BrainChip Holdings (ASX:BRN) Presentation, FNN Online Investor Event, April 2020

Company Presentations

  • Email Alerts for:

BrainChip Holdings (ASX:BRN) President & CEO Lou DiNardo introduces the concept of neuromorphic computing where only events or spikes are processed, its advantages and applications in edge computing , the company's product and business development timeline and financials.

YouTube video links:

Incremental Learning:
https://youtu.be/BLNr-KXPxhE

Gesture/Robot:
https://youtu.be/aV_P7jKQgqA

Lou DiNardo:
 
Clive, thank you very much and Martin as well as two partners. The Financial News Network and Shaw and Partners have been both loyal and big supporters and I want to thank you for that. If you could move to the presentation. While this is an introductory slide, it occurred to me as I was putting this presentation together, that this really does speak volumes about what we do. This is basically showing, you can call them spikes or you can call them events. We process events. The more important the event, the larger one of those bars would be. The less important, the bars would be less and if there's no bars and no spikes, we don't process information.

That is probably the biggest if you don't want to get into the technology at all, that is the biggest advantage we have is that once we go into what we call the event domain, when an event occurs, is it important? How frequently does it happen? How intensely does it happen? If it doesn't happen, we do no processing. That's what gives us the capacity to do ultra low power. We do not process events that are anything other than what we process events only if they're none zero. If there are zeros, we don't process. Zero times one is zero, one times zero is zero. Why bother doing the cycles of energy and burning power? Well I know this is a slick looking slide. It also has a lot to communicate.

Can we go to the next slide please?

The agenda I've only got 15 minutes. The agenda is very short. I'm going to keep it sweet. If you want more detail on market sizes, all of the data that goes with it, you can go to our website. We do quarterly webinars with the help of our friends at FNN and we do lots of quarterly updates and we're going to touch a bit on company background for those potential investors that are new to the company. A bit more on product development for those that are more familiar with where we've been and where we're going. Business development update and a finance update. I will keep this short and sweet. We are going to do an interview later this evening. Anybody who wants to listen in, that would be wonderful as well.

Next slide please.

Actually stay here for a second, please, go back, go back please. The background here, I have poached and repurposed some material from our recent redo of our website. One of the things, and I will point to it later, is when you look at robotics generally, the opportunity for us to control robotics in this case, this is actually a little dated. If you go to the last slide when we get to it and you go to our website, you'll see that we can control a robot with a hand gesture. Turn right, turn left, go up, go down and up.

The background here again, it's important because robotics and it's not a robot delivering you a martini at your couch. Robotics is a very big and heavy industry environment. But the background here tells a lot and if you go to the... I could not embed the actual video here, it would take too much time. But we will provide in a later slide the link to the hand gesture for robotics. We'll talk about business development and finance update as well.

Please go forward.

Company background, and many of you on the phone know the company well, but there are others that are new to the company so just a very, very quick background. We were founded in 2013. BrainChip has a great meaning. Peter Van Der Made, neuroscientist, Anil Mankar, chip designer. Basically what we do is we take what Peter knows about neuroscience. Peter and his team, we have a group of folks, I'll talk about them in a few minutes, computational neuroscientists. We take what we know about neuroscience and we implement it in digital logic. Peter and Anil, that's Brain and Chip. We did a public offering, actually a reverse takeover in 2015, company of Azania, we're listed on the ASX exchange. I think one of the things that investors should be cognizant of is we are in my mind and to my knowledge, the only pure-play artificial intelligence company that is public.

You want to invest in artificial intelligence, you could buy Nvidia, you could buy Google, you could buy IBM, you could buy Intel. Well you get through a whole basket. We are, in my view, the only pure-play artificial intelligence company that is public. So all of the money that's being made in artificial intelligence startups is being made by venture capitals, maybe private equity, but primarily venture capitalists. This is an opportunity for investors to have a piece of a pure-play artificial intelligence company on a public market. We've raised about AU$53 million. I use approximately because there's exchange rates and numbers get a little fuzzy, but basically 53 million.

I can only report here what we published in our December quarter, although our foresee will be coming out in the next couple of weeks. We had 7.6 million in round numbers on hand at the end of December quarter. We did put out a forecast of what we would earn in a quarter and [inaudible 00:05:47] to our update and we did a recent capital raise with our friends at Shaw and Partners, which was three million off seed and it was taken up by two large and very high quality institutional investors in Australia. It does help us bring stability to the register. They're very, very strong believers of the company, firms that I've met many times. Unfortunately, I can't come to Australia right now, otherwise, I would've been there for the race.

Our engineering resources are primarily in California and Toulouse, France. California, Southern California, we have hardware design, we have computational neuroscientists who are researchers and in Toulouse, France, it's a team that we acquired when we acquired Spikenet in September of 2015. Again, software development as well as algorithm development and our computational research scientists. We do have five people that are dedicated contractors to us in India and we will likely be opening an India design center and rather than have them as contractors, so they'll be wholly owned.

I put Japan here because of our relationship with Socionext. Socionext does the backend. Akida product is currently in wafer fabrication. We're out of development and we're into fabrication and Socionext is an absolutely phenomenal partner. Not a well known name, but if you think about Socionext, just think about Panasonic microelectronics and Fujitsu microelectronics. They were taken private by the Bank of Japan. We have administration, sales and marketing in California, primarily in the San Francisco area. Certainly in Europe and in Australia as well because we are a public company, we do have our footprint in Australia.

Next slide.

So I don't run out of time. I think it's just fair to take a few minutes, a few moments at least, to talk about what Akida is. Akida means spike in Greek. But you can think about spike as an event. Spiking neural networks are going to be an evolution. In the interim, there is a whole bunch of what we call computational neural networks, everything that's going on in the artificial intelligence world. What the team has been able to do is develop a chip that can do spiking neural networks in a native domain, but during the interim, while people are evolving to spiking neural networks, it really is an event-based processor.

We can take existing convolutional neural networks pretty much from anybody. And all of the layers that are included in that convolutional neural network can be implemented in the Akida device at ultra low power. Again, because of that first picture that I showed, where we do not process anything that has a zero. That allows us to do extremely low-power edge AI applications. You could think about automotive ADAS or advanced driver-assistance systems. You can think about autonomous vehicles, smart home applications, medical diagnostics, medical applications and I would say more of the industrial internet of things but this is what the chip is. This chip, this picture here is not to scale.

If we go to the next slide. Please.

You can see what the device once packaged, it's going to be smaller than a U.S. dime. It's 15 by 15 millimeters. The U.S. dime is about 17 and a half by 17 and a half, I put this together. I think it's reasonably well to scale, but it gives you a sense of the power that we can bring to the edge, the capability we can bring to the edge at ultra low-power. Microwatts, hundreds of microwatts, that's millions of a watt, rather than sending all of the data back to CPU or GPU or some hub in the trunk of a car or back to a data center, sucking up bandwidth, WiFi network or local area network.

Our strength here, I think the first thing that we delivered was, which is a very big differentiation, is the Akida development environment. There's a lot of companies that are laying claim to being able to play at the edge. Number one, they're not a complete solution. They need a host CPU, they act as an accelerator or a math coprocessor. Our device is complete. Everything in our device and everything in the intellectual property that we're marketing as well is a complete neuro fabric. It runs the network and it does all of the processing.

The development environment allows people to implement whatever networks they currently have, which again, can be convolutional neural networks that are existing or convolutional neural networks that they attempt to implement. Eventually they could jump to it in a native spiking neural network, but this is probably one of the most sophisticated design tools in the artificial intelligence space, particularly when you're talking about naming at the edge.

In addition, we have a field programmable gate array where you can do emulation. So you can do simulation software, you can do emulation in hardware. You can't fit everything in an FPGA that you could on a device, but it does give you the opportunity to run smaller networks and see that the logic design works. We're currently marketing the Akida intellectual property. It's very advanced. I'll talk a little bit about that in a moment. But first, commercialization.

The Akida device itself, the hardware is in wafer fabrication now. We're developing all those software support tools, all of the design support tools and evaluation board. When that device comes out of fab, we're ready to go. We've got a sales team in place that's already canvasing the marketplace for target customers. We're being very selective because it's going to be a bit of handholding, you're doing the first iterations we will have an Akida card, likely what we call an M.2 PCIE interface card and we will likely have and Akida dongle USB plugin.

Let's move on.

So I don't run out of time. Quick business update and again, we do market updates every quarter we focus on the AI edge market. I avoided putting in all the [inaudible 00:12:31] charts slides. The AI edge market is a market that is going to grow at least based on multiple forecasts from reputable sources. It's going to grow from a couple of billion or three or 4 billion to something over $50 billion in the period between now and 2025 that is our target. What that market needs is very compact, extremely low power solutions that are very flexible.

In our case, we have advantages that go well above and beyond just being low power, incremental learning. I wish I could have put the demo in here. But I can't go to the website and I have the links in a subsequent slide. We are marketing our intellectual property. We've got very advanced discussions going on with customers in Europe, United States, China. Those are probably... I just ran out of time. My timer just ran out.

We are opening innovation and research centers. The first is in Western Australia. Peter Van Der Made for personal reasons. Wanted to go back to Perth. He's starting a innovation and research center there. It turns out that there's great talent. We have a partner program that we're advancing, which is bringing people from kind of playing around with the development environment to moving them to, proof of concepts. A development license, which takes them from proof of concept to you buy a license so that you can develop on our platform and then a production license. Which is when you go to production.

Next slide please.

Just a very quick financial update. I'm not going to do a lot here because we have our 4C coming out. I can only speak to what was said in the last 4C. Commercial contracts are in negotiation. We've got one that's been on a kind of long pole out of Europe. We've got one coming out of United States in Detroit. We've got a couple that are going on in China. These are going to be either proof of concepts, which is the first step toward a true commercial, large engagement, but nonetheless it's validation that you've got very serious customers.

A Tier 1 automotive supplier in Europe, a Tier 1 automotive manufacture in the United States, a couple of [inaudible 00:14:56] in China that are more designed services but have very big relationships with major customers in China. We've got good institutional support. Thank you from our partners at Shaw and Partners again. I think that the fundraise was well done. It's a very, very difficult market. We do have a diverse shareholder base. We've got insiders, we've got institutions and we have a very long tail of retail and [inaudible 00:15:23]. Again, just because I need to be careful here. Our quarterly 4C and our update. Are scheduled I believe for 30th of April and then we'll have more financial detail there.

Next slide.

Just as a quick summary, we've got very strong potential customer engagements, some moving to commercialization of the IP now. Others that are looking forward to us coming out with our engineering samples in the early part of Q3 2020 have a very strong patent portfolio. Peter pays a lot of attention to protecting our IP and I appreciate that. We are grinding on cost containment and these things tie together cost containment because we're a company on our formative years and we're looking forward to strong commercial engagements but the cost containment is important.

With respect to health and safety. I'll use this just as an overarching comment on what's going on with the corona or COVID-19 it has made life challenging. No doubt about it. I would've spent the entire month of February on the road between Seoul, South Korea and Shanghai. Unfortunately, it's like Hotel California. I could get in but I probably couldn't get out so there's a lot of conference calls internally. The company has been working from home for several, several weeks now.

In some respects it's ironic, it feels like it's more productive because people have nothing else to do but actually keep in touch with each other. We're on conference calls constantly, but the health and safety of our employees is paramount. It's not getting in the way of us having engagements with customers. Maybe a little harder to actually close a deal because you want to look somebody in the eye. Health and safety is important. Cost containment is important.

These are the two links that I mentioned here. I don't know if they're going to be clickable or you may have to cut and paste, but I would encourage you to go look at the incremental learning demo and the gesture for robot demo. It's going to give you a good sense of just how powerful the Akida technology is.

That's it. You can reach me and I know many of you do. You can reach me at this email address or you can go to brainchip.com submit questions and we'll be happy to get back to you. With that, I thank you and we'll have a 4C and an update coming out shortly.