So hi, everyone. Thanks for joining us for this last session, but certainly not the least, with SES AI as part of Deutsche Bank's Global Auto Conference. My name is Winnie Dong, and I'm a senior associate at DB's U.S. Autos and Auto Parts Research Team. SES is looking to develop its next-gen lithium metal battery technology by combining the high energy density of lithium metal with the cost-effective, large-scale manufacturing of conventional lithium-ion batteries. The company is currently in JDA partnerships with several OEMs, including GM, Honda, and Hyundai, and has announced B-sample agreements with two out of the three, which are Honda and Hyundai. I'm very pleased to be joined by Dr. Qichao Hu, Founder and CEO of SES AI, for a side-by-side chat and Q&A. So Qichao, thanks for being with us. I will hand the floor over to you for some opening remarks and some slides that you want to discuss.
Sure. Thanks, Winnie. And then I'm Qichao. I'm the Founder and CEO of SES AI. And then we work on lithium metal batteries, and they're mainly for EV and UAM applications. I'll go with some slides. And then there are three parts to our story, and that includes EV, UAM, and AI. And then in terms of vertical integration, we are very integrated in terms of intelligence. So we start with molecular discovery, and then that is a key part to our material. And we also work on molecular discovery for the electrolyte to electrolyte formulation and into battery development. And that includes all the A-samples that we have with GM, Hyundai, Honda, and then now the B-samples that we have for both EV and UAM applications. And then we also cover Avatar, which is this AI tool that we have to predict battery safety.
In terms of global presence, currently we are headquartered in Boston. That's where we do all the materials, all the AI development, and all the A-sample lines and then B-sample lines. We have split between Shanghai and Korea. As we mentioned earlier, we announced three A-sample JDAs, and now we have two B-sample JDAs. And then we announced three A-sample JDAs with GM, Hyundai, Honda, and then also two B-sample JDAs.
Maybe before we go into some of the operational and milestone updates, I do want to sort of take a step back because this is a pretty sort of long journey in terms of getting to that commercialization timeline. So if we just shrink the timeline into maybe the past 12 months or so, what would you say is the latest update in terms of competitive landscape, and what would be beneficial for us to sort of understand to set the stage a little bit?
I think if you look at the landscape in terms of EV and UAM, and OEMs in general are becoming more serious to lithium metal because of what lithium metal can offer. And then if you look at the EV market, of course, lithium metal provides longer range. And then lithium metal, especially when it's paired with LFP cathodes, it can actually lower the cost for the same thing. So it's quite appealing both from a range perspective as well as a cost perspective to EV OEMs. And then to UAM OEMs, then it's actually quite obvious. It's unlikely that UAM OEMs will ever reach profitability with lithium-ion batteries. And so lithium metal is almost a key enabler to UAM OEMs to even get to profitability. And also in the last 12-18 months, more UAM companies are getting certified. So it's not just a dream anymore.
It's actually becoming very real. And then also a lot of the OEMs, both EV and UAM, are waking up to the need for AI for manufacturing, AI for safety. And then especially for next-gen batteries, it's not enough to just have a new chemistry, but also use new tools to predict safety and to improve manufacturing quality. I think all these trends and then more OEMs are waking up to a new trend. And also investment in lithium metal and also AI for manufacturing, AI for safety actually increase in the next 18-12.
Yeah. So yeah, I guess you're essentially characterizing lithium metal as sort of this end game for EV. Maybe if you can just remind us, what is SES AI's approach to developing lithium? There's some of your competition. It's sitting on the audience today. How would you characterize SES AI's differentiator in that approach?
Well, so we've always wanted to develop lithium metal batteries using very standard lithium-ion manufacturing processes. And that to us is important and scalable. So some of the key differentiators that we have, well, one is on the material side. We go really deep in terms of electrolyte development. So not just the formulation, but the molecular development, identifying new molecules. And right now we have one of the highest quality efficiencies for any electrolyte on lithium metal. And that's not because we use an off-the-shelf molecule, solid molecule. It's because we actually develop a new molecule. And then we're going to develop even more new molecules. So we go really deep in the electrolyte development. And then that electrolyte is one of the most important materials in lithium-ion batteries in terms of safety and performance. And another thing is the manufacturing.
So lithium metal manufacturing is quite similar to lithium-ion, but then there are lots of small details that we actually learn and accumulate through the A-sample and the B-sample practice. And the third, I would say, is AI for safety and AI for manufacturing. Because lithium metal is new compared to lithium-ion, and the OEMs don't have that level of comfort with a new chemistry, we really have to basically get to that level of comfort by collecting a lot of data and then using AI to accelerate that learning instead of just spending 10, 20 years.
Yeah. I did ask this question in my prior session, but I don't want to skip it for you because I do think it's really important to address, which is we are seeing a slower penetration basically in terms of EV, at least versus expectation. Maybe it's still growing, but does that change any sort of urgency from your customers specifically?
So I think the slowdown in penetration is more for U.S. EV using lithium-ion, right? But then if you look at foreign car companies, and then we have two JDAs with foreign car companies, I think they are not really slowing down, even in EV. And Hyundai recently became the second largest producer and seller of EV in the U.S., as after Tesla. And also we're not working on lithium-ion. We're working on next-gen. And I think quite contrary, if you look at next-gen batteries, and especially the use of AI for manufacturing and AI for safety, that is actually accelerating. And the more OEMs begin to realize the importance of next-gen batteries and also the use of AI to help predict safety and quality.
Let's talk about maybe your cell design that's unique to SES. Can you touch on the functionalities and capabilities without giving away too much of your secret sauce? And then from a technical perspective, where are you today versus 12-18 months?
So in terms of cell design for EV, we have this long health cell with tabs on the opposite side. And then for EV application, it's end of width is around 500 mm-600 mm wide, depending on the OEM, 50 amp hour-100 amp hours. For UAM, it's a narrower design, about 30 amp hour. And a year ago, we were still in A-sample, and we're still just finalizing that cell design for EV. We didn't have the UAM cell design. Now the EV cell design we are in B-sample is actually being designed to fit into demo cars as part of the B-sample. And now we have a UAM cell design.
So yes, we now do have two JDA partnerships that are in transition to B-sample, which you mentioned an example for this year. Maybe talk through how those two relationships look like today, because one with Honda, one with Hyundai, which you recently announced, right? And with Honda, it's been a much longer relationship. Can you maybe compare and contrast the two? And then potentially, when can we maybe expect a third B-sample agreement, which is presumably with GM as that last part?
So in terms of the two B-samples that we have, they include obviously a new line, a B-sample line that's dedicated to the B-sample. Also, very extensive just collaboration between our team, our material team, cell engineering team, and also their material and the cell engineering team. Very deep. And then in terms of third, we had a few in the pipeline, but now we're not that focused on getting a third JDA. I mean, I think the number of JDAs to us is less important than actually how deep we go with the two that we have. So for now, we are just focused on the two that we have.
Maybe talk more about the depth of the relationship or the depth of the collaboration. How does that look like? You have maybe teams in the OEM partners facility. Vice versa, how do you engage the two teams?
Yeah. Yeah. So this year, and then we announced this as part of the B-sample JDA with Hyundai, we will actually be for the first time building a line in their facility. And then this is actually quite significant because down the road, when we are in C-sample and then our commercial for EV applications, we are not going to build a line by ourselves. We are going to do this joint venture, likely a three-way joint venture. So we have to figure out how to build a line, operate a line, and then build ourselves in someone else's facility. And then this is actually a really good practice. And then one example, so and it actually helps lower the cost, right?
So instead of us paying for the whole line, paying for the whole facility driving and then utility on that stuff, the OEMs will actually pay for a lot of that. We still operate the line. And then one thing that's really important is we make sure we install Avatar. And then for the first time, we are actually installing Avatar, this AI for manufacturing software, in someone else's line, in someone else's facility. And then this actually opens up the possibility of a new source of revenue down the road. So whoever that builds lithium metal lines, B-sample, C-sample, we make sure we operate. And also we make sure we install Avatar AI for manufacturing software on that line.
Yeah. Maybe talk about the rest of the manufacturing lines that are in place, because you do have some that are in South Korea. You have some that are in China. What are the plans for those? What are your uses for now? And then maybe talk about at what rate are you actually producing cells right now? And why is it significant?
Yeah. Yeah. So the 2 B-sample lines we're still building, and then the B-sample lines will be more automated than the A-sample lines. A-sample lines are islands of automation. B-sample lines are more peninsulas of automation and basically much more automated. And then right now, for the A-sample lines, we are keeping 1 A-sample line for continuous development. Because while we wait for the B-sample line to be built, we're not waiting for that line. We're building B-samples on the A-sample line. And then another A-sample line we're actually converting to make UAM cells. So the A-sample lines, we can make about 30-50 cells a day with 1 shift. And currently, and these are the 30 amp hours, 50 amp hours, and even 100 amp hour cells. And then so that translates to about 500-1,000 cells per month with 1 shift.
So that's basically the A-sample lines. We're converting one for UAM. And then it's actually really meaningful because not only can we build the cells to supply to the UAM OEMs, we are also building a lot of cells to collect data, to train the Avatar, both AI for safety and also AI for manufacturing. It's good to have a line that's dedicated to building cells to collect data to train Avatar.
Yeah. So it seems like data is a very important part of your process. Maybe talk a bit more about how, what do you do with the data collection? You mentioned Avatar. Can you go a bit more into that? And how does it reinforce your testing, your validation? Maybe how does it help you reach your safety goals?
Yeah. So in Avatar, we have two parts. One is the manufacturing quality. One is basically performance testing. So AI for manufacturing and AI for safety. And then if you think about what impacts the safety of a battery, a big part of that is manufacturing quality. So the saying in the battery industry that quality equals safety. And actually, in lithium-ion, most of the recalls the industry had in the past were due to manufacturing defects. So part of the Avatar is AI for manufacturing. That's really important. And then, for example, so we hire a team of very good quality engineers from CATL, Panasonic, LG. But that's not enough because they have experience from lithium-ion. But on top of that, we add Avatar. We collect all the manufacturing data, and then we train this model. And what this model does, it's actually really important.
For example, it's going to tell you the 0.1 mm in the electrical misalignment has a greater impact on quality and safety than, let's say, 0.1 g of electrolyte in the filling process. So things like this that no one has the knowledge of because no one has made lithium metal cells at scale, but this model can actually train them. And this is only after just making 200 cells per month. And then in A-sample over the past 2 years, if we can make more cells, several thousand cells per month in B-sample with 2 EVs and also all the UAMs, then we'll have a lot more data to train this Avatar. So in lithium-ion, in the quality manufacturing, so it took the lithium-ion industry probably 20 years to develop all the quality experience.
We can do all that for Lithium Metal in maybe 18 months with several thousand cells per month with that amount of data. And then for the performance testing, AI for safety, then also when you put a battery, a Lithium Metal battery inside EV or UAM, and then you cycle this on the different mission profile, that also has an impact on the safety. And then so in the lab, most of the cyclings are pretty nice cyclings in the sense that you go from 0% to 100%, and then you charge, rest, discharge, rest, charge, very structured. But then in real life, no one does that.
So we also train these cells under a variety of mission profiles, both lab mission profiles as well as real-world EV UAM mission profiles that we get from the OEMs, that we can integrate the manufacturing data with the actual performance testing data on the different mission profiles. So that the final goal is to really be able to predict incidents. And then so, for example, at the end of last year, our accuracy was around 92%. Meaning all the cells that we had incidents, we could predict 92%. This year, our goal is to get to 95%. Basically, of all the cells that we will have incidents, we want to predict 95%.
Yeah. I mean, predictability of incidents, does it necessarily prevent the incident? Is that two different concepts that we should think about, or?
Yeah. So once you can predict it, say 5 cycles or 3 cycles before, then you can stop it. So predictability does lead to prevention, right? Because you are predicting it before it happens, so you can stop it before it happens.
Are you talking to your customers about unit economics right now? Is this the right stage to talk about that, understanding that there is still a lot of testing and validation that's going back and forth? You're building a line in the facility, etc. You're building Avatar, AI line to the line. When is that point where you talk about unit economics with your customers?
Yeah. So we have been talking about the structure of the cost. Not the cost, the exact number, but the structure of the cost. Meaning if we look at the COGS, basically you have the BOM, okay? Within the BOM, you have anode, electrolyte, separator. What's outside the BOM? You have assembly costs. You have all that. And then as part of A-sample, we actually started doing this, making sure that our anode cost is actually reasonable at scale. And now in B-sample, especially now that we are operating this B-sample line at an OEM's facility, then basically they can see the COGS minus the BOM. And then within the BOM, they can see the cathode because actually the OEMs actually source the cathode for us. So cathode is transparent. The COGS minus the BOM is also transparent. And then what's new is basically electrolyte costs and anode costs.
So these two, and then we demonstrate our electrolyte is made using mature industrial chemical process. So that's scalable. And then the only thing that we still have to prove is the anode costs. But also anode costs, we start with ingots. We primarily use extrusion. So cost of thin lithium foil. At the end of the day, the final cell in terms of $/kWh will be quite similar to a lithium-ion cell using the same cathode and then also using the same assembly process.
Yeah. You alluded to Avatar and cell predictability. Can you maybe talk about the potentials of it in a future state? Is it something that, in addition to internal quality control, something that you can actually monetize as sort of like an adjacent business?
Yes. I think so. Now we're focused on EV, and then UAM is a natural adjacent market to this. And the two parts of Avatar, AI for manufacturing and AI for safety. AI for safety, there are already UAM customers that have asked us, "Can you apply your Avatar not just to your lithium metal module? Can you apply that to lithium-ion module?" Along using the same mission profile, which makes sense because the Avatar for safety is agnostic to the chemistry. Basically, the data that is used to train it can be lithium-ion, can be lithium metal, and the model, of course, will be trained and then based on the data. So yes, we are already in conversation with several UAM customers about applying Avatar, AI for safety to lithium-ion modules along the same mission profile.
Then for AI for manufacturing, as part of the B-samples, we are already installing AI for manufacturing in this line owned by Hyundai, right? So down the road, and whoever is going to assemble lithium metal batteries, not only do we have to operate, we also have to license AI for manufacturing to that. But yes, the possibility to license AI for manufacturing, AI for safety, yes, we are in conversations with both EV and UAM customers beyond just supplying lithium metal cells.
Yeah. So I understand that maybe sourcing may not be top of mind right now in the current state. But just in light of the recent EV Terra development in terms of inflows of raw materials, what is your overall reaction to that? From SES's perspective, is there any sort of longer-term implications eventually when you do reach that commercialization stage when you do need to start sort of sourcing mass volumes? Is that sort of something that you think about?
Yeah. Yes. Obviously, we consider this together with the OEMs. And then the few OEMs that we work with have a global presence. And then, for example, today we source lithium ingots up until lithium ingots, and then we do everything else in-house. And then we do that. Basically, we have a mirror image of the whole process starting from lithium ingots in both Shanghai and Korea. So the goal is Shanghai will serve the China market. Korea will serve the basically rest of the world market. And then also you're seeing that in terms of sourcing, a lot of companies in Korea are doing that, serving as a platform for the rest of the world market. Yeah. And then in terms of intelligence, so several EV OEMs are also talking about insourcing intelligence, better intelligence.
Because in the past, they had battery companies manufacture the batteries, but they want to not only insource manufacturing, but also insourcing intelligence. So we are actually more vertically integrated in terms of intelligence than we are in terms of materials. In terms of intelligence, we start from molecules all the way to the final vehicle testing. We don't outsource anything.
Yeah. Can you talk about the relationship with GM a bit? Currently, still sort of in that A-sample stage. If we sort of rely on some of the transcripts, I've often heard Mary Barra talking about their exploration of alternative chemistries, etc. So what is it going to take, I guess, to push them to that B-sample stage? Or is it something that, from their conversation with you, not necessarily the immediate near-term focus?
I think GM has their plans, and then we're still working with them. For us, we're focused on the B-samples that we have. The specs, the performance are actually quite similar. Once we hit the B-samples, not only GM, but other European OEMs, the target specs are basically the same. We are focused on the B-samples that we have and also the UAMs.
Okay. Maybe talk about your vertical integration strategy. I understand at some point you have moved the production of Lithium Metal anode in-house because you want to maybe manage quality control, etc. But overall, how much are you looking to do in-house versus upstream, using upstream funding?
Yeah. So I think our goal is to ensure quality and then making sure we have complete data to train Avatar. So if it's a material or a process that we can outsource and still get the same quality, same data, okay, fine. But then if it's a material process that we have to insource to get the quality and the data, then we just insource. And then lithium foil, especially really thin and wide with lithium foil, that was one thing that we used to buy from other companies. But then the quality wasn't good. We couldn't figure out what was wrong with it. And we also didn't have complete data to improve that. We decided to insource that.
What are the, I guess, capital that you get to sort of do that? Because presumably, in order to ensure that you would have to invest some money upfront to do that, what was that sort of business process between the value proposition to this?
So for A-sample and the B-sample scale, lithium metal annual foil production, it's roughly less than 5 million. So it's not that much. And then down the road, if we're in C-sample or SOP, if we need to scale that up, well, first of all, in the SOP, the CapEx is going to be shared between us and the OEMs. But we always want to make sure we can control quality and get full access to data.
Yeah. So you've talked about UAM many, many times now. Maybe help us sort of frame the TAM a little bit in that. For those of us who might not be as familiar with it as you are with auto applications, right? How might the economics be different for UAM? And then is the process easier, harder in terms of timeline to reaching that commercialization time target?
Yeah. So a couple of benefits of UAM compared to EV. And we're still focused on EV, but then UAM is a more near-term opportunity for us because, for one, if we look at the current status, EV industry cost of battery is less than $100 per kWh. And for LFP, it's like $60 or even less. But for UAM, just current market, it's about $400 per kWh, even lithium-ion. That's the current price. And then also volume isn't as high. For EV, you're not going to get any contract from these large OEMs if you don't show like 10 kWh capacity. But then for UAM, a lot of the leading UAM customers will not want to do aircraft per month. And then our current A-sample lines, 1,000 cells, that's actually 2 aircraft worth of batteries per month.
So we can supply modules to them at a higher price now without building a new line. And that's one. And then number 2, if you look at UAM, most UAMs operate at a fleet business model. So like Uber, basically $ per passenger per mile. And then so Uber, United, Delta will want to operate in the U.S. and also other operators in other countries. And if you can make the battery lighter, then the $ per passenger per mile actually decreases because now you can add more passengers. You can add more suitcases. And the aircraft will fly much farther. So this benefit is actually really important to UAM companies. And the timeline to commercialization, UAMs don't really have an A-sample, B-sample, C-sample. So from a technology perspective, once you get to B-sample stage with an EV, then you are sort of similar to commercial for UAM.
Now, of course, you have to get through all the supplemental type certification with FAA, with EASA, with all the different agencies. But that's easier and faster than the cells going through A-sample, B-sample.
Yeah. You were previously talking about the three-way joint venture sort of business model. Is that something that's sort of been that you're thinking about that's been set in stone? Because I do recall that in our past conversations, it was in a wholly owned facility or licensing or the joint venture setup. It seems like you are sort of is that like a set position that's been sort of contemplated, or is that still sort of up in the air? And you're still exploring those?
I mean, both open. We are leaning towards this business model strongly just because we do want to participate in the manufacturing of lithium metal. And we talked about for EV applications. So we do want to participate in the manufacturing. But then that could mean the OEMs put up the CapEx for the line and the facility, but then our people go into operate. And then we also license this Avatar AI for manufacturing to that line. So we collect all the data. And then we're still responsible for operating the line and also the quality.
Yeah. Who are the most probable sort of partners among your UAM partners?
I think if we look at the two that we have, B-sample is where they're at.
We have just under 2 minutes. Anyone have any questions from the audience?
Yeah. Thank you for sharing. Just a really quick one. I'm just curious, do you actually plan to expand your global supply chain across ASEAN, South America, for example, across within this battery energy-driven sector? We've been seeing many companies that are looking for source rich originations and mitigating their supply chain. I don't know, is that necessarily your strategy as well, or do you see yourself in the maybe future three to five years to strategize your global supply chain? Yeah. Thank you.
So the question is, do we plan to source material from South America?
How do you see yourself in regard to the global supply chain game? Do you see, because I saw the map of your locations, so South Korea, China manufacturing, do you actually force yourself into source external resources by mitigating your factory to maybe, I don't know, Southeast Asia or Latin America because those are generally labor-cost-effective regions and resource-rich?
Yeah. We currently don't buy straight from the resource, the mines. We buy, for example, high-purity, better-grade lithium ingot. There are companies in Asia and North America that sell this. So we still focus primarily on cell sampling.
Anyone else?
Thanks for the presentation. Just a quick question around cash position, if you just talk about that at a high level. And then secondly, maybe any other product milestones that you foresee as being an imperative maybe in the next six months and heading into 2025?
So I think in terms of cash, in Q1 this year, we still have about $300 million left. Our guidance for this year is 90-100. I mean, we're good. We talk about achieving revenue from UAM first half next year, 2025, and then also from EV second half of next year. We're good through C-sample and then SOP. In terms of product milestones, for EV, we are absolutely focused on B-samples. Making sure the lines are operational this year, and then we can start collecting data from the Avatar that we install on these lines. For UAM, we do want to get to a supply agreement with an actual UAM business.