Arbe pitches 2,304-channel radar for eyes-off highway driving

By Olivia Hart

Arbe says Phoenix Radar uses a 2,304-channel array, 10 times the channel count of current leading alternatives, at roughly the same size and cost. CEO Ram Machness told Pulse 2.0 that it supports more than 50,000 detections per frame and produces a denser point cloud than any other radar solution.

Arbe is trying to turn that claim into a manufacturing business. Machness said his mandate covers production lines, a product-manufacturing lifecycle and an organization tuned for delivery as the company moves from product introduction to full production.

What must work at 80 mph?

Machness identified reliable eyes-off, or Level 3, driving at 80 mph as one of autonomous driving’s biggest barriers. He cited long-range detection, high-resolution object separation, dynamic range, free-space mapping and large-scale validation.

At highway speeds, sensors need time to brake or change lanes. They must separate debris from guardrails, identify the drivable area and detect weak objects beside reflective surfaces.

Arbe says its perception radar exceeds the 300-meter detection requirement. Machness claims it measures distance and relative velocity accurately, resolves horizontal and vertical detail, tracks vehicles across lanes and detects subtle cut-in maneuvers.

“With Arbe’s solution, this long-standing challenge has been resolved, enabling eyes-off driving at full highway speeds,” Machness said. The interview offered no independent highway test, fleet result, regulatory approval or program volume. Arbe still needs repeatable validation across rare events, dense traffic and severe weather before broad Level 3 adoption.

What sits inside Phoenix Radar?

Machness described an automotive-grade radio-frequency chipset with transmitter and receiver chips that combine into 48 receiving and 48 transmitting channels. He claimed leading isolation, noise figure, transmit power and the market’s lowest cost per channel.

Its Radar Processing Chip handles raw data with capacity equivalent to 3 Tb per second. It supports 2,304 virtual channels, scales from high to ultra-high resolution and is designed for low power consumption.

Phoenix Perception Radar provides 2K ultra-high resolution and supports free-space mapping, object tracking and SLAM. Its point cloud records range, azimuth, elevation and velocity, while the company says it can distinguish threats from false alarms across speed, range and weather.

Arbe positions radar to complement cameras in multi-sensor stacks and calls its image quality the best in the automotive market. It says the outputs come at the same size, cost and power consumption as current solutions, which use about 10% of its channel count.

No unit price, power measurement or comparative test appeared in the interview. The advantage remains unverified until production confirms that channel density does not compromise cost, power or manufacturability.

Can design wins become production?

Machness, appointed CEO in April after serving as Chief Business Officer, has 30 years in embedded systems, semiconductors and automotive. In that role, he led product strategy, OEM and Tier 1 business development, customer support and market expansion.

Before Arbe, he led automotive and wireless-connectivity product lines at Texas Instruments across the U.S., Europe and Asia, scaling them to more than $300 million. He held product roles at Microsoft and DSP Group.

Arbe was founded in 2015 by Kobi Marenko and Dr. Noam Arkind. Machness said it took two years to choose automotive after starting in robotics, and launched what Arbe called the world’s first ultra-high-resolution automotive radar system in 2019.

Key company and partner milestones include:

• Arbe became publicly traded on Nasdaq in 2021 and listed on the Tel Aviv Stock Exchange in 2024, under ticker ARBE.
• In late 2023, it released a production-version radar processor for OEM development projects and Tier 1 B-sample radars.
• HiRain, Sensrad and Veoneer, which Magna acquired, built radar systems around Arbe’s chipset.
• In 2024, HiRain’s LRR610 imaging radar moved toward mass production, with one million kilometers of data collection planned.
• In 2025, Sensrad began deliveries for defense autonomous off-road vehicles, boat collision prevention and smart infrastructure.
• In late 2025, a China-based state-owned automaker chose HiRain and Arbe’s radar for an L4 vehicle program.
• In 2026, Arbe announced an AI-based driving platform combining its radar with NVIDIA accelerated computing, featuring an automotive-grade system ready to integrate with complete perception stacks. Arbe also works to integrate its radar into NVIDIA’s DRIVE Hyperion.

Arbe said its technology won two CES Innovation Awards, the Informa Tech Automotive Award, Automotive News PACEpilot Innovation to Watch, Tech.AD Europe Award and Just Auto Excellence Awards. Machness said Phoenix operates at night, through glare, snow, fog and heavy rain.

“My favorite memory is when a customer informed us that our system was responsible for saving lives. That is something I will never forget,” Machness said. He did not identify the customer, and no supporting test or incident details appeared in the interview.

Arbe’s milestone list gives no order values, revenue contribution or production volumes. Commercial reach is visible; economic scale is not.

How large is the opportunity?

Third-party estimates in Arbe’s public filings put the automotive radar market at about $13.5 billion by 2032. Machness said L3 and higher autonomy could become an important part, but timing and scope depend on OEM programs, regulation and market readiness.

Interest extends to off-highway machines, commercial vehicles, agriculture, counter-drone, defense, security, robotaxis, robotrucks and smart infrastructure. Arbe does not break out a total addressable market for each vertical, pursues them selectively and says it will not overstate them before they mature.

Arbe’s physical-AI goal is affordable, scalable radar across vehicles, platforms and industries. Machness gave no revenue, cash or funding figures, directing investors to ir.arberobotics.com.

The next test is whether HiRain’s 2024 data campaign, Sensrad’s 2025 deliveries and the late-2025 China L4 selection translate into disclosed production volumes. Until then, Arbe’s shift from development programs to scaled delivery remains unproven.