Money For Somethin?

The perennially famous Dire Straits’ song ‘Money for Nothin’ – would be great if true in real life. Unfortunately, performance rarely comes free.

The LiDAR business is becoming real.  The recent buzz of corporate events among LUnicorns is creating increased focus on revenue, pricing and profits.  Amidst the clamor of announcements of $100-$1000 LiDAR products, the pertinent questions to answer are: “What is a LiDAR price point buying in terms of performance? Are future price projections realistic from an implementation and financial perspective?”

There are many ways of designing a LiDAR, and > 75 companies are pursuing  variations. To first order, LiDAR performance in an automotive environment can be decomposed into:

1)   R or 3D Range: distance over which a LiDAR can reliably provide 3D information

2)  PPS or Points per Second: how many points can 3D information be provided for per second. PPS in a LiDAR is similar to the number of pixels in a 2D camera, but also includes depth information (sometimes called voxels).

R and PPS are critical for sensing and controlling the car in an ADAS or AV environment. R depends primarily on laser power and detector sensitivity – in effect, the overall system sensitivity or SNR (signal to noise ratio). PPS is a function of how the view in front of the LiDAR is addressed in terms of angular resolution, field of view (FOV) and frame rate. This performance parameter is driven by the laser, detector, scanning mechanism, and the signal processing platform.

The costs and prices of different LiDAR designs is a function of:

1)    Costs of the key components that drive R and PPS – laser, detector, scanner, and processing electronics

2)  Assembly, testing, and certification costs (typically involve labor, supply chain management, facilities, equipment, tooling, quality, engineering)

3) Pricing power based on other factors other than performance – smarts, size, integration, reliability, first-mover advantage, etc.

Performance Comparisons

Table 1 shows a comparison of performance promised by 7 different LiDAR companies. The performance parameters (R and PPS) are available through their websites/publications or estimated from other public information.

Valeo, a Tier 1 automotive supplier is currently the only LiDAR manufacturer in the automobile space with an automotive-qualified LiDAR (SCALA®). The product is designed into four leading automotive OEM platforms including Audi. The company currently has orders for ~$600M at a projected price point of ~$600 (1M unit volumes). The SCALA® delivers a range of 100m (R) and 60,000 points/second (PPS).  Valeo’s SCALA® is used as a benchmark to compare performance and pricing across other LiDAR suppliers.

How Should Pricing Scale with Performance?

Range performance (R) is driven by the laser power and detector sensitivity. Using Ouster’s pricing (Table 2), the price premium for higher R can be estimated. 

To estimate how the market views pricing as a function of Points/second (PPS), it makes sense to consider other similar optical and semiconductor technologies.  For passive infrared cameras, a four-fold increase in the number of pixels (going from a VGA format or 640×512 pixels to an HD format or 1280×1024 pixels) increases the pricing by a factor of ~ 2. Similarly, for fiber optics communications, an increase in data transmission speeds from 2.5 Gb/s to 10 Gb/s, or 10 Gb/s to 40 Gb/s drove a price increase by a factor of ~2. On one hand, customers will never pay 4X for a 4X increase in performance (design change costs are too high). Convincing customers to switch requires a sub-linear price/performance dependence. This is also rational from a cost perspective since semiconductor processing costs scale similarly (a 2X increase in wafer diameter increases cost by 2X and produces 4X the number of pixels). Additionally, optical packaging costs which form a significant portion of the total costs reduce on a pro-rata basis (need one package instead of four).

Based on the above, a price premium factor for R and PPS can be derived and used to project pricing for different performance levels.

Comparing Projected Versus Announced Pricing

Figure 1 compares projected pricing based on R & PPS across different LiDAR supplier products (at high volumes) versus announced pricing.  P in Figure 1 is the composite price premium for the performance provided, and is used to compute the projected pricing. Valeo is used as a benchmark to scale projected pricing as a function of performance.

Continental, Waymo, and Ouster have not published high volume pricing. Figure 1 indicates the projected pricing they would need to be at for the performance delivered.

Robosense’s pricing is based on CES 2020 press releases.  Their announced price is ~2.5X lower than the projected price.

Velodyne and Luminar announced pricing is estimated based on public filings related to plans to go public. The announced prices are ~ 4 to 5X lower than projections.

Analysis

Price variations between LiDAR manufacturers is a function of:

1)      System performance specifications: higher R and PPS should drive higher cost and price. Similarly, higher software content should enable higher pricing power.

2)     LiDAR system design: this is a balance between how photons (laser power and detector sensitivity), atoms (fixed and scanning optics, mechanics), and electrons (signals, computing) are used to achieve the desired R and PPS performance. Design choices can include 2D flash LiDAR, 1D flash LiDAR, and 2D scanning LiDAR. It also includes other choices like the operating wavelength (9XX nm vs 15XX nm) and LiDAR physics (Time of Flight or ToF vs Frequency Modulated Continuous Wave or FMCW). Companies try hard to excel at creating IP and products that optimize performance and price.

3)     Bill of Material (BOM) costs: 2D flash LiDARs tend to increase the cost of optical semiconductors but eliminate costs of scanning. 2D scanning LiDAR reduces optical semiconductor costs and increases scanning costs. 1D flash LiDAR is in between. Processing electronics plays a big role – using expensive computing stacks lead to high costs, in many cases eclipsing the cost of the optical components. Higher operating wavelength (15XX nm vs 9XX nm) significantly increases the cost of the laser and detector semiconductors, although they also enable higher R and PPS performance because of higher eye safety thresholds.

One strategy for reducing BOM costs is to pursue vertical integration and preserving margins. Some LiDAR companies pursue this model (but it requires the right talent and higher R&D investments). 

4)     Manufacturability, Yield, Quality: this is all the yeoman stuff that does not typically garner a lot of press and buzz – creating efficient manufacturing infrastructure and processes to scale volume, minimize scrap, increase throughput, manage supply chains, implement quality systems, and guarantee contractual business agreements with automotive OEMs.  Non-Tier 1 LiDAR companies do not have the experience and relationships to excel in this area, and would need to rely on contract manufacturers or Tier 1 suppliers to provide this as a service. Of course, this would increase their costs since margins are being shared externally. Automotive Tier 1 LiDAR suppliers are masters at this, and best positioned to leverage this as a competitive tool.

Takeaway

The substantially lower prices proposed by Velodyne, Luminar, and Robosense (for the performance they are promising) in the future can be justified if the system architecture is disruptive enough to drive lower BOM costs, and there is more vertical integration so that margin dilution is prevented. 

Velodyne has indicated that they have significant IP in wafer scale LiDAR, micro-LiDAR arrays, ASIC design and process automation that can provide manufacturing efficiencies and reduction in BOM costs. However, they also indicate a capital light model and the use of external manufacturing partners which will cause margin dilution. Luminar has internally developed low cost detector solution, which helps with BOM costs, although operating at the 1550 nm wavelengths will increase laser and detector costs.

Continental and Ouster are more vertically integrated in terms of internal control of the laser, detector and some optical elements. The situation at Robosense and Waymo is not clear in this regard.

As a business proposition, non-Tier 1 LiDAR players funded by venture capital investments have to promise higher performance than Tier 1 suppliers. Otherwise, they would find it difficult to attract funding. However, they need to benchmark price/performance propositions against established players. Relying on volume scaling and production efficiencies as an argument to project future price/performance disruption relative to an automotive Tier 1 supplier is not convincing. Design and process IP, vertical integration and innovative system design as disruptive forces are reasonable arguments – as long as they can be justified to customers and investors.