High-speed data movement will be required in future vehicles, probably including optical, but challenges persist.
Key Takeaways:
The massive amount of data moving through a modern vehicle has far outstripped the capabilities of the traditional CAN bus. In its place, Automotive Ethernet is gaining ground as the logical choice for moving data between processors and memories.
Automotive Ethernet has a number of advantages. It’s faster and lighter than copper wire, well-tested under a variety of ambient conditions, and it’s already a standard. It’s also available in a wide range of speeds, with 10BASE-T1S, running at 10 Mbit/s, the likely candidate to replace the CAN bus.
“What we see a lot at the moment is the low speed for Automotive Ethernet,” said Jon Ames, principal product manager for the Synopsys Ethernet IP portfolio. “BASE-T1S is sitting there in the CAN space. Some of these automotive technologies go to higher rates, many gigabits per second, but it is at this low rate where we see the most activity, and this is at the edge. Zonal controllers throw out, essentially, a multi-hop cable. It’s a single twisted pair that can go to multiple endpoints, which can be simple things that are being switched, or there could be a thing that you are enabling and disabling. There’s not much traffic on those things. It’s all about simplifying the wiring system, having a multi-drop bus, and using a technology that can then go through a switch up to within the car’s network, so you could serve a central controller that talks out to edge, zonal areas — all the switches that go out to a multi-hop with 10BASE-T1S.”
Some OEMs may choose to use 10BASE-T1 in certain places, while keeping CAN or LIN in others due to the lower costs. “Integrating different types of Ethernet, such as 10BASE-T1S, with established standards is technically feasible but complex,” said Seung-Taek Chang, SDV solution manager in the Automotive & Energy Group at Keysight EDA.
Current automotive Ethernet has its challenges, though. Among those flagged by Keysight:
The Open Alliance’s Automotive Ethernet Specifications details various speeds and requirements. While 10 Mbps was considered high speed five years ago, the goal posts are always shifting.
SDVs and 10 Gbps
Higher speeds are not needed for all automotive domains and applications. However, they will be required in future vehicles with full entertainment options, probably at gigabit/second speeds in the near future. And in fully autonomous, software-defined vehicles with 6G technology, those requirements could rise to terabits/second.
Without Automotive Ethernet, it would be difficult to deliver on the promise of software-defined vehicles (SDVs), including advanced driver assistance systems (ADAS) and over-the-air updates.
“OEMs want SDVs because it’s a single platform that can handle all the makes and the models,” said Mike Yeager, senior vice president and general manager of Ethernet solutions, ATV Division at Infineon Technologies. “More importantly, on the technical side, it reduces the cables and reduces the weight of the car, and makes it economically feasible to have one platform that can serve the entire car with an SDV. There’s been a fundamental shift here in the architectures, and there are three major pillars that have occurred: safe and secure computing, a high-speed in-vehicle network (Automotive Ethernet), and intelligent power distribution.”
SDVs currently represent about 5% of all vehicles sold, but they are expected to account for 50% in 2030, Yeager noted. “Automotive 10 gigabits/second (Gbps) Ethernet was transformative. Before, we had all the different varieties of conductivity. Now, the bandwidth of Automotive Ethernet, which allows us to drive across a 15-meter cable with bi-directional wire, allows us to expand the network globally.”
25 Gbps and faster
25 Gbps speeds currently are not as common as 10 Mbps 10BASE-T1S, but that is likely to change in the not-too-distant future. “We do see that designed in, but less often at the moment, because is high speed really going to be there in the car?” said Ames. “I hear that is coming, and has been coming for quite a while.”
Increased video will be one reason OEMs push to higher speeds. “In terms of bandwidth, if you want to take uncompressed video from a camera, that might be some number of gigabits. And then, if you’ve got multiple cameras in the car, you can get to tens or hundreds of gigabits, so you can start to reach these kinds of bandwidths,” Ames noted.
IEEE 802.3cy specifies a 25 Gbps PHY for automotive applications, but higher speeds will be needed for camera, sensor, video, and display links for sensor aggregation. “Automotive Ethernet is adopting features from enterprise Ethernet, such as MACsec and TSN, and scaling toward 25 to 100 Gbps PHYs, alongside PCIe for backbone networking,” said William Chen, product marketing group director for design IP at Cadence.
High-speed Automotive Ethernet also will help realize the promise of fully autonomous (L4/L5) vehicles. “Future standards will likely address the integration of Ethernet with emerging protocols for diagnostics, V2X, and over-the-air updates,” said Adiel Bahrouch, director of business development for silicon IP at Rambus. “The convergence of TSN, MACsec, and high-speed PHYs will create a robust framework for next-generation vehicles. As Ethernet becomes the common language of automotive networking, standardization will drive innovation and scalability.”
Some see the future arriving sooner than expected. “I would estimate about 90% of Ethernet in a vehicle today is between a gateway and central compute,” said David Fritz, vice president of hybrid-physical and virtual systems, automotive and mil-aero at Siemens EDA. “This would be a zonal architecture type arrangement, where we’ve got high speed Ethernet between, say, the zones and the cores of the vehicle, and then you’ve got the central compute, which is getting closer and closer to HPC capabilities. This whole terabit thing for HPC absolutely is going into cars and airplanes in the not-too-distant future, once the actuators and the sensors catch up. Let’s say you’re a Sony and you’re selling cameras. You’re already worrying about your supply chain, not for next year’s model, but for five model years down the road. No matter what Sony decides to do, they’re going to have to support the CAN-based interfaces and many others that they support for at least the next five to seven years. It’s a business conundrum that really impacts the adoption rate. It isn’t the technology.”
Optical Ethernet
Another challenger to traditional protocols is optical Ethernet, which offers several advantages over copper in automotive applications.
“These include higher bandwidth, lower weight, EMI immunity, thermal efficiency, and longer reach,” said Keysight’s Chang. “Optical links can support 25 Gbps and beyond, making them ideal for ADAS, infotainment, and sensor fusion. Fiber optics is also lighter than traditional copper cables, which helps improve electric vehicle range and overall fuel efficiency. Additionally, optical cables are immune to electromagnetic interference, ensuring reliable performance in electrically noisy automotive environments. They’re more thermally efficient as well, with optical PHYs consuming less power and generating less heat, easing thermal design in ECUs. And optical links can maintain signal integrity over longer distances without needing equalization or amplification.”
Optical Ethernet is likely to be popular among new vehicles with zonal architectures and centralized computing. “Retrofitting older models is unlikely due to legacy ECUs and wiring harnesses not being compatible, the cost and complexity of integrating new optical PHYs and connectors, and OEMs are focusing R&D on future-proofing next-gen EVs and autonomous platforms,” said Chang.
Future autonomous vehicles with massive sensor fusion and V2X data exchange might eventually require link speeds faster than 100 Gbps, and these could be in optical form, Chang noted.
SerDes and asymmetrical Ethernet
While CAN and LIN may eventually lose ground to Automotive Ethernet, SerDes remains essential as it converts parallel data to serial for transmission and back.
Ethernet and SerDes are both foundational for modern interconnect technology, but their topologies and design tradeoffs create important distinctions. “Point-to-point SerDes technologies excel in localized, high-bandwidth connectivity,” said Cadence’s Chen. “Newer automotive standards like ASA Motion Link 2.0, which incorporate asymmetric Ethernet communications, may blur the lines and increase interoperability over time.”
At the same time, asymmetrical Ethernet doesn’t eliminate the need for SerDes, such as GMSL or FPD-Link. “SerDes is still optimal for camera and display links due to low latency and high reliability,” said Chen. “Asymmetrical Ethernet, such as ASA-MLE, is emerging to standardize what SerDes does today, offering Ethernet-based alternatives with asymmetric bandwidth of, say, 10 Gbps down, 100 Mbps up. While the transition is ongoing, SerDes remains important for now.”
One advantage of asymmetric Ethernet is reduced power. “A lot of power is needed, usually, if we want to run the same speeds upstream and downstream,” said Benjamin Tan, senior applications engineer at Infineon Technologies. “For asymmetric, what we’re doing is 10 gigabit speeds downstream, so now that video camera data is going to your CPU processor, and we’re only doing 100 megabits per second upstream.”
Developers are moving in this direction because of the potential for lower costs, die sizes, and power. “With this camera, we’re running low power over coax,” said Tan. “Power over coax can be considered power over Ethernet, too, depending on how you interpret it. This is over a 15-meter cable, which is industry standard, and that can enable all this processing of upstream and downstream data.”
With 10BASE-T1s, the 1 is the number of differential pairs. “For automotive, we always want less copper, so we use one single differential pair, whereas other industries use T4 or TX, depending on how much data they need down the path,” Tan explained. “Right now, we’re still at the limit of 10 Gb. That’s mostly for the camera stream. If you want to do a 4K 60 FPS camera, that’s already 10 Gb, and we usually only need one cable for that.”
Data center similarities, differences
As auto networks speed up, vehicles begin to more closely resemble data centers.
“The data center and automotive sectors are seeing ongoing technology cross-pollination, particularly around single-pair Ethernet (SPE) and SerDes design techniques,” said Chen. “The convergence of chiplet-based SoCs and advanced process nodes, 3nm and below, is accelerating integration between automotive and data center technologies, especially as AI-driven architectures become more prevalent.”
Auto chiplets may also mean more UCIe is designed into vehicles. “If you look at the car architecture that exists today, you probably wouldn’t need to have massive dies or, because a big die is expensive, chiplets,” said Synopys’ Ames. “But if you’re looking at where things are going with the kind of compute power in a car with autonomous driving, then you might start to see chiplets, and therefore UCIe become technology, just because once you go to high compute and you need big silicon, big die, then chiplets make sense.”
Looking ahead, the convergence of edge computing, industrial, operations technology, and composable architectures such as chiplets may create opportunities for automotive-style Ethernet or SerDes, especially where low-cost, lightweight cabling is advantageous, Chen said. “In core hyperscale data centers, SerDes will likely remain optimized for ultra-high speed, minimal jitter, and power efficiency. The greater influence may come from SerDes innovations — such as equalization and error resilience — feeding back into general high-speed SerDes design, rather than direct adoption of automotive PHYs.”
Compared to automotive, data center SerDes already enables the highest Ethernet speeds, including 100G, 200G, 400G, and moving toward 800G and 1.6T, as specified by standards such as IEEE 802.3ck. Reliability expectations also differ. “Data centers demand extremely low bit error rates, high uptime, and redundant systems,” said Chen. “Automotive components must operate reliably in harsh environments, with different failure modes and qualification standards such as ASIL and ISO.”
Obstacles to Tbps Automotive Ethernet
Terabit per second speeds are already found in the data center, but there are barriers to automotive deployment, in addition to the fact that today’s vehicles don’t yet require it.
“Automotive applications are moving toward 25 to 50 Gbps for now,” said Keysight’s Chang. “Tbps links are overkill for current in-vehicle needs and face challenges like power, cost, and thermal constraints.”
Reaching Tbps speeds in Automotive Ethernet is not just a bandwidth challenge, agreed Rambus’ Bahrouch. “It represents an architectural transformation across the hardware and software stack. Achieving this leap requires rethinking PHY design, electromagnetic resilience, and thermal management, especially in zonal architectures where data from radar, lidar, and cameras converge. Synchronizing hundreds of ECUs across a fault-tolerant, low-latency backbone pushes the limits of deterministic networking.”
However, the benefits are substantial. Terabit Ethernet will enable real-time sensor fusion at scale, allowing vehicles to perceive, decide, and act with greater speed and precision, and become the backbone for SDVs that evolve rather than remaining static after production.
Higher compute, more bandwidth needed
Dynamic SDVs will require more compute, which will need higher-speed Ethernet. “The technology in data centers will find its way into these vehicles,” said Siemens’ Fritz. “We’re seeing a central compute now coming out in the next couple of years with 64 or 128 CPU cores, multiple GPUs, and NPUs.”
This means ADAS and in-vehicle entertainment increase the compute requirements of the GPU. “We’re seeing a massive push in more and more processing power of the GPUs, more and more throughput required,” said Rob Fisher, senior director of product management at Imagination Technologies. “That’s driven by centralization, an increase in autonomy, and more screens, facilitated by Automotive Ethernet.”
Wireless technologies are also taking hold
Wi-Fi 7, or 8 and beyond, also may play a greater role in vehicles, not only for entertainment but even eventually for safety-critical use cases.
“We are targeting the big automotive OEMs out of North America, Europe, or Japan to reduce the amount of wiring in a vehicle,” said Ananda Roy, senior product manager of low-power edge AI at Synaptics. “They say that there are about 60 to 70 different microprocessors and controllers in a car, and there are a few hundred CAN buses over wire, which basically makes it very expensive to build. But also, when there are collisions, they are the ones very likely to catch fire, impacting passenger safety. If you can reduce those wires, there’s a tremendous advantage to both users and these OEMs who make these cars. We are enabling it through Wi-Fi 7.”
The first duties Wi-Fi will take from CAN are infotainment-driven. “If you think of the infotainment system these days, especially if you have kids watching TV — or you are a busy professional who is taking a ride with someone else, where you want to access Internet content — those rear seat entertainment systems would be pulling from the head unit over the CAN bus,” said Roy. “Now, all that entertainment and stuff can happen over Wi-Fi. We are trying to make it safety-critical, as well. You don’t need to miss anything with Wi-Fi 7, which is sub-10 millisecond latency.”
Conclusion
At whatever speed the technology in the vehicle requires, Automotive Ethernet brings advantages over CAN and is likely to replace it in many use cases.
“If you’ve got a little extra wiring in there, it weighs hardly anything even with shielding, and data’s going to get there fast enough, Automotive Ethernet dramatically simplifies one aspect of these complex system designs,” said Siemens’ Fritz.
Others agree. “Automotive Ethernet is dominating the marketplace right now,” said Infineon’s Yeager. “It is the backbone of the zonal architecture, and it is the reason why we’re able to do software-defined vehicles today.”
Finally, AI is a key driver. “We are seeing ‘AI everywhere’ from cloud and edge computing to AIoT and automotive,” said Cadence’s Chen. “Automotive system designs are increasingly incorporating high-speed interconnect protocols such as PCIe, UCIe, and Ethernet, mirroring trends in data centers.”
Related Articles
Automotive Outlook: 2026
Tariffs, EV costs and challenges, and fundamental architectural and technology improvements add up to transformative multi-year, multi-market changes.
How Fast Can Germany Shift To Software-Defined Vehicles?
Design, manufacturing, and business overhaul is beginning, but threat from low-cost, feature-rich vehicles from China continues to grow.
Often Overlooked, PHYs Are Essential To High-Speed Data Movement
From smartphones to AI factories, physical layers are the unsung heroes of data communications.
ADAS Adds Complexity To Automotive Sensor Fusion
Advancements in combining sensors enabling intelligent, distributed processing and standardized communication of object data.
Will Automotive Ethernet Win?
Why it’s so hard to provide a definitive answer, and what are the other contenders.
 |
 |
 |
 |
 |
 |
 |
 |
Leave a Reply