GSI Technology, Inc.

Ladies and gentlemen, thank you for standing by. Welcome to GSI Technologies' fourth quarter and fiscal year 2021 results conference call. At this time, all participants are in a listen-only mode. Later, we will conduct a question-and-answer session. At that time, we will provide instructions for those interested in entering the queue for a Q&A. Before we begin today's call, the company has requested that I read the following Safe Harbor Statement. The matters discussed in this conference call may include forward-looking statements regarding future events and the future performance of GSI technology that involve risks and uncertainties that could cause actual results to differ materially from those anticipated. These risks and uncertainties are described in the company's Forms 10-Q and 10-K filed with the Securities and Exchange Commission. Many of these risks are currently amplified by and will continue to be amplified by or in the future may be amplified by the COVID-19 global pandemic. Additionally, I have been asked to advise you that this conference call is being recorded today, May 6, 2021, at the request of GSI Technology. Hosting the call today is Lee Lin Chu, the company's Chairman, President, and Chief Executive Officer. With him are Douglas Shirley, Chief Financial Officer, and Didier Lachere, Vice President of Sales. I would now like to turn the conference over to Mr. Hsu. Please go ahead, sir.

Good afternoon, and thank you for joining us today to review our fourth quarter and fiscal year 2021 results. I will cover some highlights from the year and give an update on our APU products, followed by Didier Lachere with comments on products and the sales breakdowns. Afterward, Dr. Shirley will review in detail our fourth quarter and full year physical 2021 financial results. Physical year 2021 brought several unforeseen challenges to our business. We faced a global pandemic and slowing sales from our largest customers. With that said, we stay focused on our key objectives of bringing Gemini One to market and increasing awareness of our groundbreaking technology. We made good progress on both of these fronts. Despite the challenges presented by COVID over the past year, we launched remote cloud-based data centers in Israel and the US for customers to test and demo Gemini One. The Gemini APU is receiving more media coverage, highlighting its unique advantages and benefits. We have also received third-party validation of Gemini for certain applications, including our Elasticsearch plugin, which will discuss this section. Incidentally, the system bottleneck problem facing big data is getting Heightened attention, several leading semiconductor companies have recently introduced products. They are positioned a solution to eliminate system bottleneck. The system bottleneck we refer to is the volume bottleneck caused by the separate memory and the CPU in the volume architecture. The volume architecture requires the CPU to fetch data for every operation it performs. We don't believe that any of the new solutions address the actual problem. The only way to truly eliminate the bottleneck is to break the volume model. This is what the Gemini APU does. You remove the limitation of the volume architecture. With the Gemini APU, we break the volume model and usher in a new architecture that puts processes in the memory. Today, we are working on numerous applications with our customers and demonstrating how our revolutionary architecture can reduce latency, reduce power consumption, and offer scalable, flexible solutions. When we look at the trends that favor Gemini and APU architecture, ESG and the growing awareness of sustainability will be a significant trend that favors the Earth. ESG and the power consumption are increasingly topics on the radar in the press, in the boardroom, and with investors. In ever-growing data centers where AI and big data are rapidly increasing power consumption, there is a heightened focus on finding a better solution to use less power. The Gemini APU has demonstrated a dramatic reduction in power consumption for critical application in search. One exciting example is from our aerospace and defense workshop, which we held in February, where we compared a synthetic aperture radar, or SAR, image search over a similar area size, resolution, and times. Looking at the operating power, cost over five years, the Gemini AVU used 93% less power on average and can be installed in a small portable server cabinet. This is a significant reduction from the CPU or GPU-based solution. Additionally, this solution requires much greater hardware investment. Gemini not only delivers low power consumption, but lowers the overall system cost as well, decreasing total cost of our own machine. Another compelling comparison is traditional facial recognition system that use a rear edge server with a CPU and GPU to search data and send visual vector to the backend cloud CPU and memory farm. Here, the trend model search and the final ranking occur in the GPU and the match is identified. With the Gemini APU, this process is simplified. When the Gemini APU is stored along with the GPU or GPU, search takes place on the Gemini APU, eliminating sending data to the cloud, and the match is found more rapidly. This means fewer GPU and GPU are required and latency is reduced, and again, less power is used. On previous calls, I have highlighted all software capabilities. We have finalized Gemini 1 with the leaderboard for mass production, and we are on track to begin qualification soon. The software team at GSI has demonstrated their capability in numerous applications, and are always finding new ways to improve our offering and extend our capabilities. Our technology will make a difference in how search is done, like with ElastiSearch. Using KNN search could be ideal for working with a very large database, such as those at bidding scale entries and above. Because KNN is compute-exhausted, It has been challenging to use Elasticsearch due to the constant constraint of moving the database between GPU or CPU cores. With the APU, the storage itself becomes the processor. Instead of a massive array of processing cores with cache memory close by, the APU is differentiated by performing the processing in the memory array directly. The convergence of Elasticsearch, KNN, and ABU acceleration provide less latency, more query per second, and reduce the power. I found this very exciting, and we are engaging with potential customers in the SaaS vector search space who are recognizing the strengths of ABU technology. Overall, this was a year of surprise, and we lost to the challenge. COVID is still restricting business in many ways, but we are continually advancing our technology and bringing the Gemini APU to the world's attention. Today, GSR has 119 engineers worldwide, and we have received over 100 US patents, with 51 of them for the APU. I am grateful for all of our employees' commitment to delivering excellence in product, service, and support during these challenging times. Along with my team, we are bringing our revolutionary solution to our existing customers. The solution is also opening many doors for us with the potential new customers. There's lots of excitement about our progress on Gemini One at GSI. Our S1 products continue to be essential for military and defense and the telecommunication and the networking customers. Doug will comment on our second physical quarter outlook in his prepared remarks. Now I will now hand the call over to Didier, who will highlight the sales and marketing progress on Gemini One and review our business segment performance. Please go ahead, Didier.

Thank you, Lillian. I would like to touch on two things in my comments. the third-party validation of the GSI plug-in for Elasticsearch, and the recently announced Phase 1 contract with NASA. First, a little background on Elasticsearch for those unfamiliar with the name. Elasticsearch is the most popular search engine. What makes Elasticsearch so popular is that, one, it's open source and free software, and, number two, the way it stores documents with searchable references that allow them to be searched and retrieved and enables visualization tools for visual search. Very flexible and highly functional, Elasticsearch has become the go-to search solution for scalable real-time search. Recently, a third party tested our Elasticsearch plugin powered by the Gemini APU, which demonstrated the fastest vector query speed on a one million item search compared to four other methods. At 92.6 milliseconds, The Gemini APU was 82% faster than the next fastest solution and showed a 94% improvement from the slowest response solution. They also noted that unlike other methods, Gemini supported batch queries. These are impressive performance results and it's great to see them published by an independent party. Switching to the NASA contract, we issued a press release earlier this month announcing that we were awarded along with prime contractor Space Micro, a phase one contract to develop a real-time sorting inference processing unit, also known as an IPU, an IPU board for Earth observation missions. The board will feature a radiation-tolerant Gemini APU, which is the predominant force behind the IPU for satellite applications. Using inference rather than search gives the broadest potential scope for the NASA project, which is why they release references as an APU, I'm sorry, IPU. This phase one contract is the first key step to proliferating Gemini in space by giving us access to the mission. The goal of phase one is to optimize the product portfolio and demonstrate the usability of the hardware across multiple platforms for as many possible applications. Phase one is typically a six-month process. We are also very excited that the Board will be using a radiation-tolerant chip. Once we have optimized the portfolio, we hope to move to phase two, where we will receive NRE funds to cover design and qualification. We estimate 18 to 24 months of testing and qualification before we begin to ship the product. On a different note, I am pleased to report that earlier this week, we received a purchase order for our RAD-tolerant SRAM product. The PO is for demonstration satellites anticipated to be launched this year, late this year, or early next year. Ultimately, it is for a satellite constellation. We hope to be able to see more in the coming quarters. Shifting to our sales breakdown for the fourth quarter of fiscal 2021, Sales to Nokia were $2.8 million or 36.5% of net revenues compared to $2.4 million or 28.3% of net revenues in the same period a year ago and $2.8 million or 42% of net revenues in the prior quarter. Nokia sales have been slowly improving in the past two quarters and we expect business with Nokia to stabilize this year. Military and defense sales were 22.5% of fourth quarter shipments compared to 30.9% a year ago and 26% in the prior quarter. Sigma Quad continues to be our best performing product category with sales of 52.9% of fourth quarter shipments compared to 44.7% in the fourth quarter of last year and 62% in the preceding third quarter. I'd now like to hand the call over to Doug. Please go ahead, Doug.