CEO. I should mention that SES AI's safe harbor statements can be found on their website. Also, this fireside chat may not be reproduced or a written transcript distributed without the express written consent of Water Tower Research. Qichao, thanks so much for doing this. I really appreciate it. Thank you, Joe. For those who are not familiar with SES AI, maybe a quick overview would be helpful. Sure. We started back in 2012 to develop a very high energy-density lithium metal battery. We were the first to enter automotive A sample and B sample using lithium metal, and we're on track to C sample. Recently, we decided to bring this material discovery platform that we demonstrated success for lithium metal to other chemistries, including silicon lithium-ion, LFP, and sodium, to other chemistries and other applications. We also launched this platform called Molecular Universe. This platform, just in the last two months that we launched it, has been very popular. Let's talk about Molecular Universe in a minute. More broadly, you're going to a multi-strategy business model. You're going to have batteries, and you're going to have an innovative software, AI-driven platform. Does that radically change the kind of trajectory of where you guys are headed? Yeah, I know it sounds complicated to explain, but actually it all comes together and it's all connected. If you think of SES AI as like a tree, at the trunk, the foundation is the material discovery platform, Molecular Universe, which we have used and then demonstrated success for high energy-density lithium metal and silicon. In terms of the branches, we have one branch that is electrolyte. We actually produce and sell the electrolyte that users find using Molecular Universe. Also, drones. We have commercialized lithium metal and high silicon lithium-ion batteries using the electrolytes discovered by Molecular Universe for drone applications. We sell these batteries and also energy storage. We also sell the LFP and sodium batteries using electrolytes discovered by Molecular Universe for the energy storage market. While it sounds complicated, it really is one cohesive strategy. We do believe that by having this really strong software and hardware together, we're building a very strong moat, a very defensive strategy and portfolio. Actually, the revenue that we're getting and also expect to get from the materials, the drones, batteries, the energy storage batteries, and the subscription to the Molecular Universe software actually might make us probably the first, the first battery company in the U.S. to actually break even, and much sooner than we expect. Break even is also always a great thing to have. In the fourth quarter of 2024, you saw some revenues that were booked from Molecular Universe. I think those from your OEM partners. Were these the first real deals for Molecular Universe? Yeah, so back then we did use Molecular Universe to make progress in terms of discovering additives for silicon lithium-ion cells. That's different from the lithium metal B samples that we were working on. Also, other electrolyte companies were using this platform to come up with new formulations and materials. Great. Now, I know along the way you're obviously gathering a bunch of data in the course of your development. How are you using this data for the different markets you're trying to target? We actually don't gather customer data. The way we do it, it's a great question. Most great AI companies have their own data. If you look at all the big AI companies, they definitely have their own data. In the battery field, the AI companies in this field don't have their own battery data, and that's been a big disadvantage. We actually generate our own data. We build cells, we build batteries, and then we systematically test, synthesize different materials, different molecules, and then build batteries with different cathode-anode cell chemistry combinations, different electrolyte formulations, different charge and discharge environments, temperature. Every half a year to a quarter, we will go through a massive round of these different cell designs and then collect the data from the testing and then use the data to actually train this model. We actually systematically generate our own data, and that, to us, has been the cleanest and the best label, best quality data to train the model. In terms of customers, we don't take data from the customers. It's like a data privacy and security. We have this Molecular Universe that's on the cloud that's trained purely on our own data. We will have child Molecular Universe models that will be installed on-premise to each of the customers, and then that version of the model gets trained on the customer's own data. That version never gets sent back to the cloud or gets shared with other customers. Right. Okay, I get it. You really want to have customers own their own data. Yeah. Now, recently, you rolled out version, what was it, 0.5 of MU? Yes. Maybe you can give us a little color on, you know, just the regular adoption with M U and the adoption of 0.5. Yeah, so we rolled out the original version, zero, Molecular Universe-0, back in April. Since then, we've had several thousands of free users, and tens of large battery companies. Actually, most of the major battery companies are trial testing enterprise versions, and they really liked it. The feedback has been quite phenomenal, and people are impressed with the output, the accuracy, how relevant the answers are compared to just general models like OpenAI or Grok, or some of the more battery-specific models. One feedback they had was, okay, Molecular Universe 0 was great at answering questions. Can it actually solve problems? There's a difference between answering a question and actually solving a problem. The key feature we rolled out in Molecular Universe 0.5, this new version is called Deep Space. In the original version, we have this feature called Ask. Basically, you ask a question and you get an answer. It's based on a very advanced large language model that's fine-tuned specifically for the battery domain. This new Deep Space, when you ask a question, instead of answering the question right away, it will ask questions about the question to get a much more comprehensive and complete picture of the context. For example, if you ask a question about, I am using A11 cathode and then silicon anode, and I'm using a particular electrical formulation, I want you to recommend a new formulation that can improve low temperature fast charge, like that. Deep Space will first ask you, okay, are you looking for a more academic answer or a more practical industrial answer? What about cost? Is cost a factor? How fast do you actually want your fast charge? How low is your temperature? Basically, it will ask more questions. You answer these questions, and then it will go away to think. It will think for about 20 to 30 minutes. It does take much longer than the previous version, but then it will provide solutions. It will provide several solutions that are very practical that can be implemented right away. This is quite powerful. For example, in the traditional R&D process, a human scientist will basically start with a question, and then go through literature, trying to find state-of-the-art answers, and then innovate on top of the answers by finding new materials to replace, and then build cells and then test and then validate. That takes six years, sometimes even eight years. A lot of traditional R&D is basically spending resource on all the stuff that don't work. The definition of R&D is basically like trying things that don't work. If we can cut this and just go straight to the thing that works, that will save companies a lot of money. This Deep Space will provide you solutions in about 20 to 30 minutes that you would normally take six years or even more to get these answers. This is possible because of three things. One is, we are using a multi-agent model compared to just one single LLM. This will also call our molecule database, and it will call our cell performance data. We have this multi-agent LLM that's connecting this material database with a cell level performance database. Actually, this linking of material properties with cell performance properties, this is something that battery companies have been trying to do, but so far, no battery company has been successful at doing. I think Molecular Universe 0.5 is probably the closest at this. We continue to improve and get closer to this goal. I know that sounds quite innovative. Now, you said you mentioned enterprise customers were starting to get interested. Can you give us an idea of, is MU 0.5 going to accelerate enterprise interest? Is that key to them going forward? Yes, and I think this Deep Space, so back in Molecular Universe 0, it was all just, it was mostly a very smart chatbot. It would answer your question about any battery stuff. With Deep Space, it's starting to provide you with solutions, things that you can take and then go implement within six months in the line. Now this is becoming much more practical now, and we are getting more battery companies and also taking a much more serious look at this. They are actually getting ready to invest significant dollars at this effort. Actually, a lot of the battery companies, to our surprise, have internal plans to reduce R&D costs and also reduce R&D timing. Most of these battery companies spend 5% of their revenue on R&D, and 80%- 90% of R&D is spent on doing things that don't work. They definitely want to cut that part, and that will save labor costs, that will save material costs, that will save CapEx, it will save patent filing, it will save on time, yeah. The more accurate and the more practical these solutions are, these companies are starting to take a serious look at, okay, now this actually translates to dollars saved. Right. Now, you have multiple tiers in terms of your pricing model for Molecular Universe. Maybe you can give us a little bit of color on how that looks or an example of how that plays out. Yeah, so, we have five tiers. We have tier one, which is the entry tier, that's called research. It's free to anyone with a university email. There are some restrictions in terms of size of database, number of these Deep Space searches, queries you can make per month. Then there's explorer and a team that's higher. You have access to more Deep Space queries per month. Explorer is $150 per user per month, and the team is like a discount team rate for explorer. These three are the entry levels for individual users. Enterprise and joint developments are more for companies. For now, enterprise, we found mostly battery companies and electric companies. They want to use this tool, and how they use the tool and what they do with it, that's their business. If they discover new molecules, they patent that. Joint development is more for car companies, where they are less interested in using this as a tool. They are more interested in us using the tool ourselves, and it just provides them with a solution. That's more on the car companies. We're seeing, I will say, the fastest growth in enterprise interest, battery companies, electric companies, and also cathode companies. We're actually quite surprised that cathode companies are interested in this because it was originally primarily for electrics, but a lot of cathode companies are very interested in this. We're seeing most of the interest growth coming from enterprise users, battery companies. Yeah, interesting, interesting. To kind of take a step back maybe, how is Molecular Universe, from your perspective, going to change the business trajectory or growth outlook for, you know, SES? Yeah, so, before we had Molecular Universe, people were confused about exactly what we do. We had all these segments like batteries for drones, batteries for EV, batteries for energy storage. How are these connected? You are a small company, how do you compete with the big guys, right? Now with Molecular Universe, it serves two things. One is it connects all these businesses. The batteries for drones, the batteries for EV, the batteries for ESS, all the electrolytes are developed using Molecular Universe. All the cell designs are developed using Molecular Universe, are validated through Molecular Universe, all the cell design platforms. Molecular Universe is the core that connects all these market segments. We don't need to compete with the big battery players. We are supplying Molecular Universe to them. We use Molecular Universe and then come up with the electrolyte designs, and then we sell the electrolytes to the big battery producers. They will counter-manufacture the batteries for us, and then we sell the batteries to the drones, the EV, and the energy storage. With Molecular Universe, all these pieces come together. We are seeing a much faster growth in our revenue and also better margin. That's why earlier I said, I really think with this portfolio of the various market segments and this core foundation of Molecular Universe, we can achieve break even sooner than we expect, much sooner than we expect. Right, so is this also going to help with the predictability of your business model in terms of subscriptions and things? Yes, yes. Maybe you could just give us one more step back and say, what does the rest of 2025 look and what is maybe some highlights that we might see in 2026? I think, in terms of revenue, we are on track to the guidance we gave, and that's definitely quite exciting. We do want to have all these pieces, all these market segments, finalized. We also plan to add additional manufacturing capacities and additional marketing and sales channels to really grow these segments for drones, energy storage, and electrolytes for the Molecular Universe. I think we can expect more announcements in terms of how we actually add these capacities, these marketing and sales channels to grow the revenue. Great. Unfortunately, Qichao, I think we're going to have to leave it there. I really appreciate you joining us today for today's fireside chat. Everyone, to learn more about SES AI, you can go to their website or can go to research on the company on our website at www.watertowerresearch.com. I want to thank everyone for joining us today. The views expressed in this fireside chat may not necessarily reflect the views of Water Tower Research LLC and are provided for informational purposes only. This fireside chat may not be distributed or reproduced without written consent of Water Tower Research and should not be considered research nor a recommendation. WTR is an investor relations firm, not a licensed broker, broker-dealer, market-maker, investment-maker, underwriter, or investment advisor. Additional disclaimers can be found at watertowerresearch.com. Thank you, guys.