Money For Everythin’

Part 1 of this article discussed automotive LiDAR pricing across suppliers, including automotive Tier 1, ride-sharing, and venture funded LiDAR companies. Valeo, a Tier 1 supplier currently has an automotive-grade LiDAR (SCALA) for ADAS Level 2-3 capabilities. It is priced at $600/unit in high volumes and has modest performance in terms of range and PPS (Points Per Second, analogous to resolution in a camera). Companies like Ouster, Velodyne, Robosense and Cepton are promising similar price points with substantially higher performance levels for future deployments (~3-5 years from now). 

The question is whether the pure-play LiDAR companies can achieve these aggressive price/performance metrics and stay profitable in the future. The business models of these companies are diametrically opposite to Clayton Christensen’s classic study of disruptive innovation which posits that entrenched and well-funded suppliers with an existing product in a given market continue improving at a modest pace. Creative and nimble innovators seed the market with lower-priced, lower performance products. Over time, the innovators improve their product at a faster pace, disrupt the performance/pricing proposition, and “surprise” and overthrow the entrenched players. 

In the case of automotive LiDAR, however, the innovators are promising superior performance now and in the future. The challenge for them is to reduce pricing over time and accelerate performance disruption. It will be interesting to see how this new paradigm plays out, and how entrenched competitors like Valeo respond.

In general, higher performance LiDAR requires more optical and electronic resources, driving up the Bill of Material (BOM) and assembly cost (labor, capital, quality, yield). Efficient system architectures, a higher degree of vertical integration, and intellectual property (IP) that enables complex opto-mechanical-electrical systems like LiDARs to transform into semiconductor chip-level products can reduce or eliminate this cost increase. As LiDAR approaches “chip” like architectures, wafer-level manufacturing efficiencies can deliver substantially higher performance with non-linear price increases.

Table 1 shows performance vs announced pricing of LiDAR systems for Valeo (as a reference) and 4 other suppliers.

As seen in Table 1, the non-Tier 1 LiDAR players have announced similar pricing to Valeo. The $600 price point indicates what automotive OEMs are willing to pay based on their assessment of consumer demand for safety and automation features that LiDAR would enable. Promising superior performance relative to Valeo makes sense since otherwise taking a bet on a non-Tier 1 supplier does not present a compelling case to an automotive OEM. However, OEMs should consider that they may need to pay more for higher performance – and LiDAR companies should continue to push for this, otherwise business models may get untenable. LiDAR is a game changing technology and provides the ability for automotive OEMs, trucking and TaaS (Transportation as a Service) businesses to provide premium safety, efficiency and autonomy features. 

The discussion below describes why four of the non-Tier 1 LiDAR companies interviewed for this story believe they can achieve this type of price/performance disruption.

Ouster’s ES2 LiDAR is expected to launch in 2022 with serial production in 2024. It operates at 880 nm with VCSELs (Vertical Cavity Surface Emitting Laser) and SPADs (Single Photon Avalanche Detector). There are no moving parts to scan the Field of View (FOV), instead, it uses electronic scanning. Per Angus Pacala (CEO of Ouster), the highly digital nature of the ES2 enables semiconductor manufacturing efficiencies. Component count is low, the assembly is efficient and the silicon SPAD and GaAs VCSEL array can be driven to low costs through innovative wafer design, zero moving parts, and efficient system design.  The ES2 uses an architecture similar to the Apple iPhone LiDAR (VCSELs and SPADs. Smartphone volumes will mature the manufacturing base of these key components, something the company is positioned to leverage. Ouster controls > 50% of the BOM cost of the ES2 through IP ownership of critical components – this preserves internal margins. These factors will allow the ES2 to deliver high levels of performance at the $600 price point and maintain, at a minimum, the 25-30% gross margins typical of the automotive business. Since Ouster also serves the higher-margin non-automotive markets and deploys essentially similar LiDAR architectures, the overall business gross margins are expected to be higher.

Velodyne: the Velaray H800 operates at 8XX nm and advertises itself as a solid-state LiDAR. It uses GaAs EELs (Edge Emitting Lasers) and silicon APD (Avalanche Photodiode) arrays, with integrated drivers and readout circuits. Although not publicly disclosed, the Velaray presumably has some type of scanning (either a MEMs or a Voice Coil mirror) that flashes the VFOV and scans the HFOV (for the kinds of PPS performance advertised, this would need at least 64 element EELs and APD arrays). Per Anand Gopalan (CEO of Velodyne), the Velaray has been in development for the past 5 years – during which time, significant internal investments were made to integrate electronic functions into an ASIC (Application Specific Integrated Circuit). The packaging of active and passive optical elements (lasers, detectors, lenses, filters) was automated to build MLAs (or Micro-LiDAR Arrays) which leverage silicon photonics packaging innovations from the optical communications industry. This has allowed the Velaray to achieve low BOM costs and essentially zero-touch labor. Additionally, Velodyne’s model of being capital-light allows them to rely on contract manufacturers and Tier 1 suppliers for cost-effective manufacturing, typically in geographical regions with competitive labor and facility costs. 

Per their IPO filings, Velodyne expects to operate at a 45% gross margin, and > 20% EBITDA in 2025 when these ASPs are realized. This projected margin is across multiple products, software licenses, and different markets. The specific margins for the Velaray and the automotive business are not provided but expected to be a lot lower.

Robosense has launched prototypes of the RS-M1 which it advertises as a solid-state LiDAR. Multiple customers have deployed this LiDAR and production is planned for 2022. Dr. LeiLei Shinora (Vice President of Product Development) explains that the architecture uses a total of five low cost 905 nm EELs, scanned at 15 kHz by Robosense’s proprietary 2D MEMs design. The low-cost architecture relies on minimizing the laser cost and maximizing the MEMs scanning performance – something that Robosense believes is their key IP and core competence. Other IP includes the design of the optical module packaging which minimizes the number of piece-parts on the BOM (< 20) and enables highly efficient assembly and calibration processes (< 5 minutes). Finally, the RS-M1 embeds perception and point cloud processing algorithms that eliminate the need for high processing power (and expensive FPGAs and GPUs). 

Cepton’s Vista-X90 uses a proprietary Micro Motion Technology (MMT) technique to accomplish a high speed, mirrorless 2D scan of the FoV. An earlier incarnation of MMT was used by Dr. Jun Pei (CEO of Cepton) in an optical instrumentation venture. In essence, it uses a single voice coil mechanism to create frictionless and synchronized scanning of an 8XX nm wavelength EEL and APD receiver. Dr. T. R. Ramachandran, Chief Marketing Officer at Cepton indicates that MMT is simple to implement and requires a small number of components that have already been in use for automotive applications. Additionally, key functions have been integrated into ASICs, and proprietary packaging optical array for lasers and detectors reduce BOM costs significantly. The design simplicity enables delivery of high-performance at aggressive price points, which Dr. Ramachandran says is acknowledged by multiple OEM customers. Cepton’s relationship with Tier 1 partners like Koito provides access to a global automotive supply base with substantial volume cost scalability. The compact size allows for low cost and style compatible integration into headlamps, behind windshields, and within vehicle bumpers.   Future technology innovations (currently in progress) will enable higher point density performance (PPS) at much lower optical component costs. 

Analysis and Takeaways

In terms of performance, the 200 m range at 10% reflectivity (and within Class 1 eye-safety limits and the high PPS performance) still needs to be publicly demonstrated at these wavelengths. Arguments for being able to deliver at the low price are essentially generic at this point (not a surprise) – manufacturable designs, minimizing the number of components, conversion of electronic functions into ASICs, and efficient optical packaging. Telecom providers made similar arguments two decades ago as the era of optical communication networks became a reality. Their experience has shown it is difficult to “siliconize” optical manufacturing, an the problem is even more complex for free space optics deployments in public spaces.

Competing to become an automotive supplier is a grueling proposition for a venture capital (VC) funded company.  Design-in cycles are long, gross margins are in the 30% range (low by VC standards), risks are enormous and there is a constant pressure to reduce pricing. A sale price of $600 implies a manufacturing cost of about $400 – this includes about $50-100/unit to be paid to contract manufacturers or Tier 1’s for the assembly and calibration consistent with automotive quality standards (a Tier 1 LiDAR company like Valeo absorbs this internally). The balance needs to pay for all the components – lasers, detectors, scanners, electronics, computing. A difficult endeavor unless the company has significant funding and is vertically integrated. Entrenched suppliers like Valeo and other Tier 1’s will continue improving performance (although likely at higher prices). OEMs will continue to push suppliers to lower pricing (as they do for their traditional supplies like radar, cameras, metal and rubber).

LiDAR companies should not give performance away for Nothin’ in a race to the bottom. Stay strong and negotiate hard because you are providing Somethin’!