Okay, so everyone talks about this industry and, you know, I think we've seen Arya just showed us also, you know, the size of the industry. People quote the $1 trillion size.
I guess the comment last year, the blue line was what was expected. It was expected to reach a trillion dollars at around 2030. This year with AI impact and everything, we're already seeing, you know, the yellow line become the trajectory. So this industry is continually, continuously growing and growing faster and faster than people expect. And I think that's maybe the big message in general of this conference.
Now, specifically, who is Weebit? What do we do? Weebit develops a technology that's called non-volatile memory, NVM, which is a memory that doesn't require power to retain the data.
Now, if you think about it, practically every device has that magical moment that you turn it on and you expect it to work. You expect there to be some software there. You expect there to be some data there. And that's really what NVM is all about.
Now, when you look at NVM, it has a major impact on everything. I already said, you know, practically every electronic device needs some sort of NVM in it in order to operate. It needs it for code. It needs it for data. AI now with all of the weights and stuff like that. And a good NVM creates differentiation. If you have a fast memory, if you have low power, etc., all of these things impact the device that they're in. And it's really across the board. It makes a big impact for the people who design the chips. We were talking earlier in the panel about the fact that there are the people who manufacture and there are people who just design chips and they don't manufacture anything. They send it to manufacturing. It really impacts everyone.
So, what is ReRAM? ReRAM is a new technology, again in semiconductor terms. It's new. People have been developing it for 20 plus years, but it's finally coming out to the market. And ReRAM is a much... First of all, the big advantage is it can go down to the small geometries. So, as I guess people in this room should know, the technology is constantly shrinking, Moore's law, etc., and we're going down to smaller and smaller geometries and people are now focused on going down to 2 nanometer and even below.
Flash technology, which is what exists today, can't really go below 28 nanometers if you push it. So ReRAM has a big advantage in this space that it can really fill in that vacuum that today flash can't fill. But it's also significantly lower power consumption. It's faster, simpler, and easier to manufacture. It has a lot of advantages. It is good for automotive applications. And AEC-Q100 is kind of the golden standard that you need to meet in order to be applicable for automotive, which today is a huge market. So ReRAM has really been moving forward nicely and Weebit is very proud to be one of the leaders in this domain.
Now, you know, ReRAM market, you look at in the left, you have there the 2024 numbers. The market was nice. ReRAM had about a quarter of the market. But I think today it's pretty clear to the market ReRAM is the future technology. It's growing. It's taking more and more market share. You can see that it's expected to grow to be more than 50%. But by the way, note, you know, today the market is about less than 150 million dollars and in the near future it's going to be already bigger than 3 billion. And so on. So this is the embedded ReRAM market and you can just see the growth. We're talking about CAGR of about 90% and that's really the exciting thing about this market right now.
Now, when you look at, and I'll focus, you know, we don't have time here to talk about all the different applications, etc., and people always want to hear about AI. So I'll tell you a little bit about the application of ReRAM for AI. And in AI, in a very simplistic way, you have two main activities. You have the learning activity. And then you have the inference where you use whatever you learned in the field. Today it's become very clear that you want to do the inference in the edge device, in those devices that are in the field. You don't want to waste time on sending everything to the data center so that it's analyzed and then goes back. You don't have the connectivity and latency and whatever. You want to do it at the edge. Now big challenges are that you need to do a lot of work. There's a lot of data. You need to do it fast. And those are key requirements from the technology. And of course, you want to have as many weights as you can on that little piece of silicon, because the more you have, the more accurate your AI is.
So, if you look at it the way that it works right now, so today what you have, because people, you know, in the existing world with flash, flash does not scale below 28 nanometer. However, the AI world is practically all of it below 28 nanometer. So, what do you do? Well, you don't have much of a choice. You have two chips. So, you have one chip with the AI engine and a big SRAM, and then you have another chip with the flash, where you keep all of your AI weights and coefficients and so on. And when you turn the system on, you need to basically copy everything from that flash into the SRAM, and then you can start running. Now, obviously, well, first of all, you have two chips, it's more expensive. The time to load everything from the flash into the AI chip, it's time. It's power, it's everything. There's a big security issue here, because now all of your coefficients are running on that communication. It's very easy to eavesdrop.
So, there are a lot of issues with this solution, and that's why the market is really looking to embed the non-volatile memory in the AI chip. And you can actually replace part of that SRAM. So, the SRAM, by the way, is a volatile memory that runs very fast. And so, in the first solution, you copy the flash into an SRAM, and then you start running. The world is now looking at replacing a big chunk of that SRAM with a ReRAM, so that you can actually have the coefficients inside the chip. When you turn the chip on, it's instant on. You don't waste time on copying and all of that stuff. And in many systems, even those, you know, fractions of a second are critical. So, here you have instant on.
You don't have the security issue. And that's a major issue in AI. And of course, you have a one-chip solution, so it costs less. So, a lot of advantages. And this technology exists today. This is something that we have a lot of people calling us up and just asking, "Hey, you have ReRAM. Can we embed it inside our system? How can we work with you?" So, there's a lot of activity around this. And this is something that basically now the world is already looking at moving into.
Now, the future is actually very interesting because the future is doing... One of the big issues in computers is the fact that today we use what's called a von Neumann architecture where you have a compute engine and you have the memory. And the compute engine always reads from the memory, does whatever it does, and then writes into the memory. And that communication takes a lot of time, a lot of power. It's really a big issue today.
In AI, you can actually do the computation in the memory. With the ReRAM, you can actually do the computation in the ReRAM and save all of that back and forth. And that's really a key activity that the world wants to do. We're working now with a lot of research institutes, with companies that are working on this. And this is a very exciting domain that I think we'll see emerging. It's still mostly in research, but I think we'll see it in the mid-term.
The long-term goal of the industry is obviously to just mimic the brain. Have the brain do, basically the memory will act as the brain. Really work like your brain. A ReRAM bit operates very much like a synapse in your brain. And that's the long-term neuromorphic technology which is again in research. We're working with research institutes. Weebit already showed its first neuromorphic demo I think in 2019. So, we've been in this and we've been looking at it. In the beginning it was very far away and initial, now it's really starting to move much faster. So, ReRAM based architectures are really ideal for this and there's a lot of activity going on there.
Now, you know, as a last slide, just who is Weebit. And in the panel this morning Ron was asking me, "What is Weebit doing and why is Weebit successful and what can other companies look at?" Well, our focus is, first of all, the company itself. We're an independent provider of the technology. We're now growing, but what's the secret behind it? It's the people and that's what I was saying earlier. We really focus on hiring the top-notch engineers.
You know, we don't just, if we need someone we don't just hire fast just because we need them. We really go and look and find the best people. And this goes from the board and you can see the people on the board. We have an amazing board with Dadi Perlmutter who basically, he and his team invented the Pentium processor, Atiq Raza who basically made AMD. AMD today is, people talk about Nvidia but if you go to conferences in the US the one after Nvidia they talk about is AMD and all of its success. It's grown very nicely. It's beating Intel in its own game and he's the one responsible for that. We have Yoav Nissan-Cohen who's the founder of Tower, one of the big foundries and really an amazing person and of course our Australian directors that help us on all the governance side, but the whole team including the VPs or the management and the lower level, everyone is really a very strong team.
And then the technology itself, we're focused and we made it a point to have a very simple and easy to use technology, something that uses standard material, standard tools and everything, and really is easy for the market to accept and that's the key point. Many companies have not realized that and they were trying to be too sophisticated, too smart, using some non-standard materials. The market doesn't like that.
And at the end, you know, our solutions, we're working with the fabs, we're working with the customers, with the product companies, and really making good progress there. So, you know, we're feeling very bullish at Weebit. I think we're making good steady progress and that's basically the company. Oh, and of course the customers, they're very important. I can't forget them and I'm very proud that today we're already signing up more and more customers. So I can skip that one and thank you very much.
