Developing Battery Materials with AI

Vivas Kumar is the cofounder and CEO of Mitra Chem, the first lithium-ion battery materials product company focused on shortening the lab-to-production timeline by over 90%. They recently raised more than $80M in funding from Social Capital, General Motors, Alpha Wave Ventures, and others.

Vivas's favorite book: Long Walk to Freedom: The Autobiography of Nelson Mandela (Author: Nelson Mandela)

(00:01) Introduction
(02:25) How Lithium-Ion Batteries Work
(04:56) Evolution of Battery Materials
(06:46) Development of Lithium-Ion Battery Technology
(08:10) Optimizing Battery Properties: Cost, Energy Density, and Cycle Life
(10:52) Supply Chain of Battery Materials
(13:16) From Lab to Production: Key Bottlenecks
(15:26) Using AI to Accelerate Synthesis Design
(18:18) Battery Cell Qualification: Process and Importance
(20:13) Managing Risk in AI-Driven Battery Experiments
(21:20) Data Generation for AI-Driven Battery Development
(22:52) Challenges in Manufacturing Large-Scale Batteries
(25:02) Current Limitations in Battery Materials Development
(26:39) Next-Generation Battery Materials in the Pipeline
(27:50) Advancements in Battery Technology Beyond Materials
(29:52) Advice for New Founders in the Battery Industry
(31:42) Rapid Fire Round

--------
Where to find Prateek Joshi:

Newsletter: https://prateekjoshi.substack.com 
Website: https://prateekj.com 
LinkedIn: https://www.linkedin.com/in/prateek-joshi-91047b19 
Twitter: https://twitter.com/prateekvjoshi 

Transcript

Prateek Joshi (00:01.448)
Vivas, thank you so much for joining me today.

Vivas (00:04.131)
Thank you so much Prateek for inviting me and having me here.

Prateek Joshi (00:08.412)
Let's start with the basics. Can you explain how batteries are manufactured today? To the average person, obviously we use batteries every day, but we just, I think it's not common knowledge to know how batteries get manufactured. So can you talk about that?

Vivas (00:26.318)
Sure. Let's start with the basics. A battery is an energy storage platform technology.

In energy, you've got generation, which is you produce energy. You've got transmission, which is you move energy. And then you've got storage, which as the name implies, you store the energy for future use. And you have various different use cases. The world runs on a variety of different energy sources. So the most notable that we can probably think of is electricity.

I'm sitting in a room right now and there's a light turned on and that's powered by electricity. If you look outside your window, we don't live in a fully electric vehicle world yet. It's mostly still internal combustion engines, although they are clearly losing share as compared to where EVs are going. But the energy source is hydrocarbons for internal combustion engine. So a battery is an energy storage device in as much as it

And this is a relatively sort of crude way of saying it, but it accepts electrons when it's recharging and then it emits electrons when it needs to power the source or be the source of power for the device that's using it. Okay. There is a difference between power and energy. I realized that I use those words interchangeably, but any of you energy nerds out there will jump out of your seats and try to correct me for it. And power.

is a rate versus energy is about capacity. Okay, so let's establish that upfront. So how does a lithium ion battery work? A lithium ion battery works by having a cathode and an anode. One is a plus, one is a minus.

Vivas (02:25.639)
In the cathode there are lithium atoms and when those atoms get ionized the flow of electrons out of the cathode and through the leads and out of the battery is the electricity that powers the device.

the lithium ions, the Li plus that has just lost its electron, will then migrate from the cathode through the electrolyte, through the separator, through the anode, causing a charge differential.

In order to recharge the battery, electrons enter the battery, pair up with the lithium ions to make them lithium atoms again. So once again, this is very, very crude. This is like, you know, using like very simple terminology to describe an otherwise very complicated process, but we can roll with that as the assumption. In history, there have been in history, there have been many different types of batteries. So one can argue.

Prateek Joshi (03:29.514)
That's great.

Vivas (03:37.595)
that wood is an energy storage platform technology. Because if you set fire to wood, then what happens? It continuously allows the fire to burn, emitting heat and energy, right? You can make the argument that wood is a type of fuel and actually not a battery because it's a consumable. And these debates have actually gone on in the energy circles for a very long time.

What I'm focused on doing is I'm focused on building lithium ion batteries materials. So building the cathode active materials. That's what my company does.

Prateek Joshi (04:20.128)
Amazing. Love that intro to batteries. going a level deeper, there are obviously the materials used to manufacture these batteries over time. And we started off somewhere and now in the modern world, we use good materials. So can you maybe talk about the materials that were used in the past and what materials are being used today to manufacture these batteries? And also maybe what

properties makes the material suitable for use in batteries.

Vivas (04:56.732)
Historically, if you think about all the way in the past, like when the Neanderthals were around.

Fire was the best form of energy use for heat and for cooking.

So anything that's flammable, like I mentioned wood earlier, right? Like anything that's flammable was the energy source for fire. And energy generation came through lighting a fire. Energy transmission came by just carrying the fire from one point to another.

Energy storage, was more of embedded energy storage, but there wasn't really a way in which to store energy in the way that we think about it today. You always had to have continuous power, continuous fuel sources. And then energy uses were relatively primitive. At some point, it was figured out that the ideal solution would be if you could generate energy when you didn't need it, when it was cheaper.

and then store it for use when it would be otherwise inconvenient or expensive to produce energy.

Vivas (06:15.45)
Good so far? Okay. So lithium ion batteries. So the concept of a battery, I mean, there were, you know, lithium nickel batteries being used in the first ever electric vehicles back in the 1880s. And then there were nickel metal hydride batteries used for several decades. The modern lithium ion battery came into fashion because of the work of three gentlemen, Akira Yoshino, John Goodenough,

Prateek Joshi (06:16.841)
Yep. Yep.

Vivas (06:46.085)
and Stan Winningham, all three of whom won the Nobel Prize for inventing the lithium-ion battery for work that they started doing in the 1970s. In 1991 was the first practical application of the lithium-ion battery inside of a Sony Walkman. And ever since then, you've had more more consumer electronics, more and more electric vehicles and energy storage devices that are now using lithium-ion batteries. So here we are today, there's, you know, terawatt hours

of capacity being planned for batteries right now. There's millions of electric vehicles that are powered by lithium ion batteries as well. So it's been a grand slam home run that the lithium ion batteries become one of the most reliable platform technologies for energy storage.

Prateek Joshi (07:35.904)
Amazing. And when you look at different materials being used, and obviously materials, the scientists, they keep researching new, better materials. So how do they approach the development of new materials, new chemistry? Like what do they look for and what properties make it desirable? For example, why lithium ion, right? So what made it so good? And also more than that, as we look forward,

what properties are desirable here.

Vivas (08:10.093)
Earlier I mentioned this word platform technology to describe lithium ion batteries. So any platform technology is defined by two things. Number one, what other product sets are built on top of it. So for lithium ion battery, the product sets that are built on top of it is various different forms of electric vehicles and energy storage devices and consumer electronics.

The second question is how do you keep continuously optimizing a certain parameter that allows for larger and greater proliferation of this technology to power other product sets that haven't been invented yet?

So what you do is you pick a certain metric and you optimize around that. the three that matter, there's multiple different metrics that matter for batteries, but the three that have mattered the most recently in commercial conversations are cost, energy density, and cycle life. So cost. Lower cost is better.

If you look at the greatest societies on earth, they've always been built on having a cheap energy advantage. If you look at who wins wars, people who have highly developed energy supply chains and the lowest energy costs are more apt to win wars.

Vivas (09:32.639)
Energy density. If you have higher energy density, you get more bang for buck. You get to store more energy within the same space or the same volume.

That's how I think about energy density.

Cycle life is a proxy for how long your battery lives. If you have a battery that lives longer and better, then you're in a more privileged position because then there's more reliability that that battery will last through all of the use cases that you put it through.

So any scientist who's working on lithium ion batteries is working on optimizing one of these three metrics, if not a combination or all of them.

Prateek Joshi (10:22.024)
And when you look at the critical elements in the supply chain of our battery materials, can you talk about what that looks like? Meaning, obviously it's great that if a scientist comes up with something that's great, but when you have to manufacture in big numbers, large quantities, you have to source from around the world. So can you talk about the supply chain of these materials?

Vivas (10:52.733)
Lithium ion batteries are aptly named because they use lithium. So let's start with lithium. Lithium is the lightest metal. It is widely abundant everywhere around the world, but that does not necessarily mean that you can get lithium from anywhere. There are very specific parts of the world where lithium is available in enough high concentrations and an economically accessible manner with governments that have

policy to support lithium project and infrastructure development that have emerged victorious. So the most notable lithium mining jurisdictions are what's known as the lithium triangle in South America.

And also Australia, hard rock. Canada is emerging as a jurisdiction, the United States is emerging as a jurisdiction. There's some really interesting work happening in Africa right now as well. And the reason for that is there's just so much more demand for lithium now that there's always new resources and new projects being discovered.

Now, let's go back to the cathode for a second. There's two major types of cathode, nickel rich and iron rich, and there's associated supply chains for them as well. So nickel rich generally use various high purity nickel sulfates, for example, and cobalt sulfates or cobalt hydroxides. So there's a mined material, the nickel or the cobalt, and then there's a specialty chemical, which is the conversion of that mined material.

to a usable specialty chemical that can be made into the cathode-active material. So there's 150 plus parts that are either mechanical or chemical in nature to be used inside of the lithium ion battery.

Vivas (12:45.434)
and each of them has its own supply chain with its own complicated geopolitical dynamics to follow.

Prateek Joshi (12:51.902)
Right. And let's move the conversation a bit towards what it takes to go from lab to production. So just as a starting point, going from lab to production takes a long time. What are the key bottlenecks here just kind of in that process?

Vivas (13:16.284)
So the first is that the synthesis process, so the process of actually making materials is very different from a lab to large scale. Why? Because in a lab you have experimental flexibility. You have more permission to take different approaches to synthesizing the material with a relatively minimal cost versus at large scale you have to deploy hundreds of millions of dollars in buying capital equipment.

where the flexibility and the optionality of how you synthesize something is minimized.

So the big mistake that is often made in the thinking on the subject is just because you can make battery material really well in a lab doesn't mean that it will translate to large scale based on the available large scale production equipment. And also second of all, the supply chains that feed the materials also have to be large enough. So you can make a really whiz bang new chemistry in a lab using very esoteric materials. But if you have to build a supply chain

all the way from scratch, along with also building your own production infrastructure, more than doubles the complexity.

Vivas (14:35.864)
Another big consideration is if you build a factory, you have to deploy lots of capital upfront and then build infrastructure to make products that have to get qualified for a customer before you can even make cash flow. So typically it ends up not only being a technical manufacturing challenge, but also a financing challenge that needs to get solved.

Prateek Joshi (15:03.038)
Right. And when it comes to exploring synthesis design spaces, you mentioned how ML-driven techniques can make it faster, better, cheaper. So how do you utilize, how do you and your team utilize AI to explore synthesis design spaces?

Vivas (15:26.842)
This is one of the founding theses of Mitraken, is that we saw that there was an opportunity to radically reduce the design-build test cycle to bring a battery materials product from lab to market about 10x faster than what the conventional industry allows for. The way that we do that is we take every single step of the battery materials R &D journey, from synthesis to testing the chemical and physical characteristics

building battery cells and to testing the battery cells. And we found various optimization opportunities to radically reduce the time by increasing our ability to forecast failures so that only the experiments with the highest value and the highest probability of succeeding propagate all the way through to the most expensive steps of the process, which is cell testing. Generally speaking, you have to

cycle a battery cell for 10 months before you can determine whether or not the materials inside that battery cell are useful to the application that you're designing for. What we're able to do, just one example, is we can figure out within just a few days of data, rather than waiting 10 whole months, whether or not this battery cell is likely to succeed. If it's not likely to succeed, then we just de-prioritize the experiment

and prioritize a different experiment, which could have a higher likelihood of succeeding. So that's just one example. There's plenty of others that allow us to rapidly down select experiments so that we can come up with key design rules and synthesis steps and material specifications that are relevant to a customer.

Now, secondly, to reconcile this with what I said earlier in this podcast about how just because you make something in a lab well doesn't mean that it translates. What we're trying to figure out is that scale up translatability of design rules from the lab to large scale infrastructure. So how can we make at large scale?

Vivas (17:41.62)
the same or better material than what we've made in the lab.

Once we do that, the next question is going to be, how can we faster get to the type of yield rates that we want to ensure long-term sustained profitability even at large scale?

Prateek Joshi (18:00.544)
And while we're on the topic, can you briefly explain cell qualification and also with versus without AI? How do you do it and why is that process important?

Vivas (18:18.71)
Cell qualification means assessing the performance of a battery material within a battery cell. So you can make, once again, the best whizbang chemistry that passes all the chemical and physical characteristics and metrics that you set. But if it doesn't perform within the system for which it's designed, then there's no point in even continuing further in developing that material or producing it at large scale.

What we do, even though we are a battery materials company, is we place a lot of emphasis on battery cell building, in knowing how to make cells effectively and then making them on an automated cell assembly line that we have in our lab in Mountain View, California. And in having thousands of channels of battery cyclers to cycle those battery cells that we make. Because when we go to a cell customer, when we hand them a material to test,

We want to be able to say to them that we have tested that material in representative cells ourselves, and we know that it will pass the test that they give to us.

Prateek Joshi (19:27.294)
And when it comes to using AI, for example, I just want to take a quick example. You can use ChatGPT to generate text for marketing. And if it doesn't work, you go back, you generate another text. It's fine. You're not missing too much. The cost or penalty is pretty minimal or almost zero versus the way in which you use AI if you don't.

estimate the right outcome, then the price is bigger, meaning the penalty is high for getting it wrong. So what checks and balances do you have to put in place to make sure that you're maximizing the likelihood that, 10 months from now, we're not discovering something that will force us to start and go back to day one?

Vivas (20:13.76)
This goes back to the point that I was making about predicting experimental failure faster and assigning a probability of success much earlier in the life cycle of the experiment so that we can down select only those experiments which will succeed. There will certainly be some that get all the way to 10 months and make us realize that we did not hit the metrics that we want. But we want that to be the simple minority rather than the vast majority, which is unfortunately the reality most

battery labs in the world today is lots of experiments run to failure before it's realized that they're not going to be suitable for the application that they're designed for.

Prateek Joshi (20:55.072)
And to build these systems internally and to deploy them in a reliable way, you need good, high-quality data. So can you talk about all the ways in which you are generating data or gathering data to use here? Is it a combo of internal and external? Are you buying? Are you getting systems? So how do you get your data?

Vivas (21:20.726)
Where we get our data? So we generate data from our own in-house experimentation at this point, because we have, like I said, we've cycled several thousand battery cells at this point. There's a few open source sources of battery data that we use as well, but our in-house generation has involved not only, data generation has involved not only our own battery cells that we've made.

It also has involved us benchmarking against the current industry standard cells that are readily and widely available for electric vehicle applications. So we've just gotten our hands on the cells and we've bought as many as we can from as many sources as possible and done a teardown to understand what are the materials that are used inside and what are the specifications of those materials, but then also run cyclers to generate a data set internally that we can compare our own battery cell cycling against.

Prateek Joshi (22:22.081)
Going into the process of manufacturing itself, obviously, let's say you have your materials, you've tested, it's looking good. Can you talk about all the things that come up during a real true production run? You're servicing a large order, it's out of the lab, and you have to deliver now to a big customer. What are all the things that come up during manufacturing?

Vivas (22:52.586)
One of the most difficult parts of manufacturing is you have to buy several different pieces of complex equipment.

tighten them all to work together to hit a certain yield target to ensure your line's profitability. Yield is just a question of how much can you make, how much do you make versus how much can you make. There is a yield threshold at which profitability can happen for any manufacturing line at the price point. And if your price point, this is a simplification.

But how do you calculate the total cost of making one of these products? The cost of the raw materials plus the cost of manufacturing, so the cost of running the equipment plus the cost of financing associated with the equipment. So how do you minimize the cost per kilogram of running equipment and the financing cost per kilogram? Is number one, try to make as many kilograms as possible.

and making sure that the number of kilograms you make is as close as possible to the number the number of kilograms you do make is as close as possible to the number of kilograms you can make from something.

Yield optimization is the differentiator between somebody who wins and loses in manufacturing to get to their cost target. Yield and also selecting the process flow that can consistently make your product at lower cost over time. If you can master the art of these two things, you can win in manufacturing products over time.

Prateek Joshi (24:38.688)
Amazing. Now, looking forward to the future of battery materials, if you look at everything that's being done inside and outside your company, what are the current limitations in battery material development? Meaning, what needs to be unlocked so that we can see the next big wave of innovation?

Vivas (25:02.741)
We are already living in the beginning of the renaissance of battery materials because there are so many more applications that are using lithium ion batteries than ever before. There's so much interest from consumers, from companies, governments on incentives and policy towards more electrification and automotive and more proliferation of clean energy generation technologies that need to be coupled with energy storage like batteries.

that this is the most exciting time ever to be in our industry. What we need to see though is the excitement from the supply, from the demand side needs to be matched with innovation from the supply side. Okay, so let me repeat that. The excitement from the demand side needs to be matched with innovation from the supply side. So radically reducing the time to bring a new material from lab to market.

is one of the ways in which we're contributing towards that challenge. And that's two parts. Number one, like improving experimental speed, and the second is scaling up faster.

Prateek Joshi (26:16.948)
Now, obviously, the research on new materials is an ongoing thing. What new materials right now that's still in the lab phase but are showing potential in the next-gen batteries? Or maybe like what is something that in about three to four years, we'll see that, that started now and now it's just scaling up. So what's exciting to you?

Vivas (26:39.457)
We believe very strongly in a thesis of iron-rich cathodes. The first product that we're making is called LFP, lithium iron phosphate. This is a well-known, well-industrialized chemistry used at scale for almost a couple of decades at this point. Within mass market EV, within grid-scale energy storage, it's increasingly becoming the product of choice in the automotive markets in the United States as well. The next product that we're working on is called LMFP, lithium manganese iron phosphate.

which is a higher energy density product which is applicable to more premium and higher end segments of the market that we're choosing to enter. Then our final product is something that's still in the lab. It's called LMX. And this is where we can get to pack level energy density competitiveness with a nickel rich cathode, which was the previous standard but is rapidly losing share within the cathode market.

Prateek Joshi (27:36.07)
And going beyond battery materials and looking at the battery technology as a whole, what advancements in battery technology are you most excited about today?

Vivas (27:50.803)
What I'm most excited when it comes to batteries is that there are multiple different companies that are working on optimizing various different parts of a battery cell. There's cathode companies, there's separator companies, electrolyte companies, anode companies, there's even cell design companies. And the integrated value of all of these companies innovating on various different parts of the system coming together to make a super battery cell is what's exciting.

If we think about what batteries have done in the last 15 years, we've talked about, we have had a 10X decrease in costs, which is what has allowed batteries to proliferate as the platform energy storage technology of choice within electrification.

There's not going to be a 10x decrease. There's a physics-based limitation to a 10x decrease in cost happening again for lithium ion batteries. But the continuous improvement and the integrated value over time of all of these improvements that result in an exponential outcome of the technology proliferating even faster. I've already been in this industry for the last nine years and I've seen this daily in my work. And I think that we're still at the beginning for how far

batteries can be pushed for various different applications and for new product categories that we have not even thought of.

Prateek Joshi (29:19.068)
And I have one final question before we go to the rapid fire round. And this is around, how do you start something new in batteries? It's capital intensive. Obviously now, now we have raised a lot of capital, we're on a different trajectory, but going back to the very early days, how do you advise a young founder who's just starting out, wants to work on batteries, has something, but what do you need to show in those early days to prove the viability?

of what you're working on.

Vivas (29:52.493)
The first suggestion would be that there are multiple different points of innovation that need to happen. Figure out what you uniquely capable of doing and where you have specific knowledge. And choose early whether you're targeting a very large market that takes a long time to sell into or a specific niche market that requires lots of upfront development.

I chose to go after Ironrich Cathode in our company because that's where the innovation was most lacking in the Western world. And that's also where the strongest tailwinds were going to be from customers who wanted to have a cheaper alternative to get more mass market EVs on the road. That doesn't necessarily mean that, you know, Nickel Rich is bad or innovating on Anode is bad, for example. And I hope that there are other entrepreneurs who go out and do that. The second is recognize

R &D and manufacturing are two very different things. And the skill sets and competencies to do R &D effectively cannot just be assumed to be applicable directly to manufacturing. The United States has lost a lot of this manufacturing talent. And that's because of 30 years of industrial policy from Asian nations who have been very, very good at building large scale energy storage technology industries where we lag behind. So bringing back these skills to the US,

It's going to take probably another 20 more years. And we need to start now. We need start now. We need have more companies that are working on building in the battery supply chain. There's never been a stronger amalgamation of legislation and policy and funding from the government to make it happen. So I once again still feel like it is still just the beginning of a long story that needs to play out for the battery supply chain.

Prateek Joshi (31:42.72)
Amazing. I love that. I love that. I love that guideline around how to approach this industry. All right. With that, we're at the rapid fire round. I'll ask a series of questions and would love to hear your answers in 15 seconds or less. You ready? All right. Question number one. Question number one. What's your favorite book?

Vivas (31:59.021)
Let's do it.

Vivas (32:03.959)
Nelson Mandela's biography, Long Walk to Freedom.

Prateek Joshi (32:09.352)
Next question, what has been an important but overlooked AI trend in the last 12 months?

Vivas (32:17.45)
This is really good question.

Jensen Huang's sovereign AI speech at the World Government Summit in Dubai.

Prateek Joshi (32:28.73)
That's amazing. Yes, I think that's so good. That's a separate topic on its own. All right, next question. What's the one thing about batteries that most people don't get?

Vivas (32:44.419)
Battery fires are a lot less common than you think. And there have been tremendous advances in safety towards avoiding them.

Prateek Joshi (32:54.81)
What separates a great AI product from a merely good one?

Vivas (33:02.347)
invisibility.

Prateek Joshi (33:04.158)
I love that. That's great. And I agree. Next question. What have you changed your mind on recently?

Vivas (33:16.557)
Wow, so many things, but we're in the middle of an election season right now. And I have come to appreciate a lot more the impact of local government relations and local government elections. I had previously, I think, overhinged on how much policymaking from the national level bleeds down to state and local in the United States.

but I've learned about the power of federalism. This is very relevant because of the work that we've been doing in multiple communities in the US in our own business.

Prateek Joshi (33:57.128)
What's your wildest battery prediction for the next 12 months?

Vivas (34:05.259)
you are still going to see a very significant cost decrease, even though analysts are predicting that because of supply chain trends, battery costs may go up for the first time in nearly 20 years this year.

Prateek Joshi (34:22.144)
All right, final question. I think we covered a little bit earlier, but what's your number on advice to founders who are doing batteries or not doing batteries, but founders who are starting out today?

Vivas (34:37.549)
You need to do three things. You need to sleep eight hours a night. You need to exercise every single day. And you need to eat a whole foods, highly nutritious, protein rich diet. I don't care what business you're running and how many people you manage. If you can't be the healthiest version and highest performance version of yourself, you will not be able to create the value that you want to create for the world.

Prateek Joshi (35:02.528)
That is brilliant. 150 plus episodes, not a single time has somebody said that. And that is so very, very important because when you do this for like years and years and years, I think mentally and physically, you've got to keep yourself at the highest level. So that's brilliant. So Vivas, thank you.

Vivas (35:18.576)
You want to know how you win in this space. You don't necessarily have the best technology. You just outlast the competition. That's true for companies and that's true for individual entrepreneurs as well.

Prateek Joshi (35:23.508)
God.

Prateek Joshi (35:28.448)
All right, right.

Prateek Joshi (35:33.056)
That's amazing. This has been a brilliant, brilliant discussion. I just loved the incredible crispness and depth of your knowledge and batteries and also using AI to make that faster, better, cheaper for yourself and for the world. thank you so much for coming onto the show and sharing your insights.

Vivas (35:49.082)
And thank you for the service that you're doing to the world and spreading this knowledge Prithik. Hope to see you again soon.