Automotive software has a complicated journey to a typical car’s electronics systems. Even to many auto industry observers, the auto software cost is hard to measure and it is opaque at best. This is largely because it covers the variety of software components and platforms and systems, which are also complex and diverse. These features also make it difficult to estimate the costs and values of software in the auto industry. I will offer some perspectives and ranges of estimates of the auto software cost and value flow.
The first step is to get a picture of how the many software components flow through automotive supply chain, which is shown in the figure below. In the middle of the figure are four levels of software flow — from processors to modules and systems and eventually to ECU networks.
At each level, the supplier has to develop software as part of the software-based product that is sold to the next level. This effort is usually paid for by a software royalty as indicated on the right side. Each block also has an estimate of the software R&D share of total electronics system R&D. The first R&D percentage is for the 2020 timeframe with the second R&D percentage an estimate for 2030 timeframe.
The upward lines between the software flow level boxes have an estimate of the software value as percentage of the system cost that the buyer pays to the supplier. There is a range because the software content varies greatly between the systems that flow up the chain.
The left side in blue shows the automotive supply chain from processor suppliers through Tier 2-3, to Tier 1s to OEMs with high-tech included. The left side also include perspective on software levels at each step in the supply chain.
The right side in green gives some perspectives on software types added at each step in the software flow. All of this have more details below the figure.
The processor level includes software that are included with MCUs, GPUs, MCU-based controllers for sensors and other electronics plus the emerging AI-specific processor chips. The growth of system-on-chips (SoCs), system in package (SIP) will accelerate the amount of software at the processor chip level.
At this level, it is mostly high-tech companies supplying the processors and the special purpose processors for sensor control. The Tier 1 companies are starting to design their own processor chips for sensor control and OEMs will also do so. A few OEMs are expected to develop their own processor chips for AVs, with Tesla leading the way.
The software at this level currently tends to be focused on hardware control such as software drivers for CAN, Ethernet and other electronic buses. Special control chips with software drivers for sensors such as cameras, radars, and lidars are growing in importance. Middleware software components are also included along with bootloaders and similar code to get the processors started. Embedded cybersecurity software along with cybersecurity hardware will see increasing usage.
The software R&D cost to create these products is estimated to be in the range of 15% in 2020 and growing to 35% in 2030 as a share of total system R&D budget.
The value of the software in the products that go to the next level is estimated at 3% to 8% of the price paid by the customer. The range is due to the large software variety flowing to the next level. It is likely that the percentage will grow as the processors have increasing memory with more built-in software.
Modules to ECU Level
At this level the processors are becoming sub-systems, systems and modules to create ECUs. Much more electronics are added and multiple sub-systems with multiple processors may be included. Communication capabilities are often part of the sub-systems to modules.
A variety of suppliers compete at this level ranging from Tier 3s and Tier 2s to high-tech companies. System and software integration to the module level are often included.
The software at this level start with more middleware and software drivers for hardware sub-systems and additional sensors and actuators. Some level of operating system such as AUTOSAR is often included. Application software for specific functionality is also added. All of these software components may become ECU software platforms.
The software R&D cost to create these modules or systems is estimated to be in the range of 10% in 2020 and increasing to 30% of total system R&D effort.
The value of the software in the products that go to the next level is estimated at 8% to 15% of the price paid by the system customer. The large percent range is due to software variety by product category going to the next level. The percentage is likely to increase as the modules have increasing memory with more added software functionality.
This level includes a lot of ECUs and many MCU-based systems that are managed by the ECUs. There is a lot of ECU system and processor integration taking place. Software integration is another major activity as more software is added and the modules are becoming connected systems.
This is primarily a Tier 1 supplier domain and mostly led by the leading Tier 1 companies. A few high-tech companies may compete with likely growth as AV systems go into volume production.
The software that is added at the system level is comprehensive with major applications such as infotainment, telematics, navigation and ADAS. They all need advanced operating systems, application software platforms with communication capabilities for cloud functionality.
The software R&D cost to create these products is estimated to be in the 20% range in 2020 with growth to 40% in 2030 as a share of total system R&D cost.
The value of the software in the products that go to the next level is estimated at 12% to 25% of the price paid by the OEM customer. The range is due to the large software variety flowing to the next level. It is likely that the percentage will increase as the ECUs will have increasing capabilities with more software platforms for client and cloud activities.
ECU Network Level
This level includes all ECUs as a system and connected via a network using multiple electronics buses such as Ethernet, CAN, LIN and others. Domain ECUs are emerging at this level and creates new software platform to manage the resulting system changes.
This is the auto OEM’s territory with help from some Tier 1 suppliers. The main tasks are system and software integration at multiple levels. The emergence of domain ECUs will help, but there are probably a lot to be learned and experience to be gained before the domain ECU advantages are fully realized.
The software is now quite extensive as all the software listed earlier have flowed to the top. All the ECU apps platforms are connected and communicating via electronic bus networks. Domain ECU software platform will become prevalent along with Ethernet-based networks in the next decade.
The operating systems for most ECUs and domain ECUs will be quite complex with few products meeting the required capabilities. OTA and cybersecurity software platforms are becoming required capabilities.
The software R&D cost to develop and integrate all these products is estimated to be in the 15% range in 2020 and grow to 35% in 2030 of total electronic system R&D effort at this level.
Car Buyer Level
The software value included in the car buyers’ price is not included in the above figure. The next table has some estimates for 2020 and forecasts for 2030. The estimates have large ranges due to a great variety of electronics per car sold. The 2030 forecast is speculative, but the trend is clear—more software value per car will continue for a long time.
The bill-of-material (BoM) is the software royalties that are paid for various software used in ECUs. In 2020 it is estimated that software royalties for an average car range from $150 to $300 and will grow to $250 to $500 in 2030.
The software development cost allocated to an average car is projected to grow from $500-$900 in 2020 to $900-$1,500 in 2030. This includes the software value estimates flowing through the software supply chain.
In the car ownership phase or the car use-phase the auto OEMs still have software costs for recalls and software maintenance or bug fixes. These cost estimates are $150-$450 in 2020 and growing to $250-$650 in 2030. The increasing use of OTA is probably slowing the growth of this segment.
The total cost range for these three categories is $800-$1,650 in 2020, which is forecasted to grow to $1,400-$2,650 in 2030. This is the OEM’s estimated life-time software cost per average car in the U.S.
I also like to estimate what I call the software free-ride-on-hardware or the cost of hardware needed to run the software. For 2020 I have an estimated range of $4,000-$9,000 for the electronics hardware in car segments. The electronics hardware costs are expected to grow to $7,000-$15,000 in 2030. The future addition of ADAS and AV technology is a key reason for the large growth by 2030.
The good news for OEMs is that the car use-phase will bring new software opportunities to potentially match or exceed the software costs. I have made a guesstimate of the potential life-time software revenue range for cars sold in the U.S, in 2030. The range is $2,000 to $5,000.
The post Projections for Rising Auto Software Cost for Carmakers appeared first on EETimes.