Aeonsemi has introduced the Praxium family of multi-gigabit Ethernet PHYs, targeting the connectivity requirements of robots, autonomous vehicles and intelligent machines. The new chips support data rates up to 10Gbps in a 6 × 6 mm QFN package, combining low-power Ethernet connectivity with support for cameras, LiDAR, zone controllers and AI compute networks.
Aeonsemi, a fabless semiconductor company focused on Ethernet connectivity and timing technologies, has launched the Praxium family of multi-gigabit Ethernet physical-layer transceivers (PHYs) designed for emerging Physical AI systems.
The product family is aimed at machines that increasingly combine distributed sensors with centralized or distributed AI compute. Robots, autonomous vehicles and industrial machines can contain multiple cameras, LiDAR systems, sensors and actuators, creating a networking challenge as the volume of data moving between edge devices and AI processors increases.
Praxium is designed to address that challenge with Ethernet data rates reaching 10Gbps while keeping the PHYs small enough for space-constrained embedded systems. Single-port devices use a 6 × 6 mm QFN package, according to Aeonsemi, while the family also includes a four-port device intended for higher-density network applications.
The portfolio comprises four devices: the AS31010, a single-port 10G/5G/2.5G PHY; the AS31510, supporting 5G/2.5G; the AS31210, supporting 2.5G; and the AS31040, a quad-port 10G/5G/2.5G PHY.
The technology builds on IEEE 802.3ch, the automotive Ethernet amendment that specifies 2.5Gbps, 5Gbps and 10Gbps operation over a single balanced pair of conductors. The standard was developed for automotive applications and provides a common foundation for multi-gigabit networking between electronic systems.
Aeonsemi says Praxium extends that foundation by supporting additional physical cabling options, including coax and miniature twinax, alongside shielded twisted pair. For system designers, the choice of cable can affect weight, diameter, routing flexibility, electromagnetic compatibility and cost.
That flexibility becomes particularly relevant in robotics and autonomous machines, where cables must pass through mechanical joints and constrained spaces while carrying high-bandwidth sensor data. Instead of deploying different networking technologies for different sections of a machine, designers can use Ethernet as a common architecture from distributed sensors and actuators through network backbones and AI compute.
Power consumption is another focus. Aeonsemi says its asymmetric 10Gbps sensor-link configuration can operate at as little as 350mW of total chip power when Energy-Efficient Ethernet is used on the lower-speed uplink. The PHYs also support Power over Coax (PoC) and Power over Data Lines (PoDL), potentially allowing data and power delivery to be consolidated for some distributed sensor designs.
The Praxium family also incorporates timing and security features aimed at machine networking. IEEE 1588 and IEEE 802.1AS/gPTP support provides precision time synchronization, while an integrated IEEE 802.1AE MACsec engine is designed to secure Ethernet communications at the link layer. Aeonsemi says the devices are designed for AEC-Q100 Grade 2 qualification for automotive applications.
The need for this type of connectivity is growing alongside robotics deployment. The International Federation of Robotics reported that 542,000 industrial robots were installed worldwide in 2024, more than twice the number installed a decade earlier. Asia accounted for 74% of new deployments that year, while India ranked sixth globally for annual industrial robot installations.
In the United States, preliminary IFR data shows industrial robot installations increased 11% year over year to 38,000 units in 2025. The automotive sector remained the largest adopter, with 13,500 installations.
As robots become more capable, networking requirements are changing alongside compute architectures. Traditional industrial networks were often designed around relatively predictable control signals, while Physical AI systems can require continuous streams of visual and spatial data from cameras and LiDAR to feed perception and decision-making systems.
That makes Ethernet an increasingly relevant layer between sensing and AI inference. High-speed links can aggregate data from multiple endpoints, while technologies such as Time-Sensitive Networking and precision time synchronization can help coordinate distributed systems.
The competitive landscape includes established automotive Ethernet and networking semiconductor vendors, as well as suppliers developing connectivity specifically for AI-enabled machines. The differentiation is increasingly shifting beyond peak bandwidth toward power efficiency, package size, deterministic behavior, security, synchronization and the ability to support multiple physical media.
For Aeonsemi, Praxium places its PHY technology at the intersection of AI infrastructure and embedded networking. The company is not supplying the AI processor itself; instead, it is targeting the connectivity layer that moves sensor and control data to the processors responsible for perception, planning and machine intelligence.
That layer could become increasingly important as Physical AI expands from factory automation into autonomous vehicles, mobile robots and humanoid systems. For these machines, AI performance depends not only on compute capacity but also on how efficiently data can move between the physical world and the systems processing it.
Market Landscape
Physical AI is increasing the amount and diversity of data moving through robots, autonomous vehicles and intelligent machines. Ethernet offers a scalable networking architecture spanning sensor links, control systems and high-speed backbones, while IEEE 802.3ch provides standardized 2.5Gbps, 5Gbps and 10Gbps automotive Ethernet operation.
The robotics market provides a growing deployment base. IFR recorded 542,000 industrial robot installations globally in 2024, with Asia representing 74% of new deployments. India installed 9,120 industrial robots in 2024, a 7% increase from the previous year.
This expansion increases demand for compact, power-efficient networking components capable of handling high-bandwidth sensor traffic while meeting automotive and industrial reliability requirements.
Top Insights
- Aeonsemi’s Praxium family provides up to 10Gbps Ethernet connectivity for cameras, LiDAR, zone controllers and AI-enabled machines.
- Single-port PHYs fit into a 6 × 6 mm QFN package, targeting space-constrained robotic and automotive designs.
- Support for STP, coax and mini-twinax gives designers greater flexibility over cable weight, routing and electromagnetic performance.
- IEEE 1588, gPTP, TSN-related timing and MACsec capabilities target synchronized and secure machine networks.
- Global industrial robot installations reached 542,000 units in 2024, highlighting the expanding infrastructure base for Physical AI.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
