Editor’s Note: This article is part of Joanna Makris’ Fireside Chat series, where she provides retail investors with the scoop on the hottest technologies and trends from today’s business leaders, industry experts and money managers.
Our latest Fireside Chat is a front-row seat to to one of the biggest investment cycles of the next decade: autonomous driving and vehicle safety.
Fresh off the company’s presentation at the IAA Mobility show in Munich, I had an opportunity to talk lidar with Dr. Matthew Weed, Senior Director of Product Management for Luminar Technologies (NASDAQ:LAZR). Weed, an optical scientist and member of the original engineering team at Luminar, has spent his career focused on commercializing advanced optics and photonics technologies in drug delivery, medical diagnostics, and now, the automotive industry.
When it comes to automotive driving and safety, most cars today use sensors including cameras and radar to help guide features like automatic emergency braking and lane-keeping assist. But, now there’s a new technology that promises to take automotive intelligence to the next level: lidar (Light Imaging Detection and Ranging). Basically, lidar uses laser light sensing to help a car develop a 3D map of its surroundings.
Lidar helps cars do the “seeing,” making cars safer as a result. But this technology also has the potential to completely change the way we think about driving — and even potentially create an “uncrashable” vehicle in the future.
Opinions certainly differ on whether self-driving cars are part of some dystopian Orwellian future or simply the logical progression of the merger between humans and machines. Regardless, you won’t find many cars using these high-powered laser sensors, for one very simple reason: they’re too expensive. The high price tag of early-stage technology means that fully-autonomous vehicles have been out of reach and way above budget for the general public.
But Luminar says that’s changing. It’s one of a handful of lidar companies saying they’ve have found creative ways to bend light and use technology like network processors to bring down their sticker price without sacrificing performance.
Eyes on the Prize With a Laser Focus
Luminar is the pack leader when it comes to lidar: it’s the only company with a production contract right now. It also has a $6 billion market cap, despite having generated only $14 million in revenues last year (remember, we said it’s still early).
The Palo-Alto-based company may be bending light, but it isn’t bending the truth when it comes to its relationship with Volvo (OTCMKTS:VLVLY). The Swedish automaker has partnered with Luminar to use lidar in new vehicles capable of hands-free, eyes-off highway driving beginning in 2022. That includes the upcoming Polestar 3 from Volvo’s new electric-car division as well as a range of Volvo-branded cars and SUVs.
That’s no small endorsement. Meanwhile, the rest of the lidar pack is still in development-stage relationships with no orders to show just yet.
Luminar’s Iris sensor uses 1550-nm wavelength lasers, as compared to the 950-nm wavelengths used by competitors. Critics say a major downside to 1550-nm lasers is that they require the use of more exotic semiconductors, such as indium-gallium arsenide, which tend to be more expensive. Weed discussed how Luminar has managed to bring down the cost and why the company prefers it over 950-nm, saying in part, “shorter wavelengths are much more hazardous for eye safety concerns” (For more background on the great “1550 versus 950” debate, check out my recent Fireside Chat with Omer David Keilaf, CEO of Innoviz Technologies (NASDAQ:INVZ)).
Regardless of their technology approach, lidar stocks have been under recent selling pressure, and LAZR is no exception: Luminar is down roughly 44% year-to-date.
A few factors are at play, ranging from a rotation away from high-growth futuristic stocks, to several new names to choose from in the self-driving space, including Ouster (NYSE:OUST) and Innoviz, both of which went public earlier this year.
But the promise of lidar technology, coupled with endorsements by nearly every major automaker, make this a perfect time for investors to do some diligence and make sense of this space.
Read on to hear Weed’s comments about modern car safety and why Elon Musk is wrong about lidar. We’ll get a better understanding of how the tech works and discover how Luminar addresses some of the shade thrown from other suppliers in this hotly contested space (950-nm lidar players, we’re talking to you).
Oh, and that hot gossip from the IAA show? About Luminar “pivoting” into safety from autonomous driving? Weed responds. Finally, he shares with us his vision of the future of autonomous driving: the ultimate merger of man and driver.
Luminar recently showcased a rather provocative video showing Tesla (NASDAQ:TSLA) vehicles which appear to be on autopilot crashing into standing crash test dummies. Tell us more about that. And tell us about your views on the state of autonomous driving today.
Dr. Matthew Weed: Good question. And there was a lot of subtext that [isn’t] always obvious to pull out from everything that we’ve done. So what we’ve shown, and it’s a good culmination of what we’ve been driving to as a company for many years now, is really to show what’s possible at the vehicle function level with this new sensing technology called lidar. As you mentioned, everybody in the market is trying to figure out now, “what is this lidar thing?”
Really, what it’s all about is giving a vehicle better understanding of what’s going on in the environment around it, in three dimensions and the full context. And what we showed at IAA in Munich last week is the result of our team building an entire software stack, from raw sensor data to actuation of a vehicle, the kinds of safety systems that are possible with this kind of sensing.
And so what we showed there was an automatic emergency brake event at much higher speeds than most vehicles that are out in the marketplace can actually achieve now. And that’s one of the examples that we showed. We did show Tesla. They’re one of the leaders in the market space right now with vehicle capability. But the difference is, they’re still [unable] to address a lot of the core safety opportunities that lidar technology brings. And that’s really what we were illustrating there with that demonstration.
So break down lidar for us very simply. What is it, and how does it work at a fundamental level?
So you’re probably familiar with cameras — I mean, you’re looking at the result of camera data right now. It’s an image, it’s two-dimensional. You can do all kinds of processing to it to kind of “guess” at how far away things are from each other. But ultimately, it is two-dimensional data.
Radar is kind of the opposite. Radar is two-dimensional data as well. But instead of having lots of information here, you only have information radially. So you don’t have the ability to understand what things are. But you can see how far away there might be something.
What lidar does is, it kind of bridges that gap. It allows you to [have] the detailed understanding that a camera would give you, of the size and structure of objects, but also exactly how far away they are. So lidar is a technology that uses lasers to understand the distance to objects– but at camera-like resolution. And so we do that by timing the speed of light, and all kinds of really cool technological things, but that’s ultimately what it does and how it works.
So speaking of Tesla, Elon Musk is notorious for despising lidar. And Tesla says that they can go it alone with cameras. So talk to us about your thoughts on that. Clearly every lidar manufacturer disagrees with Elon, but help us understand, what are the limitations of camera technology versus lidar?
So camera technology is ultimately limited to two dimensions. Anything that’s extracted with regards to three dimensions, which is what we live in, what the road space is, is inferred by the data. Now, the argument is that if a human with two cameras and a really good computer can drive, then why can’t a car?
There’s two issues with this. The first is, human eyes are very good. And the human brain computer is very, very good as well. And so the equivalent of hundreds of megapixels of camera data is really what we have to bring to bear.
The other thing, and honestly probably more importantly, is that at the end of the day, this whole thing is not about whether it might be possible to drive a car with cameras only.
The real question, and what Luminar has really been focusing towards is, will it be safer if it uses lidar technology?
And that’s really what we believe in, and what we’re starting to demonstrate now. Because ultimately that’s the goal. The goal isn’t to prove some academic case. The goal is to make vehicles more capable than what a human can do.
You know, humans, for as good as we are at driving — and it’s really quite incredible how a good driver paying attention can drive — we’re not always paying full attention. The conditions don’t always lend themselves to good driving. And so bringing new technologies like lidar to make these vehicles more safe is the real goal. It’s not about what can possibly be done with different sensors.
This space is also kind of an acronym soup of different wavelengths used by different companies, different architectures including MEMs-based. So help us understand some of the key differences. And what you see as kind of ‘pros and cons’ of some of the different approaches to the space.
So as you said, there is alphabet soup, there’s a lot of depth… [But] at the end of the day, for most of the world, I would [focus] on what is possible with the sensing technology, and focus less on trying to figure out for yourself who’s got the best tech. Let the markets kind of speak for themselves: who’s got traction? The hardest critics in the world are our customers. They’re the ones who are basing their business models off of the sensors they select. That’s the biggest thing I would say.
But at a high level, wavelength is a really important one. We operate a longer wavelength than most of the landscape –1550 nanometers — which allows us to send a lot more energy into the world safely. Shorter wavelengths are much more hazardous for eye safety concerns, because they’re much closer to visible light. We use much longer wavelengths, further out in the infrared, [which] allows us to measure much further and much more resolution.
Other than that, there’s a whole host of architectural decisions, which we’ve recorded, [and] actually run through. If you go to our website, you can see a kind of a sensory overload review of different architectures. But for now, I think that’s probably a good place to leave it.
Obviously, cost is an issue in this space. And critics will cite the 1550 nanometer wavelength as being prohibitively expensive. So I would love for you to address that and talk about where you see costs going and how close we are, from a price-point perspective, to getting to commercial mass adoption [of lidar].
A lot of work that Luminar has been doing over the last number of years is toward, exactly as you’ve said, to break down the assumption that it’s too expensive to operate these longer wavelengths.
This assumption comes from two sources. One, comes from the fact that you can’t just go to the Internet as a random person and buy a bunch of 1550 nanometer components. And most of the lidar industry are not actually optics experts. They’re trying to buy what they can off the shelf, [and] their electronics experts are pulling these things together and [trying to] build something that is lidar.
Luminar, knowing that we needed to go to these longer wavelengths to serve the market, invested really early in integrating that technology ourselves. And so, building that supply chain, leveraging what existed from telecommunications and from the defense community, to a technology base that is custom-designed for the automotive application space, but stands on the shoulders of these existing markets.
The other challenge is the architecture stuff… if you design the architecture right, you end up minimizing your cost-reliance on the opto-electronic components — the receivers, the detectors, the lasers. Those bits will drive the cost of the sensors. And the fewer of those you can have in a sensor, the better.
Historically, there’s a lot of [these] technologies that cascade a whole bunch of receivers and lasers, and single point-by-point “brute force” the problem. And those have really fundamental cost floors. Luminar’s architecture is a “one-to-one” and we scan that around the scene to capture the whole environment.
So architecturally and supply chain, being able to pull all this stuff together is really what’s allowed us to get to the cost points that we’re driving to market.
And one thing I’ll say is, we’re not [even] being undercut by technologies that [we’re outperforming]. So we’re able to go in with very compelling technological and performance-based proposals that are also very cost competitive, and ultimately enable the business case of our market.
I think we can agree that there’s a tremendous amount of activity in this space. But investors that may take a look at lidar stocks will see that many of these names, including your own, have come under pressure. I think even Luminar stock is down something like 50% year to date, despite having a commercial OEM contract with Volvo, and a lot going on. So talk to us a little bit about the catalysts to getting that relationship to scale. What are some of the milestones for you?
The volatility is partially life as a public company. But I think that, you know, you touch on a really good point: our goal is to drive long term execution.
If you look at the whole space that we’re often compared to, these lidar companies, the first thing is that most of them have jumped to this stage of their business without having a book of business that really requires them to build this much of a runway effectively. We did that well before [we became] a public company, which gives us a lot more strong stability of basis to stand on.
And this pipeline is really quite rich. In these new emerging markets, success begets success. The work that we did with Volvo, we spent a long time with them over a lot of years — way before they even wrote an RFQ, Request for Quote, for our program. And so we do this with all of our customers, or at least as many as we have the opportunity, to spend that early time working with them on their problem, understanding what they need to achieve, all the way through function, so that we can make sure that our technology serves their needs.
Now, the fact that we’ve been doing that for long enough, aligns our roadmap very well with what we can do and what we can deliver now through our ecosystem — to what they’re trying to achieve.
And so while the rest of the marketplace is trying to figure out how to put their sensors together to try to compete with what we’re doing, we’ve already moved two steps beyond that. We’re showing the OEM what’s possible with their vehicles. You know, we’re able to detect a child (dummy, for now) jumping out from behind a hidden cover, behind a car — 25 meters away from a car going 30 miles an hour. And we’re able to identify that, assess it as a threat that’s real, apply the brakes and stop the car.
This is not something that vehicles on the road today can do, because there’s too much uncertainty in their detection.
And so bringing that whole thread of requirements down from, “what do you detect, how much confidence do you need, how much time do you need to detect it, all the way down into how many measurements and points and at what range, what quality of measurement do I need in the sensor?” [That’s] really probably the biggest thing that differentiates us, which is why I urge [investors] to try to look [beyond] the spec sheet battles that exist out there and look [at] what’s being done with the sensors [now].
Lidar is sufficiently complicated and too easy to spin into, into certain configurations that allow you to do certain things. But if you focus on what’s possible, the functions… And that’s really what we’re executing towards, and working that pipeline [for], to make sure we can bring as many folks in the marketplace along with that mission.
Most companies in the space seem to be focusing on Level Two, Level Three autonomous driving, where there’s still driver intervention and attention required. What do you think needs to happen from a technology perspective, to get commercial OEMs to feel comfortable, to meet safety standards, and to truly see that kind of mass commercial adoption that we want?
I think it has significantly less to do with autonomy and what people think about as “getting the human out of the loop,” and it has a lot more to do with what we can do with these technologies to make the vehicle operate more safely, regardless of who’s in control.
We all know, driving our cars… I have a nice luxury car with lots of features. And I still don’t trust it to do most of these things, because I’ve seen it disengage enough. I’ve seen it not react to something that I probably figured that it should have. And it’s that trust, I think that’s really what drives what we call “proactive safety” and using this technology, because of how much confidence it has of knowing the environment at long ranges, without all the false alarms that come along with cameras and radar. That allows us to build a safety envelope from which self-driving can emerge, because that’s ultimately mostly a luxury feature.
There’s of course some safety benefit to be had from a really good, always attentive robot driver. But if we can create that safety envelope with the technology in the first place — which is honestly even “Level Zero”– that doesn’t sound very sexy, but Level Zero are functions that are noncontinuous. So automatic emergency braking is technically Level Zero. And a lot of these things that really will move the needle in the economic benefits of safety — avoiding crashes, not just making them less bad — is what’s going to happen and move the industry forward.
And as soon as we can create that safety envelope, is when we can start confidently deploying Level Three, Level Four technologies. And that’s really the kind of base that we’re driving toward. And, you know, look at Volvo. That’s what’s convinced them to standardize this technology in their vehicles– not the highway pilot that we’re going to be driving towards.
That’s the upside, and I think that’s really more and more the right way to think about it — both as a citizen of the world and as a business. Finding economic motivations to standardize technology is a much more long-lasting movement as a business than hoping for government regulation to say that “now thou must all have lidar” — [because that’s] just going to drive the technology towards low-cost at the expense of performance.
So if we can maintain performance, convince the economic case of safety, and the upside of autonomy in these systems, that’s where we see very exciting, very rapid adoption, across vehicles, across automakers and across the world.
So you touched on this idea of the “safety envelope” as being the largest really addressable market for this technology. But what are your thoughts on the futuristic world of robotaxis and Level Five autonomy? Is it possible? Is it a pipe dream? What technology changes do we need to get there?
So a lot of people are talking about how Luminar is “pivoting from autonomy.” And at the end of the day, it’s really not. Our mission has always been to enable self-driving, to enable autonomy. Along that path is the identification that there’s this need to deliver safety for autonomy, and that it’s tremendously valuable to the consumer market.
And so it’s really something we definitely believe… that robotaxis, robotrucks — all these kinds of things are possible. And we’ve seen our customers do it. And I think that’s really the thing that motivates. So as a business, the real question comes down to where are the markets that are viable and have strong demand now. So we’re serving our technology partners like Mobileye (NASDAQ:MBLY) and Pony.ai doing robotaxis. We spend a lot of attention trying to drive them forward. We look at our partners like Daimler Trucks and trucking and a whole host of others, to drive that market forward.
The group that we’re seemingly focusing most on now with consumer markets, is what we need to do anyway — from a product, cost, scaling perspective — to be able to not be the long pole when those markets are mature. And actually pulling for demand and volume.
And so everything we’re doing now, still aims toward full higher levels of autonomy, because the sensing technology demand is still there. But the level of autonomy has very little to do with the sensor performance as much as vehicle speed does.
And so because we’re looking at delivering safety at all speeds, because we’re looking at doing highway autonomy, that drives the sensor requirements for all of these applications. You just need more of them to cover a larger field-of-view and other sensors for redundancy and things like that, to do higher levels of autonomy. But the sensor itself needs to do the same thing.
So our roadmap to enable all of these great things in consumer vehicles, is what we need to do anyway to make sure that autonomy can be realized down the road.
Are there any other adjacent technologies that need to improve for autonomous driving to see mass adoption?
So I think the real long pole is ultimately going to be the processors. So, the market is anchoring itself for this early wave of technologies in effectively, supercomputers. These are rather expensive, high-power consumption, very performant computer systems that bring all of this data in, process it to understand the environment in high confidence, and then build routes and things.
And that is the point at which will ultimately be the cost-driver and the power-driver for these systems. And so that’s the area that is being developed. But that’s kind of the way this all works. Once we define the systems, the architectures and the compute can come down, because it can be simplified.
One of the cool things about lidar is it’s actually a much lower processing-intensive data stream, because it doesn’t have all the superfluous data that a camera stream has. I don’t need high resolution right here on this wall if it’s a wall. It’s flat, it’s the same color. And so lidar data gives you that more immediately actionable 3D data. But there’s actually less data than a camera, which gives you lots of not very actionable information. So being able to find ways like that to leverage the different sensors to build the processing requirements, gives a really nice roadmap to driving that down as well.
So you obviously have seen a lot of technology. What technologies on the horizon most excite you? What do you think is interesting right now?
Oh, man, that’s a good one. Honestly, I think that the more we do with vehicle function, the more important it’s going to be to understand how the now passenger, or changing between passenger and driver in the vehicles, are going to be interacting with the vehicle. And so this is something that, you know, we look at with our automaker customers, as well as other partners in the ecosystem to see, “how can the system really come and be completed into the vehicle?” We announced last week our partnerships with Webasto and Inalfa to figure out how to get all the sensors really nicely and cleanly integrated into the vehicles.
But we need to think also into the future, how do we integrate the person? So thinking about that, as a system, is really kind of an exciting space. How can we give this data to a human who wants to be driving, but wants the confidence and the assurance that the vehicle system around them knows what’s happening and can give good alerts, more than just lights on your dash. So that space is really kind of cool, as a consumer and as a technologist to see what’s going to come down the pipe to that driver experience when the vehicle knows more than you do.
Your comments and feedback are always welcome. Let’s continue the discussion. Email me at firstname.lastname@example.org.
Disclosure: On the date of publication, Joanna Makris did not have (either directly or indirectly) any positions in the securities mentioned in this article.
Click here to track her top trades, where she sheds light on market psychology and momentum, while leveraging her deep knowledge of fundamental analysis to deliver event-driven trading strategies.