Intel and Atsign announced a partnership this week that claims to have eliminated the encryption bottleneck that has been holding back AI agents running on edge devices. The two companies demonstrated an 8,700% performance improvement in hardware-accelerated, end-to-end encrypted communications using the Agent2Agent (A2A) standard, a result that could reshape how autonomous vehicles and smart city infrastructure handle secure inter-agent communication.
The encryption problem at the edge
AI agents operating in distributed environments face a basic tension. They need to exchange sensitive data, coordinate workflows, and make decisions in real time. But every one of those operations requires cryptographic protection, and cryptographic operations are expensive. On resource-constrained edge devices, the overhead of encryption, key management, and secure communication has been a practical limit on how many agents can run simultaneously and how fast they can coordinate.
The problem gets worse in autonomous systems. A robotaxi fleet, a network of self-driving shuttles, or an intelligent traffic management system all depend on agents that must communicate securely with each other and with cloud infrastructure. If encryption adds even milliseconds of latency to each exchange, the cumulative effect across hundreds of vehicles making thousands of decisions per second becomes a safety concern, not just a performance one.
What the partnership actually built
Atsign's contribution is its zero-exposure security platform, which provides identity management, namespace isolation, scoped data sharing, end-to-end encryption, and full auditability. The platform is designed around a zero trust model, meaning every agent must prove its identity before accessing any data, and data sharing is scoped to the minimum necessary for each interaction.
Intel provided the hardware foundation. Protected key storage keeps encryption keys in hardware rather than exposing them to software. Secure boot ensures the platform has not been tampered with before agents start running. Intel's total memory encryption and full disk encryption protect data at rest and in memory. Trusted compute features verify that the execution environment is genuine.
The combination is the point. Atsign's software handles the identity and access management layer while Intel's hardware handles the cryptographic acceleration. Previous approaches relied on software-based encryption, which left a performance gap that grew wider as agent counts increased.
The benchmark numbers
On Intel Xeon processors, the partnership measured an almost 88x performance improvement, reaching throughput of 5 Gb/s for encrypted agent-to-agent communication. On Intel Core Ultra 9, the uplift was 60x, reaching 6 Gb/s. The 8,700% headline figure comes from comparing these hardware-accelerated results against the software-only baseline.
Those throughput numbers matter because they represent the ceiling for how many agents can communicate simultaneously. At 5 to 6 Gb/s of encrypted throughput, a single edge node can support a large fleet of agents exchanging telemetry, sensor data, and coordination messages without encryption becoming the bottleneck.
Why autonomous vehicles are the target
The companies specifically framed the results around smart city and autonomous transport use cases. Waymo's robotaxis, Toyota's e-Palette shuttle buses, and South Korea's autonomous vehicle testing programs powered by Ericsson and SK Telecom all represent deployments where edge AI agents must coordinate in real time with strict security requirements.
Edge AI chip vendors like Blaize have been signing smart city contracts across Asia, and the infrastructure buildout is accelerating. What has been missing is a security layer that does not slow down the agents running on that infrastructure. The Intel-Atsign partnership targets that gap directly.
Atsign CEO Aparna Rayasam described the result as a "genuinely secure foundation for agentic edge AI deployments in smart transportation environments," emphasizing that the platform lets devices "operate autonomously at the edge, where both trust and speed are essential."
What the partnership means for developers
Atsign joined Intel's Industrial Builders partner program as part of the announcement. Intel customers can now access Atsign's pre-validated solutions through the program, with a direct path to request for proposals, reference architectures, and proof-of-concept deployments.
For teams building edge AI systems, the practical takeaway is that hardware-accelerated encryption for agent communication is no longer a theoretical possibility. The performance numbers suggest that the cryptographic overhead that has been a persistent pain point in distributed agent architectures can be reduced to the point where it is no longer the limiting factor. The question now is whether the autonomous vehicle and smart city markets adopt this quickly enough to matter as the infrastructure buildout accelerates.