memory. Both ReRAM and AI memory has been verified with proven results. In the first 3 quarters of 2020, for NeoBit, the licensing revenue decreased by 8 -- by 17.5% year-over-year, but the royalty increased by 6.7%, accounting for 53.2% of total revenue. For NeoFuse, licensing and royalty revenue grew 27.1% and 133.7% year-over-year, contributing to around 41.3% of total revenue. For PUF-based security IP, licensing revenue increased 14x, 14x year-over-year, contributing to about 0.7% of our total revenue. For MTP technology, the licensing and royalty revenue declined 41.9% and 0.3% year-over-year, accounting for 4.8% of total revenue. Now looking at the royalty from 8-inch and the 12-inch wafers. 8-inch wafers, which accounted for 50.2% (sic) [ 60.2% ] of royalty, increased 1.7% sequentially and 27.6% year-over-year. 12-inch wafers contributed to 39.8% of the royalty, increased 19.8% sequentially and 70.5% year-over-year. For the first 3 quarters, 8-inch wafers accounted for 52.7% (sic) [ 62.7% ] of royalty, an increase of 16.9% year-over-year. 12-inch wafers accounted for 37.3% of the royalty with an increase of 55% compared to previous year. There were 136 product tape-outs completed in the third quarter, a record-high quarterly number, reflecting increasing demand of -- for our IPs. We will provide more information in the management report that will be released later today. In the next section, I will address our future outlook. We expect the growth of revenue to continue in the fourth quarter and beyond. For licensing revenue, the main contribution will be from NeoFuse and NeoPUF as we continuously build technology platform among foundry partners and increasing demand for design license activities. For royalty revenues, we expect both 8-inch and 12-inch royalty to continue their growth momentum. In -- 8-inch royalty will grow due to PMIC content increase in 5G smartphone and the demand pickup for automotive and IoT applications. 12-inch royalty will have strong growth as customers are ramping up production for ISP, OLED driver, networking-related such as WiFi 6, multimedia-related, DRAM, SSD controller and others. For new business development, our new applications are centered on the security business development. NeoFuse in advanced process is adopted for secure key storage to replace conventional e-Fuse. We expect this will be the trend for security requirements. Business activities of PUF-based security solutions are in progress in the applications of IoT; industrial IoT; AI; blockchain; FPGA; data processor unit; mobile storage, that is Universal Flash Storage, UFS; and automotives. Partnership with processor IP vendors, that is ARM and RISC-V-related, continue to extend our PUF-based security solution platform and bring in more consumer adoption. For new IP technology development, we are developing 6-nanometer and 5-nanometer plus N5P technology with our leading foundry partner and have already demonstrated 6-nanometer silicon results successfully. In Q3, we have announced a crypto processor, PUFiot, which integrates our root-of-trust, that is PUFrt, with crypto engines. PUFiot is targeting to provide IoT and AI chips with comprehensive but easy-to-use security solutions. We will continue to develop our PUF-based solution to implement HSM, that is Hardware Security Module, which can be embedded in chip to provide security function for network applications. Now I will pass the time to Charles.

Okay. Thank you, Rick. Please turn to Page 14. Let me take a few moments of your time to introduce to you how PUF-based root-of-trust can secure AI. Recently, many industrial developments on AI has significantly advanced the operation efficiency of many industries such as factory automation, medical applications, autonomous driving and et cetera. However, there are also many reports that show adversarial compromises on AI applications, particularly with edge devices, where the comprehensive security is unlikely due to insufficient computing power and the low cost of the device deployed. In the AI system, the security concerns are to protect the device at rest, as shown in this figure -- this Page 14, to protect the device at rest, to protect data in transit, to protect data integrity and to authenticate authorized party and to sign the authorized software. Okay. Turn to Page 15. Okay. In this page, and we -- I would like to explain the foundation of AI security, which is the root-of-trust. In order to perform these security functions, cryptographic operations on data by a secret key become indispensable. Therefore, to generate a key and to store a key and performing the cryptographic using this secret key becomes the foundation of AI security, and this is referred as the root-of-trust. Okay. So that is also shown in the Page 16. Okay. Basically, there are 3 essential problem we would like to solve, and from our IPs, how to generate key, how to store a key and how to use crypto engine to protect your data. They are all the 3 major steps which is the foundation for the AI applications. So please turn to Page 17. So by using our PUF technologies, we can do the usage for the key generation, key storage and building the root-of-trust and integrate these functions with cryptographic to perform the security operation to protect AI applications. Okay. So in Page 18, which is PUFrt, and you can see that inside the PUF, PUFrt has a function of unique ID. It has true random number generator. It also have the key storage. And in Page 19, please turn to Page 19. For the PUFiot inside the box, you can see that PUFiot actually integrate PUFrt, which is on the lower corner of the right-hand side, which integrates the PUFrt and with crypto engine, which is on the left side of the box, to perform secure boot, encryption, inclusion and authentication in addition to the key generations. Okay. So please turn to Page 20. So as illustrated in this table, so for the IoT on the -- in this table on the right-hand corner -- left-hand corner, which is operation procedure of the IoT in the IoT operation, which you have to put up the device. Okay. Then the second step is then you -- when the system is up, then you start to input the data for model training. And after the model training, then you deploy this model into the edge device. So the edge device can use the model to do the inference. So when the edge device receive the feed, the input data, and because there's model there, so you can use it, the model, to do the inference and to do the judgment for the subject you want to inference. Okay. So left hand is the operation procedure. And on the table, which shows the threat model of this -- during this operation, and we also show the countermeasure to take care of this threat model, and then we will show the security functions, which implemented to secure the operations. And finally, we lift the solutions, which uses our PUF IP to -- as a solution to take care of the security of these operations. Okay. So in these tables, okay, you can see that for the -- each operations, the PUF-based root-of-trust for the first -- first of all, you can see that the PUF-based root-of-trust can provide key and unique ID to protect the boot program and also the authentication during the boot process and also the device-to-server authentication. And for the second row and the third row in these tables, you can see the PUFiot, which integrates PUFrt with the cryptographic engine, can encrypt data in transit and data at rest to provide -- and provide the signature for the software and also provide a unique ID and also provide a tampering-resistance function for the model training and inference operations. Okay. So please turn to Page 21. So in summary, for AI applications, NeoPUF-based hardware root-of-trust provides unique ID, secure key storage, true random number generator and the anti-tempering solution to protect AI assets and operations. In addition, our solution is very -- is highly manufacturable with ultra low cost. So our technology can provide very high value proposition for secure AI applications. And in Page 22, in these figures, I also demonstrate that -- we also demonstrate that our solutions, our PUF-based crypto processor, PUFiot, can perform unique ID-generation, secure storage, protect data and software, secure boot, authentication, and also secure communication and secure update over the air and anti-counterfeit to safeguard our products throughout the entire life cycle. So it's not only in AI application but for our PUF-based solution. It can also apply to any other electronic device which need security. Okay. So this concludes my introduction on the AI security.

[Operator Instructions]

We have a question from [ Mr. Chen ] regarding our PUF solution in Chinese, and then I will translate to English. He mentioned, okay, NeoPUF currently are the only PUF manufacturer that meet the system requirement of the U.S. Department of Defense for perfect random number generator. Does this have a significant impact on the future global information hardware security?

Okay. Regarding these questions, I think the PUF requirement, basically, there's a different -- 4 different categories. One is whether the -- our PUF meet their specs; and the second is, is our PUF manufacturable; third is, is our PUF reliable. I think you have to meet all this requirement, so -- such that it can be applied to the products. So our PUF is -- amidst all this requirement, system requirement, the [ needs are lifted ]. And because the PUF is very new technologies, and I think for the security application, I think we -- it just began to be -- to get noticed from those customers who need the securities. So I think that we have started to promote our PUF-based technology in the North America. And I think we get very well assessed on our -- the features of our PUF, I think, both basically on the reliability and also stability and also particularly the availability of many foundries. So if the -- because this result, we just start to promote our technology to the big company in North America. I think if the -- eventually, if those big companies also assess our PUF for their new generation of the product, I think, basically, all the other customers, if they see that this become the popular technology for -- to take care of the security, I think it will be very important and also very -- it's also become a very significant impact to the worldwide securities applications and also security standards.

Another question is regarding the application side. Okay. The business of the -- our subsidiary, PUFsecurity, and the new [ employee ] are both developing and increasing rapidly, and the innovative business model is worth looking forward. Can you explain this core technology and the business progress in the area of key management in the cloud and the Internet of Things and the digital signature in blockchain and the security authentication in financial transition?

Okay. I think for the application, I mentioned the -- in the questions, I think our solution can all be applied to these applications because our solution is very fundamental root-of-trust. And all these applications, they do need a very high secure root-of-trust as the basis for their security applications. I think, for example, the key management, as I mentioned a few times, our PUF, one of the -- our PUF IP, we call it the PUF root-of-trust, it basically can generate the key and also store the key. And also, we can provide a true random number generation, and -- which is basically the -- which is basically can use in the key management and also blockchain signature and also the authentication of the transaction in the financial business.

Another question from [ David ]. Okay. There is currently tightening in foundry business. And so could you discuss the trend in wafer pricing for 8-inch and 12-inch? Rick? Pricing.

Okay. Regarding the trend of the wafer price for the 8-inch and 12-inch, as you know, the capacity in foundries now is quite tight. So as we learned from our customers, there is more than double-digit percentage of the wafer ASP increase in 8-inch as well as 12-inch. So because we -- as you know, the -- there's over 70% of the revenue come from royalty. So we expect the wafer ASP increase will contribute to our royalty income growth.

Yes. So because our royalty rate is [ not fixed ], so the wafer royalty we collect as a percentage of wafer price. So proportionally, we will benefit from the increase of wafer price.

All right. Thank you for the -- these remarks.

Okay. And the second question is regarding NeoPUF. Currently, it's around 6% of licensing revenue in Q3. So do we have any guidance about the licensing revenue in a year time and when NeoPUF can contribute to our royalty?

Okay. This is a very good question. As you know, [ like exactly for ] -- if you followed eMemory for years, and you will find that to be -- in general, we took 5 to 6 years to roll out our new technology to the market. So the NeoPUF start to contribute to the license revenue from last year, and we see the more and more customer adoption with our NeoPUF IP. So we expect the license revenue in next year will grow further. And then the royalty from NeoPUF, we expect to have the contribution by the second half of next year. So because we already have several customer adoption in this product, and they are doing the verification, so we expect the production will be happen in next year.

Well, okay. Another question from [ Haas ]. Can we talk about the progress in the advanced to -- no, I think it's the nanometer. And what's our target application? And the revenue -- when we expect the fast ramp. And the other one, share gains in China foundry. Okay. And the other one is our DRAM progress. I think you mentioned about the licensing and the royalty, both. And the other one will be power management-related. Okay. What's our -- I think he wants to ask is what's the content increase compared to 5G and the 4G. And the other, I think we just answered, regarding the foundry patents, which will be benefit for the better pricing and the royalty. Okay. So...

Okay. Let me answer the question one by one. For the first one is the progress on advanced 12-inch nodes, the target applications. So currently, as I report, we already marked into 5-nanometer process node. And we will -- and for the 7-nanometer, we already qualified 2 years ago. So in advanced node, we now -- we are now engaging the customer in several fields, including the AI chips, automotive and multimedia and the FPGA-related applications. So as to the revenue contribution now in the 12-inch, currently, it's mainly from the OLED and TDDI and some multimedia chip, including DTV and set-top box. And we expect the fast ramp will come from more adoption in OLED and WiFi and ISP-related product application. So we already observed our customers in the ISP product application already moved into the production stage. So I think the ISP-related product will contribute royalty very quick. Another question is about the share gain in the China foundries. For the China foundries, the -- in terms of the revenue contribution is less than 10% of our revenue. So because the China foundry still work in the legacy 8-inch-related, just legacy process [ loads ], so the contribution to [ eMemory ] on the [ percentage ] is quite low. As to the DRAM-related progress, in the 25-nanometer, our customer, DRAM customer, already moved into the production, and the wafer are produced every month. And we now also work with our customer in 20-nanometer, and we believe the project will start soon. So the royalty contribution will be happened in next year. As to the power management-related question, when the smartphone move from 4G to 5G due to the more complicated system requirements, so more PMIC will be needed. So in terms of the content, the numbers of chips, there will be 3 to 5x compared with the previous generation. But in terms of the area, the wafer area, we believe there were around 3x wafer consumption compared with 4G. So that is my answers to the questions.

So can we talk about the reason of the increase in our operating expense? Is this because of higher spend of opportunity in PUF?

Okay. So in terms of the operating expenses in the third quarter, as I reported previously, there is a tax exemption approval flow deferment caused by our China foundry. So if we exclude this factor, I think the expense is mainly caused by the manpower increase for the PUF and the PUFsecurity-related RD development efforts.

And the other one comes from [ Peter ]. Okay. Is the [ same ] PUF is gaining good traction? Can we talk about how big a percentage of our total addressable market? Because potentially, we'll adopt the solution. And I think we just mentioned the first royalty will come from next year, second half. So Charles, can you talk about how big the opportunity for PUF?

Okay. I think the -- our PUF will be mainly used for the securities. So in the future, the market size for security will be very huge, but the real number, I cannot tell. Actually, it's from very low end, like an edge device of IoT, to the very high end, like CPU data process and unit and GPU and also the automotive. They all need securities. I think if you have noticed the United States government has started to promote so-called Clean Network. So so-called Clean Network, they require every devices in order to connect it to the network has to be secure, and it has to have some level of security. And also, the devices also need to be certified as a secure device. So I think with the -- both U.S. government and also European community also promote this security requirement. I think the market size will be very big.

Another question come from him. Can we talk about the competing technology like MRAM threat to our business? Charles, can you talk about it?