NVIDIA Corporation (NVDA) Earnings Call Transcript

Good morning, and welcome to How 5G and AI Improved Traffic Flows, Safety and Sustainability at UBC. Before we begin, we wanted to cover a few housekeeping items. At the bottom of your screen are multiple application widgets you can use. All the widgets are resizable and movable. [Operator Instructions] A copy of today's slide deck and additional help materials are available in the resource list. We encourage you to download any resources or bookmark any links that you may find useful. You can find additional answers to some common technical issues located in the help widget at the bottom of your screen. An on-demand version of the webcast will be available approximately 1 hour after the webcast and can be accessed using the same audience link that was sent to you earlier. Now I'll pass it over to Joao.

All right. Thanks, Megan. I hope everyone can hear me okay. Good morning to everyone. We have a very exciting topic to share with you today about how AI and 5G are coming together to help us address the traffic problem that I'm sure we all experience way too often. As Megan highlighted, please feel free to ask questions along this session. We will look forward to discussing this again at the final part. Let me start with a brief introduction of the speakers. Maybe I start with myself. My name is Joao Gomes, and I work at NVIDIA. My job is to partner with telecom search providers to create new AI services and also help with AI adoption. We'll have the chance to talk a little bit more about it as we go along this session. And with me today, I have a few distinguished colleagues from NoTraffic, Rogers and University of British Columbia. I would like to let them introduce themselves, and maybe we will start with Krista. Krista, why don't you introduce yourself?

Hello. My name is Krista Falkner. I'm the Manager of Transportation Engineering at the University of British Columbia in the campus and community planning department, which is essentially like our own municipality at UBC. Over to you, Neel.

Thank you, Krista. My name is Neel Dayal. I'm with Rogers Communications here in Canada. It's one of the largest telecommunications providers in the country, both offering wireless and cable services. I lead the innovation and partnership portfolio within the technology office here at Rogers. Over to you, Tom.

Hello, everyone. I'm Tom Cooper. I'm the Vice President of Public Sector for North America. And we're honored to be working with Rogers, NVIDIA's metropolis partner and the University of British Columbia as a customer and partner of ours.

Awesome. Thank you, Tom and [indiscernible]. As the [indiscernible] in general, you can see on the right part of the screen. I will start with a brief introduction on how AI, 5G are helping us drive market spaces, which is the new topic of the talk. Then Neel and Krista will dive into how Rogers and UBC are driving 5G innovation with the Aurora project. We'll learn more about it as we go. And finally, Tom will share the traffic solution, how it works and what are the problems that it addresses. And we will close with the outcomes of this project, which is pretty exciting. With that, let's get it started. All right. We are -- I think we are mesmerized at least I am with the capabilities shown by large language models and applications like ChatGPT, they're using AI and AI deep learning to create really super human-like content generation capabilities to assist us with pretty much multiple daily tasks that we go through like writing e-mails or like create computer code or even like plan for vacations and much more. It's really amazing. AI, no doubt about it, is the most powerful technology of course of our time and is starting to transform the way we live. And I think what is interesting is that deep learning, which is this foundational AI technology is not only transforming the way we live, but also the way companies and enterprises operate and manage their business and care for their customers. We can see here in this slide some examples of it. Stadiums like the Patriots Gillette Stadium in New England is using computer vision, AI computer vision to increase the convenience for fans when they go and watch a game, right? Use case that I really like about it is how you can use AI to optimize lines and queues when you want to get your beverage, right, which is the main reason many people in those 3 states. Airports are also tapping into AI big time. For example, Gatwick in London is using AI to compress the time it takes for an airplane to be able to fly again. So the turnover of our planes, which is obviously a big opportunity for improvements. And it's only restaurants -- it's not only airports but also restaurants. McDonald's, for example, is using AI to mitigate the labor shortage that really started during the pandemic. So they are using computer vision to anticipate orders. They are using natural language processing to automate the ordering process and also using different types of AI to reduce the losses. These are a wide variety, a wide range of examples, and we could really spend a session on it. Since time is limited, I put in the bottom of this slide a link to one of our e-books where you can learn more about it. That's kind of an introduction towards [ mark ] spaces. And if you think about one of the most mature type of AI that there is, is computer vision, which we highlighted or I highlighted in some of the examples before. Computer vision is really an amazing capability to think about it, right? We're basically getting the computers to be able to see and detect objects and patterns and video frames. And this really creates a limited opportunities to automate and optimize process. But then the question is really like how do you turn pixels in video frames into actionable insights with AI, which is the title of this slide? And then also, what is the role of NVIDIA? How does NVIDIA make that happen? What are the capabilities that we offer to make that easier, to make that a reality? As the example for me to do this, let's use the example of today's webinar, which is intelligent traffic management, right, that you see on the left here. You see this representation of intersected car and people. And in this example, we want to use video cameras as the sensors, right? They are deployed in strategic locations across the city like, for example, the traffic lines and then use this footage and this computer vision to detect the world users, to detect patterns and behaviors and then making decisions about it and optimize the whole thing, right? So that's kind of the use case that we are here to talk about. And then now if you think about from a developer perspective, what do you need to create something like that? Well, you'll have to define what are the objects and patterns you want your AI to detect. You did want to gather examples of data, right, and prepare it. Of course, you don't need to train a neural network, an AI neural network that is able to be deployed and then infer and detect those patterns. And you're going -- you're not finished yet. Once you [ done ] out there, you still need to optimize it, right? Because along the day, this will be deployed in thousands if not tens of thousands of intersections so we need to be able to implement this at scale, right? And we need to consider the architecture behind this, so on and so forth. So there is a lot of things that the developer needs to do. And NVIDIA approach to this and what we do is really to build tools and capabilities to make the developers more productive as they go along that process. And the way we do that is with what we call application frameworks, right? We have really application frameworks that tackles specific AI problems or technologies or modalities. In case of computer vision, the example in case here, the framework is called metropolis, which you see in the middle of this picture. Metropolis enable developers, for example, to leverage pre-trained models that NVIDIA created and that obviously not only speed up the development process but also reduce the costs significantly. And we also provide tools to help you take those pre-trained model and augment it to your own data so you can make it more specific. We also offer you the capability you often don't have enough in the examples of data of what you want to detect. So we also offer tools like [indiscernible] that you create those situations digitally and incorporate synthetic data generation into your pipeline. So those are the types of things that as a platform company, developer-focused that NVIDIA offers through our application frameworks. And then obviously, when you're done, you still need to optimize, right? And every case is different. Cities are different than airports, right, and they are factored behind it. So depending on the target architecture, you might want to deploy your AI into the [indiscernible], into the [ drones ] or into your max server or into the cloud. And all of those things are important aspects of it. I think coming back to our example here. I talked a lot about metropolis. But in our example here, right, we are talking about cities, intersections like urban spaces and put on cameras in strategic locations to do all of that magic, right? So for sure, we need partners to do that. We need partners like NoTraffic that use their expertise in traffic management to create applications that uses computer vision as the underlying capability to make that happen. So NoTraffic is a great example of a metropolis partner that is using -- leveraging that vertical expertise, combining with AI to create something really amazing that we learn more about it. But that's not enough, as I mentioned, like the sensors, the cameras are deployed in strategic locations in the city, in the lighting poles or the traffic lights, right? How do you collect that? Those locations are not really easily -- don't really have easy access to connectivity, power and other aspects. So we not only need application providers like NoTraffic, but we need companies like Rogers that are able to provide not only the connectivity for those sensors but then also access to an end-to-end infrastructure that can make it all happen, especially when you think about scaling this up not only to a city, but to a state and perhaps a country, right? And then not only NoTraffic is needed because they're bringing the applications, Rogers is needed because they're bringing the end-to-end capabilities. And then we have University of British Columbia, which is the proven ground where we can bring it all together, learn and figure out what it will pay to bring this to a national scale, right? I probably stole a little bit of the thunder that Tom, Neel and Krista are going to talk next, but I hope this gives you a general perspective of the capabilities that each one of us is bringing in this case and get you excited about we are going to cover next. With that, let me transition to Neel next. Neel, over to you.

All right. Thanks very much. Great introduction. So when Rogers started thinking about deploying 5G back in 2019, we were really on the search for a place where we could explore the potential of this technology. And so we conducted a bit of a search in terms of who would be the right partners where we could actually do some validation and testing of these capabilities, exploring new use cases but also conduct some, some meaningful research as well. And it became very obvious that UBC was going to be the right place to do a lot of this work. Many folks may not know this, but University of British Columbia operates as its own municipality. And for that reason alone, it really presents itself as a great place to become a living lab for us to test the 5G technologies. So in addition to the size and sort of scale of the campus, there was another few elements that were really attractive about partnering with UBC, and that's the breadth of the faculties. So they have a number of different disciplines and areas of research that were very appealing that could really help us exploit these various use cases. And then it also has an amazing engineering department as well. So there was a lot of strong capabilities in wireless engineering that would help us understand not just how to validate use cases but also to explore the potential of the actual network itself. And so since that time, we've been able to deploy 5G on that campus. In fact, it was the first 5G smart campus in North America. And we since had some other first as well, including the first 5G drone flight as well as the first 5G smart city deployment as well. But what is 5G? So maybe we can go to the next slide and really kind of understand what are the aspects of 5G? Because I think a number of folks think of 5G as simply speed. And really, there's lots of different dimensions to 5G, which include what we would sort of call the 5 main characteristics. And all of these are really important as we think about the difference between 4G and 5G. So primarily, I think most people appreciate that the capacity improvement from 5G over 4G. And that's a result of different technologies around how well we can use spectrum and the efficiency around the antenna, et cetera. But we are able to achieve much higher speeds, and we're able to serve more customers, both from a consumer perspective but from a B2B perspective as a result of that increased capacity. The other element, I guess, talked about quite extensively with 5G is something called latency and responsiveness. And so this is really important in a case where you need to be making some real-time decisions or you need less latency between the application and where the computational capacity is residing. This is for more low latency type of applications, which we'll talk about in the context of NoTraffic. Other elements that are really important are densities, so we can think about an intersection where you have a number of devices, whether they're people or vehicles, the ability to serve more devices in a particular cell location. It becomes increasingly important with 5G as well. The other pieces that get talked about a little bit less are uniformity, and this is really important as you get closer to the edge of the network. And so having that consistency across the entire network is really important for the types of applications that we can start to imagine when it comes to more of the B2B type of critical applications. And then finally, reliability becomes increasingly important. And as a result of the improved reliability, we have a number of mission-critical use cases that we can start to validate and test that are both important for B2B, but you can also think of things like public safety as well, taking advantage of this increased reliability. So we'll move to the next slide. And I think it's important for us to understand another key technology that's very adjacent to 5G, and that's multi-access edge computing. And Joao talked a little bit about this in his previous slides. But what we see with respect to 5G is as a result of this sort of increased capacity, you open up a number of use cases that then require a lot of computation because they're collecting a lot more data off of a lot of these devices. And as a result of that, you need to process that data in a manner that's requires that lower latency. And so while the network can produce and perform well, you need to bring some of that compute closer to where the application is. And in some cases, because of efficiency or security, you may choose to move some of that compute off of the actual device, but still somewhere in an area where you can access that computational horsepower. And oftentimes, what we've understood and realized and validated is the public cloud is a little bit too far for some of these applications to work effectively. And so this has created the emergence of what is called edge computing. And we partner with a lot of the hyperscalers to help us deliver that. Often these edge capabilities are powered with NVIDIA GPUs and really allows for -- as you start to build out more complex systems a lot of that computational capability to be present and delivering in the sort of sub-20 millisecond, sub-10 millisecond range, allowing for -- an enablement of a whole set of new use cases. So what have we done at UBC? And I think it's been really exciting over the last 4 years because we've been able to create a number of test beds. And those test beds range from smart buildings to smart transportation to drones and a number of different test beds. So we have -- and this is sort of a 3-dimensional view of the campus, but we have 10 scattered around the campus. So we also deployed 9 macro cell towers, which are delivering mid-band and low-band cellular frequencies. And then we've also deployed 6 small cells, particularly at certain intersections where we're then able to test improved capacity and lower latency using millimeter wave as an example. So a number of really kind of interesting test beds that have made a lot of different use cases available to us. And we can talk a little bit about some of the projects that we've done. And obviously, the key area that we're going to talk about today is around transportation. But adjacent to that, we've been able to do a lot of different types of use cases around connected vehicles and transportation, including measuring air quality through various air pollution sensors that we've deployed and are leveraging 5G as well. And over time, this is going to help us shape recommendations around the ITS infrastructure and help us configure some of our navigational patterns within the municipality to reduce the congestion and the pollution that we see. We've also done a number of projects around AR, VR. So we've been able to do some really interesting things around telepresence and understanding how to improve compression algorithms so that we reduce some of the downsides of using VR in terms of some of the nausea, et cetera, but also improving the efficiency of the network to deliver improved experiences. We've been doing some really exciting work around tele operations and using 5G to help deliver remote ultrasounds over a 5G network without having the need for an expert ultrasound operator in various locations and being able to deliver this remotely. Been doing some exciting work on smart forest, where we've been able to deploy about 50 sensors in some of these wildfire locations to help us use different sensors and inputs to predict the possibility around wildfires, and we're seeing some really interesting results around that. And then finally, a lot of exploration around drones and doing beyond visual line of sight deliveries and using that for 5G for data collection to also improve some of the traffic management that's required once we see kind of an environment where we're going to have more and more drones flying in the sky. So these are some of the exciting projects that 5G is enabling. We've done about 20 projects altogether so far at UBC. But maybe we can dive a little bit further into the Aurora project that we've been collaborating with Krista and the UBC team on. So one of the key technologies related to 5G is something called C-V2X. And we really believe that 5G is going to be important in that context in terms of delivering exciting new use cases where using cellular technology, we're going to be able to communicate between a number of different objects. And so on the slide, you'll see in a future world, we'll have the opportunity to have the vehicles communicating with pedestrians, to the network, to other vehicles and obviously to the infrastructure. And so we partnered with Professor David Michelson at UBC to help us deploy this technology and this infrastructure so we can really start to explore the benefits of cellular vehicle to anything technologies. And this is really going to help us not only explore the value of the technology and the benefits of it, but also to help us develop the standards and look at how we might remove any barriers to deploying more of this technology. Some of the other aspects that we can test in this environment are different sensors and really the properties around some of the wireless technologies and some of these limitations. And so to dive deeper in terms of the specifics of what we've done at the Aurora test bed, I'll hand it over to Krista for a little bit of a deep dive.

Thank you, Neel. So this map here shows really the extent of the implementation of the Aurora test bed at UBC. There's a few differentiating aspects to it, mostly because of the complexity around who owns what at UBC, but there are 14 intersections where there are instruments on the campus. 6 of them are on ministry roadways, and 7 of them are at UBC intersections. And then there is one research intersection, which is the test bed. And that is kind of a mobile test intersection that students and Dave Michelson use for their learning. And then we have of note, there's the 3-millimeter wave small cell deployment at 3 intersections on campus. And then in addition to that, there's the Transport Canada drive test vehicle that Professor Dave Michelson and his students use to do further testing around the campus. Next slide, please. Over to you, Tom.

Thanks, Krista. So NoTraffic is the world's leading mobility and traffic management company. And we're being recognized globally for innovations to reduce congestion and emissions and make roadways safe for everyone. Next slide, please. The time to change dramatically -- there's massive growth in our inner cities during late 1800s and early 1900s and also the invention of automobile changed things forever there was traffic multi-modes of the horse and buggy and cars. And to control this, the traffic and help the most effective and safe utilization of roadways, traffic control officers went on to intersections to manage the traffic. What they're able to do is observe what's there. I see pedestrians, I see horse and buggy, and car, and they made decisions to move traffic based upon their experience and also their best judgment. During the industrial revolution, there's a big push to change repetitive tasks and push them down to machines. So it was generally thought that, this is a repetitive task, moving traffic. So devices were built to essentially replace the traffic control officer, drive light bulbs that brought up different colors based upon observations from the past programmed into this machine to move traffic based upon preset times and intervals. Well, times have changed dramatically, right? So if you go -- most of the major urban areas, they've had massive influxes of exodus and influx of people in various times in the COVID period as well where traffic was completely -- [indiscernible] variable, unpredictable and that continues today down to new modes of personal mobility. Scooters were not something that was contemplated in traffic in roadway planning 30 years ago nor was this phenomenon of [indiscernible] deliveries and home deliveries of goods. Then you've got the unintended consequences of all this confluence of events from distracted driving down to infrastructure challenges has resulted in an unprecedented level of accidents, injuries and fatalities in intersections. The technology exists today from the culmination of all the pieces we're talking about to solve these fundamental challenges. Next slide, please. So the requirements for agencies and for intersections and for the population, what we use in general, is manyfold from making the roadways as safe as possible for pedestrians and bikes and also optimizing the traffic flow for them. And the use case that Krista is going to talk about later really focuses on that. The previous mode of moving traffic and management traffic was focused on vehicles. That pushes many vehicles through with the most efficiency and the least amount of congestion. Well, that needs to change, and it doesn't have to be a balancing act anymore. Agencies need to know about accidents, insights about where they're having them to be able to proactively put countermeasures in place to prevent them, but also allow emergency vehicles to respond more efficiently because the quickly -- the more quickly an emergency vehicle can go out of first [ responder ] -- go out and respond to an accident or injury, the higher the survival rate. So that's key. So things like emergency vehicle preemption, then we have this upcoming world of connected vehicles and connected objects that needs to be supported as well. So this and many other specific focus points while reducing emissions and congestion are a major, major component of moving traffic. Next slide, please. So first, I want to stop, and I want to thank traffic engineers everywhere for the limited tools they've been given, and they've been doing an amazing job. They no longer serve the purposes of the modern requirements. So some problems at the intersections. That device that you saw earlier, the electromechanical timer had a very specific purpose, just changing lights at preset intervals. Well, that, of course, migrated to an electronic version, which essentially still emulates the same thing. There wasn't a requirement to have a lot of compute power at the intersections because the decisions that had to be made were very limited. And this is fact that less than 1% of the intersections have more compute power than your average microwave oven or washing machines. So -- but again, specific products built for specific purposes. The bigger issue is there's very limited data available about what's happening with the intersection. So traditionally, for detecting that you have something there would be either a pedestrian push button to say, hi, I'm here. I am a pedestrian, or something called [indiscernible] loop, which is a piece of wire on the ground that's triggered when a big piece of metal sits on top of it. It doesn't give you a lot of insights about do I have pedestrians? Do I have, let's say, disabled roadway users that need to cross -- pedestrian cross way in what time. So the fact that we don't have the high-resolution object tracking and knowledge of roadway users prevents rollout and deployment on scale of applications that can give significant safety insights and operational improvements. Another major piece is that less than 70% of our intersections are even connected. So without being able to roll out situational awareness at multiple intersections, connectivity is a major piece of that. That fundamentally comes back to traffic delays typically related to there's capacity issues certainly, but the typical construction is for [indiscernible] traffic based upon preconceived ideas and observations in the past. Well, when things change next Tuesday, this whole architecture has limitations that leads to traffic congestion and also creates major safety issues. So a solution set for this, next slide, please, is taking the best that we can get from artificial intelligence from sensors, from algorithms that can effectively move traffic and also wrapping that in a bundle with support because a big part of this whole equation is you have to have products that supported this or mission-critical. So NoTraffic platform is based upon a couple of fundamental building blocks. The first thing that we do is we provide high-resolution, high-accuracy tracking of all the objects at the intersection by using a plug-and-play IoT device sensor that uses video and radar to deal with all sorts of lighting and climatic conditions. That is the first piece of it, processing down the edge. Second piece is using all that data to create a very deep, broad understanding and situational awareness of what's happening at the intersections from pedestrians, bicycles and vehicles. The third piece is support. Most agencies don't have a very large support staff, and that's a traditional challenge for mission-critical systems. What enables all of these pieces is something we call Mobility OS. So this is our marketplace for applications. This is a software-defined architecture for various traffic applications. This is very, very different in the past. And in the past, different functionality was built based on different hardware pieces. That could be disparate vendors. But the challenge is taking different pieces of hardware, trying to integrate and support as an agency to roll out additional applications. So we've taken this. We've used the sensors and the software architecture to create which is essentially a cell phone model for the intersection, the baseline of hardware that provides high situational awareness of high accuracy that allows agencies to enable different features through software rather than going through complex construction upgrades and a lot of additional costs. Next slide, please. So this is a fully software-enabled platform for traffic management. Again, it's a single hardware platform that's future-proof, that has the sensing capabilities for the high-level situation awareness, connected vehicle support, but also allows the future-proofing architecture where different features are defined in software, which is very, very different in our industry. And it solves a lot of major challenges for agencies. Last slide, please. What this does is enables agencies, again, to select different features through software and some key features. I think some of the most exciting benefits of what we're able to deliver with our technology is giving agencies deep insights into safety issues and not just waiting for them to happen, but giving them deep insights into where I can apply countermeasures to improve safety and performance but also measure the outcome of that. I think there are some exciting things that we can do with this. We've had the opportunity to work with the University of British Columbia on a use case, which is very interesting. And back to the point earlier of traditionally with traffic signal management, the focus was on moving vehicles. And if you wanted to move bicycles or some other modality transportation, it was -- you had to do a balancing act. So fortunately, we're able to serve the objectives of primarily moving bicycles and pedestrians first, but also allowing the congestion and the travel experience for vehicles to improve as well.

Okay. So I'll get into the -- really the overview of what we did at UBC as this combined team here and then really what the outcomes were for everyone to understand. So a bit of background. So I gave you a light background of where instruments were installed across the campus. So that's replicated again here in that lower right-hand map. We deployed Rogers and NoTraffic solutions at 9 intersections, and we applied the optimization at 5 intersections along Wesbrook Mall. And these were signalized intersection, of course. We did deploy the equipment at roundabouts as well, which has also been a really good learning opportunity for us and the NoTraffic team. So for the optimization task, I worked closely with the NoTraffic team to identify what our goals were with this deployment. Wesbrook Mall at UBC is our people-moving corridor. It's one of the main routes in for buses, and it's a heavily used corridor mostly for cross-street pedestrian and cyclist activity and also cross-street movements for some vehicles, but although very limited. So it puts a bit of a unique challenge for us along this corridor where it wasn't just about moving traffic as Tom mentioned. We were very interested in prioritizing our active modes, biking, walking and rolling as well as our transit. And so with this task, we optimize the Wesbrook Mall corridor for people moving -- almost moving along and across that corridor, which was unique at the time. But the NoTraffic team put in the optimization goals, and we were able to achieve some wonderful results. So a few other things about this deployment. That top-right corner shows a screenshot of what the NoTraffic solution sees. So that's the sensor detecting the vehicles, pedestrians, and it would detect cyclists as well. So you can see how it captures the images. In the detection mode, the object-based detection classifies and tracks the objects. And what is interesting to note or important to note is the radar-based detection helps in poor weather conditions. There is no periods where data is not able to be collected due to rain, snow or other weather conditions. And then, of course, the sensors are connected to the cloud using the 5G LTE communications and accessed anywhere using that virtual management center dashboard. So me as the client or UBC as the client is able to go on to this dashboard to get data information, counts at any time. It gives access to those video streams. There's also the platform that Tom was mentioning, monitors and provides real-time alerts. It identifies when there's communication error, when there's an accident. And that is all immediate. And for a small municipality like UBC, we do not have the resources for people to be watching all of our intersections 24/7. So this has been very useful for us to notify our operations team when there are errors and for accidents, knowing a bit more into the accident history and understanding, meaning why an accident occurred that may require some civil or lane configuration or lighting or anything like that, we're able to better understand why these things are happening. Next slide, please. So getting into the results. We collected data for analysis a 2-week before and 2-week after period to kind of get a before and after understanding of the corridor. As you can see in the table to the right, there were substantial savings in all modes. So that we analyzed, as I mentioned, 5 intersections. And we compared kind of very common metrics in the industry. We have transportation targets here at UBC, much like every other municipality with regard to our greenhouse gas emission reductions as well as our pedestrian optimization or priorities for our active modes and transit. So looking at the results from the 2-week period, we can see that there is 2.8 tons of greenhouse gas emission reductions just along this corridor over that 2-week period. And pedestrian or active mode delay was over 93 hours just in that 2-week period. And then similarly, the vehicle delay was over 181 hours. When you translate that to an annual number, the numbers are astronomical and equate to a total economic value of 165,000 plus just for those 5 intersections at UBC. If this were to be scaled to a city, a full city environment like the City of Vancouver, where there's approximately 727 intersections, you start to see these results pay out immediately. And this was really important for municipalities, particularly as we're trying to fit more people in the same space. So it allows us to prioritize those active modes transit over our vehicles. But it's important to note, it's not one or the other. It's -- there's no compromise here. Our optimization that we applied with NoTraffic, it was to prioritize the pedestrians and cyclists just as much as the vehicles that we still were able to achieve a significant reduction for vehicle delay as well. And this has been one of the biggest outcomes of interest for many people to see that it is not just one or the other. So in summary, the NoTraffic system works well in a small environment -- just go back to the last slide, please, at the UBC context and can have very scalable and major improvements on city-wide basis. In terms of our next steps with the NoTraffic team, we are continuing to work with them to explore adding transit priority to the optimization and then scaling it up to a few more intersections on campus as well. Next slide, please. One intersection of note that I wanted to bring up is the intersection of [indiscernible] Road and Wesbrook Mall. And the reason why I want to focus on it is this is a bike route into campus and one of the more major bike routes, where cyclists and pedestrians cross the intersection across Wesbrook Mall to get into that core of the campus. You can see before and after results of that 2-week period, on the x-axis is hours, kind of starting at midnight on the first day, and that's over a 5-day period. And then the pedestrian cyclist delay is on the y-axis. So if we just looked at the daytime average, before the optimization was applied at the intersection, we're seeing pedestrians, cyclists wait at that intersection to cross Wesbrook Mall for approximately 38 seconds. That's a really long time. And what happens is people tend to take it into their own hands to cross at will and find gaps. This, of course, creates other safety concerns. So when we applied the optimization with NoTraffic, you can see that the daytime average went from 38 seconds to 22 seconds, a substantial difference. If you think about it the seconds, if you can kind of picture yourself waiting in an intersection, knowing that, that's 14 seconds of savings for you, it's pretty substantial. And so we're really excited to see this, and I personally ride in to UBC every day, and I've experienced the savings. And so I have to say I'm very pleased with it as well. And I'm really looking forward to seeing kind of where we can improve this across the campus more. Thank you. Next slide.

Thank you, Krista. That was impressive, right? So we are reaching the end of this webinar. I hope you enjoy and learn from this combination of partners and components, right? And the power of it, I think Krista articulated so nicely the potential and the outcome of a project that UBC has done, right, and how flexible that is to adopt to different targets and different goals that different places might have and the potential that can deliver if you bring it to scale. So I think that was really, really nice opportunity for me. And this combination of connectivity and access to infrastructure that telcos like Rogers can provide with applications that relies on AI and vertical expertise that companies like NoTraffic brings with the intellect and the capabilities to learn and articulate those learnings that companies or universities like UBC has, you can see how powerful this combination is, right? And I think intelligence traffic system, as we just have seen, it's a great use case. It has immense potential. But if you think about it, the same combination of 5G, access to compute, access to ecosystem, right, can also transform many other industries, right? It's not only traffic management in university, cities or countries, but it can be like airports, stadiums, it can be manufacturing plants, retail, right? So the potential is immense, and I think this is the journey that we are on with all these new technologies and the ecosystem coming together. And that's one of the main takeaways of this webinar in my opinion. With that, I'd like to thank you all for listening in, and let's now jump into the question and answers. Give us a minute. We're sorting through the question, and we'll get into it very quickly.

All right. We have some great questions here. I hope you enjoyed the content. Let's address a few questions that popped up during the session. Question number one, this is a great use case that brings real results. Why isn't this deployed everywhere if there are such great benefits? That's an awesome question. I think let's -- maybe, Tom, I think you're in a good position to kick us off, and then we can invite the others to comment.

Thanks, Joao. Yes, that is a great question. So in time, advanced technologies like NoTraffic, they will be deployed everywhere. The agencies are necessarily conservative by their nature, right, as a primary mission safety for all roadway users and also being [indiscernible] as taxpayer dollars. So before any innovation is void at scale, has to be proven, validated. In many cases, you have to go through a rigorous approval process with agencies to ensure compliance with the requirements. So use cases like UBC, which we've just gone through, it further validated technology and the value to the public. And simply, agencies need new cost-effective and scalable tools to continue serving the changing demands, which are continuously changing and also prepare the future of infrastructure for our future requirements.

Awesome. Anyone else wants to comment? Neel, Krista, anything you guys want to add?

Yes. I would just add, I mean, it's unusual. I mean, one of the reasons I think I mentioned this as well is just how we were able to collaborate with UBC because they operate as a municipality, but because they're also a research institution, they have the ability to do more experimentation. I think once we start to see municipalities beyond kind of the UBC-like model, where they're willing to create these innovation zones, these demonstration zones to have these test beds where they can actually in an isolated manner, I'd see the results of some of these applications and then deploy them more broadly. Once we start to see that type of sort of mindset from municipalities, I think you'll start to see these solutions scale quite quickly. But today, it's really -- I don't think all municipalities have wrapped their head around that type of innovation model, which they need when you're talking about early-stage technologies like this.

That makes sense. And then we have another question here that I think relates to the first one, right? Why this isn't deployed everywhere still, right? And the second question is about what is the role of 5G into this solution, right? I think 5G has -- 5G and connectivity in general has a very important role here to play, right? Because it's not necessarily easy to put these sensors as intelligence everywhere. You need to reset connectivity. So maybe what is the role of 5G in this solution? Maybe, Neel, you want to kick us off with some comments, and then we can go around the room?

Yes, absolutely. There's sort of what is 5G providing today versus what is 5G going to provide in the future. And so today, 5G, the way the solution is architected today, I think was a question earlier around whether it's on NSA or SA as well. So today, it's deployed in a nonstandalone environment, the solution NoTraffic is running in the public cloud. So 5G is helping to minimize the deployment costs of providing fiber connectivity to the intersection. So this way, you're able to deploy the solution much quicker because all you need is power, and you need to be able to mount the sensors. You need 5G because you need the capacity and the bandwidth to be able to stream some of the data sets to the cloud. But over time, the idea here is that you will scale this solution across many intersections. And you're going to need a really high throughput, highly reliable, low latency network to be able to optimize these traffic corridors. And so the amount of data that's going to be required to stream all of that data sets and then to do the optimization, that is where you're going to be able to really lean on the multi-access edge computing capabilities that I talked about a little bit earlier. And then finally, if we start to think about the sort of connected vehicle, the C-V2X type of use cases where you're now communicating with vehicles, pedestrians, bicyclists that are all approaching the intersection and that are all requiring data from that intersection to make decisions, you're going to have another requirement around 5G as well. So hopefully, that answers the question.