Access Official Technical PDF

Share Summary:LinkedInEmail

Spaceflight computing places demanding constraints on electronic systems: hardware must operate in a challenging environment while designers manage component count, size and power consumption. In September 2020, BrainChip Holdings and VORAGO Technologies announced a collaboration through BrainChip’s Akida Early Access Program to support a Phase I NASA program for a neuromorphic processor intended to meet spaceflight requirements. The announcement described a proposed combination of Akida processing and VORAGO’s radiation-hardening expertise; it did not report a completed processor or program outcome.

Why a neuromorphic processor is relevant to spaceflight

Neuromorphic computing takes inspiration from the way the brain processes information. BrainChip described Akida as an event-based neural network processor designed for edge applications, where sensor inputs can be analyzed at the point of acquisition rather than first being sent to a cloud service or data center. In a spaceflight system, local processing is relevant because the computing task must be considered alongside the available hardware, power and system architecture.

According to the announcement, Akida is designed as a complete neural processor that does not require an external CPU, memory or Deep Learning Accelerator (DLA). That level of integration could be relevant to aerospace designers evaluating how to limit the number of components in a system. The announcement identified component count, size and power consumption as important considerations for spaceflight and aerospace applications, and presented Akida’s integration and low-power design as potentially aligned with those requirements.

This is an architectural rationale, not a quantified performance result. The announcement did not provide specific power figures, radiation test results, or details of a flight-ready implementation. Those distinctions matter when assessing early-stage technology collaborations: design intent and suitability are not the same as qualification or demonstrated performance in an operational mission.

Incremental learning at the edge

Another capability highlighted by BrainChip was incremental learning. The company said that new classifiers can be added to a network without retraining the entire network. In practical terms, this approach is intended to let a system adapt its classification capability as new discoveries or circumstances arise, while continuing to operate locally.

For aerospace applications, the proposed value is the possibility of responding to changing conditions without relying on a full network retraining process. Local learning and inference also fit the broader edge-computing model described by BrainChip: analyze sensor information where it is acquired. The announcement characterized real-time local incremental learning as beneficial when new discoveries or circumstances occur. It did not specify which classifiers, sensors or mission tasks would be involved in the NASA Phase I program, so those implementation details should not be assumed.

VORAGO’s role in radiation-hardened electronics

VORAGO brought a different part of the proposed solution: experience with radiation-hardened and extreme-temperature semiconductor solutions for high-reliability markets. The company described its patented HARDSIL technology as a process for hardening commercially designed semiconductor components for operation in extreme environments. Its stated application areas included aerospace, defense and industrial systems.

Combining that expertise with Akida was the central idea behind the collaboration. Neuromorphic processing addresses how a system handles data and performs AI tasks; radiation hardening addresses the semiconductor’s ability to operate in demanding environments. These are distinct engineering concerns, and a collaboration can bring them together for evaluation without implying that every integration, qualification or mission requirement has already been resolved.

The announcement said the agreement was intended to support Phase I of a NASA program for a neuromorphic processor that meets spaceflight requirements. BrainChip’s Early Access Program provided selected customers with early access to the Akida device, evaluation boards and dedicated support. The agreement also included payments intended to offset BrainChip’s expenses in supporting partner needs. This early-access structure offered a basis for technical evaluation, rather than evidence of a completed product deployment.

What the collaboration means for system designers

For engineers and technical decision-makers, the announcement illustrates a systems-level approach to edge AI in harsh environments. Assessing a processor for spaceflight involves more than evaluating its learning architecture. Teams also need to consider integration, the surrounding semiconductor technology and the requirements of the target application. In this case, the stated areas of interest included:

  • Integration: Akida was described as a complete neural processor that does not require an external CPU, memory or DLA.
  • Power and physical constraints: The announcement identified power consumption, size and component count as key aerospace considerations.
  • Local processing: Akida was designed to analyze sensor inputs at the point of acquisition.
  • Adaptation: Incremental learning was presented as a way to add classifiers without retraining the entire network.
  • Environmental suitability: VORAGO contributed experience in radiation-hardened and extreme-temperature solutions.

These points can help frame an evaluation, but they do not replace application-specific requirements, testing or qualification. The announcement supplied no detailed test plan or measured results, making it important to distinguish the collaboration’s stated objectives from verified system performance.

Conclusion

The BrainChip–VORAGO agreement brought together edge-oriented neuromorphic processing and radiation-hardening expertise in support of a NASA Phase I program. Its significance lies in the engineering questions it put in focus: how an integrated processor might reduce system demands, how local incremental learning could support changing conditions, and how semiconductor technology must be suited to the operating environment. For automation and aerospace decision-makers, the practical takeaway is to evaluate these capabilities as parts of a complete system—and to treat early-access collaboration as a step toward assessment, not proof of mission readiness.

Access Official Technical PDF

About The Author