Tesla's Trendsetting Semiconductor Move and Why You Should Care
May 11, 2023 Dr James Jeffs
Microcontrollers (also known as controllers, MCUs and other name variations) are the backbone on which modern vehicles are built. Regardless of whether a car is powered by petrol, diesel, electric, hydrogen, PDG, or anything else, its functionalities will be heavily dependent on controllers. It has also been seen how the automotive industry can be bought to its knees when those controllers are in short supply. IDTechEx's report, "Semiconductors for Autonomous and Electric Vehicles 2023-2033", finds that there is currently lots of movement within automotive controllers, offering more computational power and new vehicle architecture options. One company that has been bucking the trend with controllers is Tesla, and here is how.
Semiconductor wafer value distribution in a Tesla-like electric vehicle with level 2 ADAS. Source: IDTechEx
Tesla started designing controllers as far back as the Model S, but only 20% of the Model S's controllers were designed in-house. This has risen through the years, with the Model Y being 61% internally designed controllers and the Cybertruck going up to 85%. Tesla's 2023 Investor Day presentation stated that the next-gen vehicle will have 100% internally designed controllers. One of the benefits of bringing controller design completely in-house is that it gives Tesla total autonomy over the design of its wiring harness. Here is why that's a bigger deal than maybe first thought.
Tesla is leading the industry in wiring harness reduction. With the amount of electronics in modern cars, it's no surprise that they contain a near-infinite maze of wiring so every component can communicate with each other. However, big wiring harnesses cause a number of issues, firstly, the weight. Even though cars are only a few meters long, when having to make that run a few hundred times for different wiring, it soon adds up, and modern vehicles can have harnesses with literally miles of total length. Five miles of insulated copper cable is heavy! IDTechEx's research found that a wiring harness can weigh more than 60kg, making Tesla's 17kg weight saving impressive.
Another reason that a large wiring harness is not desirable is the labor costs that it takes to build one. A single wiring harness has hundreds, sometimes thousands, of terminations. At each one, the wire needs to be cut, stripped, crimped, and fitted into the correct pin in the connector, and this is done by hand! These are fiddly tasks that require high levels of dexterity, attention to detail, experience, and skill, so it is not something that is likely to be replicable by machines any time soon.
Bringing controller design in-house allows Tesla to move from a centralized control architecture to a localized one. In a centralized architecture, one, or a small group of controllers, needs to communicate with devices all around the vehicle. So say, for instance, there are a group of five devices at the opposite end of the vehicle, requiring maybe a 4m cable run. As an illustration, let's say each device will have at least four connections, one for power, one for ground, and two for CAN; some devices have fewer, and some have a lot more. So now there are 20 wires and 80m of cable, just for one small group of devices. A localized architecture means a smaller controller is acting as a hub in the middle of these devices. It has one ethernet connection to the central controllers of the vehicle and only very small runs to the individual devices. So now that 80m of cable might be reduced to, say, 40m for example.
In addition to changing the controller architecture, Tesla is transitioning to a 48V architecture for the low voltage devices, bringing further benefits. The 48V system or low voltage system is used to power components such as sensors, lights, infotainment, pretty much everything electronic except the drivetrain. Most vehicles use a 12V system powered by that lead-acid lump underneath the bonnet, a technology that is over 150 years old now. According to Tesla, a 12V system, with all the auxiliary loads around the vehicle, needs to be capable of supplying over 200 amps on modern vehicles. But since power is voltage multiplied by current, the other can be quartered if one is quadrupled. This means Tesla's 48V system only needs one-quarter of that current around the vehicle, which lets them use much, much thinner cables.
Tesla have built a reputation for relentlessly optimizing every facet of vehicle design and production. IDTechEx thinks these moves to completely internally designed controllers, and the adoption of a 48V low voltage system will mark the start of even more impressive accomplishments in vehicle optimization from Tesla.
Not only does complete control of the controllers throughout the vehicle give Tesla more freedom to refine its architecture it also gives them more bargaining power over the supply chain. The automotive industry was given a rude awakening to its sensitivity to semiconductor supply during the COVID pandemic. Many production lines came to a grinding halt when vehicle controllers failed to turn up. By bringing controller design in-house, Tesla is less susceptible to these disturbances; it can quickly re-design chips to improve supply or look for capacity with other semiconductor foundries to keep the flow going. Over the past couple of years, other OEMs have made noise about also bringing some semiconductor chip design in-house, but Tesla is the trailblazer by raising the bar to 100% in-house design.
Tesla previously demonstrated this strength when it moved from hardware version 2.5 (HW2.5) to HW3. In HW2.5, the central autonomous brain contained four chips, two Nvidia Parker SoCs (system on chip), one Nvidia Pascal GPU (graphical processing unit), and one Infineon MCU (microcontroller unit). HW 3, on the other hand, only had two chips, both Tesla SoCs and both fabricated by Samsung using its 14nm process. This gives Tesla more bargaining power as TSMC and GlobalFoundries also have 14nm processes, meaning more options and more potential supply. It also shortens the supply chain as Nvidia is fabless, meaning it designs the chips but then needs to get them made at one of the big foundries, the same as Tesla; this will likely result in cost savings for Tesla.
Tesla also says that bringing more design in-house lets them specify features and functionalities beyond what the tier 2 suppliers are offering. When OEMs buy off-the-shelf products, it will no doubt come with compromise. A chip architecture from a tier 2 will be intended to be as marketable as possible and meet the needs of many. That will normally mean that it is not tailored to the needs of an individual. Bringing the design process in-house is not trivial, but now that they have, Tesla can build chips designed only for Teslas. IDTechEx's "Semiconductors for Autonomous and Electric Vehicles 2023-2033" report details some of the existing semiconductor supply chains and explains which other OEMs are investigating in-house controller designs and why.
Although this article focuses on microcontrollers, "Semiconductors for Autonomous and Electric Vehicles 2023-2033" gives a holistic and comprehensive coverage of semiconductors throughout the car, including ADAS, autonomy, LiDAR, radar, cameras, 4G connectivity, 5G connectivity, electric powertrains, MCUs, SOCs and more. IDTechEx can help businesses understand all the new technologies coming to vehicles, the technologies on the horizon, and how the evolving automotive industry will impact semiconductor markets. For more information and downloadable sample pages, please visit www.IDTechEx.com/AutoSemi.
This research forms part of the broader mobility research portfolio from IDTechEx, who track the adoption of autonomy, electric vehicles, Semiconductors for Autonomous and Electric Cars, battery trends, and demand across land, sea and air, helping to navigate whatever may be ahead. Find out more at www.IDTechEx.com/Research/EV.