2025 Automotive SerDes and Bridge Chip Report: Market Outlook to 2035

Disclaimer: this article is for information and education only. It reflects publicly available data, industry trends and professional analysis as of the publication date (August 2025). Every effort has been made to be accurate, but no guarantee is given that it is complete, correct or still current.

The report includes interpretations, forecasts and opinions based on public information from vendors, consortia and research sources. They are not financial, investment, legal or engineering advice, and no OEM, semiconductor company or industry body has endorsed them.

All trademarks, logos and product names belong to their owners. Companies, protocols and standards are mentioned only for analysis, and this does not imply endorsement or affiliation.

Please check official specifications and consult analysts and domain experts before making strategic, engineering or investment decisions based on this report.

For a more detailed and accurate report, please contact us.

Executive summary

Automotive SerDes (serialiser/deserialiser) chips carry camera, radar, LiDAR and display data over a single thin cable between sensors, ECUs and screens. Bridge chips translate between the protocols on either side of that link. Today the market is split between two established proprietary families, TI's FPD-Link and Analog Devices' GMSL, and a group of open standards (MIPI A-PHY, ASA Motion Link, China's HSMT and the newer OpenGMSL). Our view is that the market consolidates to two or three main protocols by about 2030, with open standards taking a growing share of new platforms.

Zonal architectures, safety and cybersecurity rules (ISO 26262, UNECE R155 and R156) and higher-bandwidth sensors are all reshaping it. SerDes links sit at the centre of advanced driver assistance (ADAS) and in-vehicle infotainment (IVI). This report compares the main protocols, reviews the main suppliers, looks at what OEMs need and projects adoption to 2035, covering both the technology and what customers value.

Key takeaways

• A modern car can carry ten or more camera and display links, and each new sensor generation pushes the per-link rate up.
• FPD-Link and GMSL have the largest installed base and ecosystem in the near term.
• Open standards (A-PHY, ASA Motion Link, HSMT, OpenGMSL) compete on interoperability and second sourcing.
• Published roadmaps reach about 32 Gbps per link, over reaches of up to about 15 m.
• Diagnostics, functional safety and link security are becoming as important as raw bandwidth.

1. Definitions and applications

• SerDes: high-speed interface chips that serialise data for transmission over a cable and deserialise it at the far end. They are essential for camera-to-ECU, ECU-to-display and sensor-to-SoC links.
• Bridge chips: interface ICs that connect different protocols, for example converting MIPI CSI-2 to FPD-Link or GMSL, so older and mixed-vendor parts can work together.
• ADAS uses: camera streaming, radar data and LiDAR point clouds. These need high bandwidth, low latency and ASIL-rated links.
• IVI uses: infotainment displays, instrument clusters, head-up displays and rear-seat screens. These need high resolution, good EMI performance and low-cost cabling.

How an automotive SerDes link works

A camera module contains an image sensor with a MIPI CSI-2 output, which is designed for a few centimetres of PCB trace. The serialiser in the camera converts that parallel or CSI-2 stream into one high-speed serial lane and drives it over coax or shielded twisted pair. At the ECU, the deserialiser recovers the clock, decodes the data and presents CSI-2 again to the SoC. Most automotive links are asymmetric: a fast downlink for video and a slow back channel (typically tens to a few hundred Mbps) for I2C or SPI control, GPIO and triggers. Many also carry power over the same coax (power over coax), which saves a wire per camera.

Worked example: sizing a camera link

Take an illustrative 8-megapixel front camera at 3840 x 2160, 30 frames per second, RAW12:

• Pixels per second: 3840 x 2160 x 30 = about 249 million
• Payload: 249 million x 12 bits = about 3.0 Gbps
• With an illustrative 20% allowance for blanking, protocol overhead and line coding: about 3.6 Gbps

That fits a 6 Gbps class link with margin. Move to 60 frames per second or to RAW16 with HDR and the same camera needs 7 to 8 Gbps, which is why 12 Gbps and faster links are arriving for front cameras. The overhead figure is illustrative; real numbers depend on the protocol and sensor mode.

2. Protocol comparison

Protocols covered:

• FPD-Link III and IV (Texas Instruments)
• GMSL2 and GMSL3 (Analog Devices)
• ASA Motion Link (Automotive SerDes Alliance)
• MIPI A-PHY (open standard from the MIPI Alliance)
• APIX3 and APIX4 (INOVA Semiconductors)
• HSMT (a Chinese industry standard)
• OpenGMSL (an open-standard effort based on GMSL)

Protocol	Owner or body	Type	Published downlink rate (approx.)	Notes
FPD-Link III / IV	Texas Instruments	Proprietary	About 4 Gbps (III), about 13 Gbps (IV)	Large installed base in cameras and displays
GMSL2 / GMSL3	Analog Devices	Proprietary, moving to open via OpenGMSL	6 Gbps (GMSL2), 12 Gbps (GMSL3)	Widely used in ADAS cameras
MIPI A-PHY	MIPI Alliance	Open standard	Up to 16 Gbps (v1.0), 32 Gbps (v1.1 over two ports, v2.0 on one cable)	Reach up to 15 m; IEEE 2977 adopts it
ASA Motion Link	Automotive SerDes Alliance	Open standard	Up to 16 Gbps in the first version	Ethernet-friendly, security built in
APIX3 / APIX4	INOVA Semiconductors	Proprietary	Not re-verified for this report	Display focus, combines video and Ethernet
HSMT	China Automotive Industry Standards Committee	National industry standard	2 to 12.8 Gbps downlink in the published grades	Video standard QC/T 1217-2024
OpenGMSL	OpenGMSL Association (launched June 2025)	Open standard effort	Based on GMSL rates	Compliance testing planned for multi-vendor interoperability

Rates in the table are the headline figures published by vendors and consortia, rounded, and were not re-measured for this report. Check the current specification for any design decision.

What we compared:

• Bandwidth: up to 32 Gbps on next-generation roadmaps
• Reach: up to about 15 m over coax or shielded twisted pair
• Topology: point-to-point, daisy chain or star
• Functional safety: ASIL B to D, depending on how mature the protocol is
• Ecosystem: broadest today for FPD-Link and GMSL
• Interoperability: strongest for open standards such as ASA Motion Link, A-PHY and HSMT
• Outlook to 2035: strongest for open, scalable solutions

Key point: we expect the protocols to consolidate. Open standards (ASA Motion Link, A-PHY, HSMT) are gaining ground in Asia and Europe, while the established protocols (FPD-Link, GMSL) keep strong momentum in the near term. The OpenGMSL Association, announced by Analog Devices and partners in June 2025, is an attempt to carry GMSL's installed base into the open-standard camp.

3. Competitor benchmarking

Company	Strengths	Weaknesses
Texas Instruments	Broad portfolio, ASIL-ready parts, large installed base	FPD-Link is proprietary
Analog Devices	GMSL widely used in ADAS, strong IVI experience	Licensing limits, ecosystem lock-in (OpenGMSL aims to address this)
Valens	Leader in MIPI A-PHY, open-standard support	Market still maturing
INOVA	Combined automotive Ethernet and display links with APIX4	Limited visibility on ASIL D
Sony	Image sensors with integrated SerDes	Not a full-range SerDes supplier
Intel / Mobileye	Own interfaces, tight SoC integration	Less open than industry standards
HSMT backers	Government support, adoption by Chinese OEMs	Little traction outside China so far

4. Customer needs and emerging solutions

What OEMs need

• Interoperability across vendors
• Standard diagnostics
• ASIL D readiness
• Low-latency, high-resolution data transfer
• Flexible topologies: star, daisy chain and ring

Solutions now appearing

• Bridges that work across protocols
• Smarter, AI-assisted diagnostics in SerDes ICs
• Open standards such as ASA Motion Link and A-PHY
• Link layers that can be updated in software
• PHYs designed to move to optical links later

5. Emerging technologies

• AI and ML in SerDes chips: in-band diagnostics, link-quality monitoring and early warning of failures.
• PCIe and CSI-2 tunnelling bridges: for sensor multiplexing and camera aggregation.
• Zonal gateways: bridges that aggregate links from smart sensors.
• Cloud-linked diagnostics: SerDes health data feeding fleet management.
• Multi-gigabit optical SerDes: longer reach and immunity to EMI.

6. Customer challenges and the features they want

Main challenges

• Protocol lock-in and limited interoperability
• EMI sensitivity at high bandwidth
• Certification effort for ASIL B and D
• Thermal budgets and PCB space

Features in demand

• In-band diagnostics and telemetry
• Fail-safe behaviour ready for ASIL B and D
• Compatibility with common cables and connectors (STP, coax)
• Automatic link negotiation and recovery
• Multi-channel aggregation

Common mistakes when choosing a link

• Sizing for today's sensor only. Cameras and displays usually get a resolution or frame-rate bump within one vehicle generation.
• Ignoring the back channel. Sensor control, triggers and diagnostics all ride on it, and some functions need low latency.
• Treating the cable as free. Cable, connector and harness cost and weight often exceed the cost of the chips at each end.
• Leaving security late. UNECE R155 makes cybersecurity part of type approval, so link authentication and encryption need a plan early.

7. Assumptions and risks

Assumptions

• Zonal architectures ramp up between 2027 and 2030
• Two or three main SerDes protocols dominate by 2030
• Open standards grow as MIPI and ASA gain support
• Cloud diagnostics grow with connected vehicles

Risks

• Regional politics keep the protocols fragmented
• Regulators are slow to adopt new in-vehicle network standards
• Vendors resist open ecosystems to protect their IP

8. Measures of success

These are our proposed targets for judging how the market develops, not published industry figures.

For vendors

• Design wins at five or more OEMs by 2030
• ASIL certification on at least 80% of the portfolio
• In-band diagnostics in more than 80% of ICs
• Ecosystem integration with ten or more partners

For OEMs and Tier 1s

• Interoperability on 80% of platforms
• Zonal architecture on half of new platforms by 2030
• 15 to 20% savings in cabling and power
• Cloud diagnostics that cut time to root cause for field failures to under five minutes

For the market

• Consolidation to three or fewer main SerDes standards
• Open protocols on 70% of new platforms
• Cloud-linked SerDes data in 60% of fleets by 2035
• Similar protocol choices across China, the US and Europe

FAQ

Why not just use automotive Ethernet for cameras?

Multi-gigabit automotive Ethernet is symmetric and packet based, while cameras and displays need a fast, low-latency stream in one direction and a slow control channel in the other. Asymmetric SerDes is usually cheaper and lower power for that job. The two increasingly meet in zonal gateways.

Is MIPI A-PHY the same as MIPI CSI-2?

No. CSI-2 is the short-reach camera interface on the board. A-PHY is the long-reach physical layer that carries CSI-2 and other protocols across the vehicle.

Can GMSL and FPD-Link parts talk to each other?

Not directly. A bridge or a matching serialiser and deserialiser from the same family is needed, which is the lock-in the open standards aim to remove.

What is HSMT?

HSMT (high-speed media transmission) is a family of Chinese automotive industry standards. Its video link is published as QC/T 1217-2024.

Final takeaway

The automotive SerDes and bridge market is fragmented today but is moving towards open standards and smarter links. FPD-Link and GMSL will coexist with newer standards such as ASA Motion Link, MIPI A-PHY and HSMT. The vendors that do best will be those that make interoperability easy, meet the safety standards and help OEMs move to zonal and software-defined architectures. As diagnostics move to the cloud and AI reaches the physical layer, SerDes will change from simple links into critical, intelligent infrastructure.