AEye, Inc. Class A Common Stock J.P. Morgan Automotive Conference

Thanks, Matt. Okay, great. Oh, yeah.

Good to be here. Fourth year for me, and no hurricane coming or anything like that, so nice relaxing time. We have a few slides to go through in the presentation. We have a little bit of a product introduction that we will share here, and then save most of the time for Q&A. Yeah, the question since I have been here last year, key question is LiDAR still essential? Is it still central to autonomy? I love this recent quote from Rivian. It is not the only one. But still fits nicely between that spot. It is complementary, and especially in automotive autonomy between radar and camera. It sees very far, like radar can, but it sees a lot sharper and a lot resolution. When you see the comments here about the L4, so we had Don here on the stage earlier.

I think that is an important factor of how the direction for the tech stack gets influenced, moving forward. Quick picture, this is our car that we work on jointly with the University of Toronto. It is their aUToronto driving program. LiDAR is still the only sensor that can see into direct sunlight. Rain is practically invisible to our sensors for driving in bad weather. LiDAR is still really a key piece to the puzzle in the tech stack. One of the things we learned since we were here last year about the OEM needs, especially on the automotive side. I would say when I was standing here on stage last year, there was this notion of having a LiDAR in a passenger vehicle. I think that has changed quite a bit. It is called a monolithic of LiDAR. It is one size fits all that takes on many tasks inside the vehicle.

I think what we are seeing now is a much more specialized set of use cases, which play well to the architecture that we have. We will get into that in a second here. Supply chain resiliency. We are an American company. We manufacture in America, and this has become a front and center topic when talking with purchasing departments in OEMs, regardless of whether it is passenger vehicle, robotaxi, and trucking. In Western Europe and the U.S., it is table stakes at this point to have a very resilient supply chain. Well, good thing Don is not here, but no intention to support any particular trucking manufacturer. But one of the things we saw in the first half of the year is an enormous amount of capital coming into the L4 players, particularly autonomous trucking. Uber sponsored a lot of cash for that.

This L4 mindset that you saw early at the beginning of the presentation is starting to influence and strengthen a position about LiDAR being essential in this space. One of the things that also ties back to what I mentioned earlier about moving away from monolithic LiDAR. We talked a lot about cost and integration cost of a new sensor last year. There are a couple things. One is just the BOM cost about adding hardware to a vehicle, be it a truck or a passenger vehicle. The other piece is just the overall integration cost. Don mentioned it earlier here when he was on stage about you need to train sensors. There is software integration, incompatibility that needs to happen. This is one of the areas that AEye with our Apollo sensor and what I am going to be talking about next. SDV, software-defined vehicle. We are the SDV of LiDAR.

Imagine a sensor, we talked last year, we have a kilometer of range in our sensor. Think about that as a checkbook. Where do you take all that horsepower and performance? Does a car need to see a kilometer ahead? Maybe not. We can do other things with it and be super flexible. It is a new industry. That was great, huh? The understanding of what exactly the sensor needs to do is evolving month to month and year on year. We are super flexible in that regard, and we can sort of Here is a case where the red is we are spending a lot of our budget in the sensor, and the cooler colors are we are spending less. We can rebudget this, depending on the particular OEM passenger vehicle versus truck.

We are seeing this becoming a very important factor as we are moving forward. I mentioned supply chain resiliency. Same as last year. We are partnered with experienced tier one automotive supplier. Our footprint is global and flexible, but most importantly today, we are able to do manufacturing in North America. Again, this is a checkbox. You walk into an OEM purchasing department today, if you are not able to check this box, it is "See you later." Capital light. This is a value we have clung to at the company. Because our assembly and supply chain process is so very modular, we are able to scale up and scale down manufacturing as needed. Really what drives the cost for us is the working capital and the components, not the initial investment of the line itself. We are ready to scale.

The current line that we announced it late last year is ready to ramp up to 60,000 units a year. Again, this is a checkbox when you are walking into a purchasing department at an OEM is, "Okay, can you produce thousands of these things to support our first vehicle line?" I know there was a question that came in the pre-notes about NVIDIA. Integration cost. That is really what everybody needs to think about and what we are thinking about here. It is not just how much does the hardware cost, but it is also how much software integration and training work that you have to do to integrate a new sensor into the self-driving stack. NVIDIA has really been a great partner for us in this place. They have got their hardware, at least they have stated, in 35 plus OEMs.

It is really powerful to be able to walk into an OEM and be pre-qualified and very compatible with that platform. It is in essence, making a statement about the maturity of the product and the integration cost. They have been such a great partner, not only just about pushing us on the representative automotive sensing requirements, but also the automotive grade piece. We are a tech company. Automotive reliability is something that is evolving and developing for a tech company like us. NVIDIA has been a great partner in helping make sure that we are hardened and robust by being a part of their NVIDIA Halos AI Systems Inspection Lab. We got smaller since last year. I am happy to show here. This is our Stratos sensor. It fits nicely in the palm of my hand. We had a one-kilometer sensor last year called Apollo. This is now 1.5 kilometers.

Essentially double the budget, double the checkbook size of performance. You will see it tucked neatly under the rearview mirror in a vehicle application. That is actually slightly larger. That is the Apollo sensor. It is slightly larger. This is just really an output of that learning, meaning, monolithic sensors not required. That was keeping the cost of LiDARs in general higher. Imagine, for example, an OEM who would use a long-range sensor, be up above the mirror, and then two shorter range sensors in the side view mirrors or the headlights. That is actually a cheaper solution than one single monolithic sensor. This has allowed us to really take out some of the over-engineering that has been done for this.

This one LiDAR has to do the entire work for the passenger vehicle. It allowed us to increase in place performance where the OEM needs it most, and also drive down the cost of the product and make it smaller at the same time. Last but not least. When we were here last year, we talked about autonomy as the market. That has expanded since then to more generally what we call physical AI. LiDAR has certainly had a place in the broader market of physical AI. That is where the thinking part of the machine interacts with the physical world. I will tell you, defense has been a very hot topic for us. There has been a lot of inbound for us with unmanned ground vehicles in high risk situations, drones in flight, and also manned vehicles.

These are not sort of the hobby level drones, but larger drones that need to avoid things like power lines. Like we have in the picture there, we are incredibly good due to the tech stack to see power lines that are 3 centimeters in thickness at hundreds of meters away. The power of long distance sensing we brought into the automotive space is now paying off in defense. These guys fly low altitude missions. Power line is a big hazard. These are very expensive drones. They are not disposable. For example, we have been able to add a key technology there. Last but not least, counter UAS, that is the swarm size drones because of our long range and ability to focus energy out far and ability to see very small objects. We have been very busy in this space.

In fact, it has been making up the largest chunk of our revenue here the first half of the year. As physical AI ramps, we feel like we are in a solid position. We continue to have strong differentiation, the ability to point performance in the way that an OEM or other markets need it. They have a big budget. We give them a large budget, and they can spend it how they want. If you followed our earnings, this is the recent commercial announcement that was the underpinning. We had competition there that was fierce. Our ability to put the performance into a high frame rate allowed us to be unmatched in that particular market. Manufacturing, North America, and that high flexibility as we continue to learn. Physical AI market continues to develop. Our balance sheet is solid. We have a clean balance sheet.