Skip to main content
Question

Strategic Edge AI Hardware Evaluation for KOPO / Nathan AI Ecosystem

  • Oct 2, 2026 5:20 PM UTC
  • 2 replies
  • 147 views

Hello Axelera AI Team,

I’m Nathan Davies, Founder and Chairman of KOPO, a U.S.-based technology company developing an integrated digital ecosystem for the U.S. and African markets.

We are currently evaluating edge-AI hardware and acceleration technologies for Nathan, KOPO’s AI platform, across several classes of devices and infrastructure.

Our interest in Axelera AI is broader than a single development board. We would like to understand whether your technology could support a scalable edge-AI architecture across applications such as smart devices, merchant/payment hardware, intelligent terminals, mobility, vision-enabled systems and other KOPO hardware platforms.

In particular, we are interested in understanding:

  • Which current Axelera products/platforms would be most appropriate for embedded and edge inference deployments.
  • Supported AI frameworks, models and development toolchains.
  • Performance and power characteristics for on-device inference.
  • Availability of SDKs/APIs and developer hardware for technical evaluation.
  • Options for integrating Axelera technology into third-party/OEM hardware.
  • Production availability, lifecycle support and volume deployment.
  • Whether Axelera works directly with companies developing their own AI-enabled products and ecosystems.

We are not looking to disclose proprietary Nathan architecture publicly. If there appears to be a technical fit, we would welcome a direct discussion with the appropriate Axelera technical/business team under NDA.

Our objective is to determine whether Axelera could become a technology component within KOPO’s longer-term edge-AI hardware strategy rather than simply evaluate an isolated development kit.

Thank you, and I look forward to connecting with the appropriate team.

Nathan 
KOPO

2 replies

Spanner
Axelera Team
Forum|alt.badge.img+4
  • Axelera Team
  • Oct 8, 2026 3:32 PM UTC

Hi Nathan, thanks for the detailed post and for your interest in Axelera AI! Welcome on board.

The community is a bit more of a technical space, while some of this probably leans more towards the product or sales guys, but let’s work through your points nonetheless and see what I can do:

  • Products: our current Metis-based line covers a range of form factors: PCIe cards (Axelera Edge 130p, Axelera Server 150p), M.2 modules (Axelera Embedded 110m/113m) and a single-board computer (Axelera Embedded 111c). Which fits best kinda depends on the host system and the power envelope of each device class you have in mind, really. It’s hard to be more specific, as the accelerator is more like one link in the chain, rather than the deciding factor. Product details are at https://axelera.ai and https://store.axelera.ai for more detailed info though.
  • Software: the Voyager SDK is public on GitHub (github.com/axelera-ai-hub/voyager-sdk) with release notes, a compatibility matrix and a model zoo of computer-vision networks (detection, classification, pose, segmentation), so you can review the toolchain and supported models before committing to any hardware. Documentation lives at https://docs.axelera.ai.
  • Evaluation hardware is sometimes available, and the community is the right place for technical questions once you're hands-on.

For the commercial topics, OEM/third-party integration, production availability, lifecycle and volume, and any discussion under NDA, those are best handled directly by our business team rather than on the forum. Please reach out via the contact form on axelera.ai and reference this thread; I'll also flag it internally so the right people are aware. 👍

Looking forward to seeing what you build!


  • Author
  • Cadet
  • Oct 12, 2026 12:15 PM UTC

Hello Spanner/ AI Team,

Thank you for your thoughtful and detailed response, and for welcoming KOPO to the Axelera AI community.

We appreciate the clarification regarding your Metis-based hardware portfolio, the Voyager SDK and the distinction between technical community discussions and commercial partnership discussions.

Your response is helpful because our interest in Axelera AI extends beyond evaluating individual accelerator modules. We are exploring how advanced, power-efficient AI computing technologies could become foundational components within a broader ecosystem of KOPO-developed intelligent products.

I would like to provide additional context regarding our objectives so that your technical and business development teams can better understand the potential scope of our interest.

  1. ABOUT KOPO AND OUR LONG-TERM VISION

KOPO is a U.S.-based technology company developing an integrated ecosystem spanning artificial intelligence, consumer electronics, digital commerce, financial technology, communications, mobility and intelligent connected devices.

A central component of our technology strategy is Nathan AI, which we envision as an intelligent software layer capable of supporting multiple product categories and interacting with users, devices, applications and services across the KOPO ecosystem.

Our longer-term objective includes developing Nathan OS as a KOPO-controlled software and operating-system environment, with increasing ownership of the user experience, device intelligence, system integration and core software capabilities.

We recognize that different devices will require different operating systems, processors, accelerators and software architectures during the development process.

Accordingly, we are not seeking a single chip or accelerator to solve every AI computing requirement. Instead, we are evaluating specialist technology partners whose products can support a modular, scalable and commercially sustainable architecture.

Axelera AI is particularly interesting to us because of its focus on efficient edge-AI acceleration.

  1. POTENTIAL APPLICATIONS WITHIN THE KOPO ECOSYSTEM

We see several areas where Axelera's technology may be relevant.

A. Intelligent Cameras and Security Systems

We are evaluating intelligent camera and security technologies that could incorporate:

  • Real-time object and event detection.

  • Person, vehicle and activity recognition.

  • Edge-based video analytics.

  • Intelligent alerts and automation.

  • Local AI processing to reduce dependence on continuous cloud connectivity.

  • Integration with KOPO-controlled applications and device-management systems.

Our preference is to retain control over the customer-facing software, device-management experience and associated AI services.

We would be interested in understanding how Axelera accelerators could be integrated with third-party camera hardware, image sensors, video-processing systems and Linux-based computing platforms.

B. Robotics and Autonomous Systems

KOPO is also exploring intelligent robotics and autonomous-device applications.

Potential workloads include:

  • Object detection and classification.

  • Visual perception and environmental awareness.

  • Pose estimation.

  • Semantic segmentation.

  • Real-time video and sensor analysis.

  • AI-assisted navigation and robotic task execution.

We understand that an AI accelerator does not replace the host processor, robotics controller or safety-critical control systems.

Our interest is in determining how Axelera could provide an efficient perception-processing component within a broader robotics architecture.

C. Embedded AI Devices and Consumer Electronics

We are evaluating opportunities to integrate local AI capabilities into compact computing devices, intelligent home products and other KOPO-branded electronics.

Depending on the product, these systems may require combinations of:

  • Low-power AI inference.

  • Local computer vision.

  • Privacy-conscious on-device processing.

  • Integration with Nathan AI services.

  • Secure software updates.

  • Reliable operation with intermittent internet connectivity.

We would welcome guidance on which Axelera products are most appropriate for compact, low-power embedded systems.

D. AI Computing and Edge Infrastructure

We are also interested in distributed AI computing, where selected workloads can run locally rather than relying exclusively on centralized cloud infrastructure.

For KOPO, this may eventually involve edge gateways, intelligent business devices, local AI processing systems and commercial computing products.

We would like to evaluate where Axelera's current hardware can provide meaningful acceleration and where complementary CPU, GPU or NPU platforms would be required.

  1. AXELERA PRODUCTS WE WOULD LIKE TO EVALUATE

Based on your response, we would initially prioritize the following products:

Embedded 110m and 113m

We are interested in these modules for compact embedded systems where physical size, power consumption and host compatibility are important.

Embedded 111c

This appears potentially useful for early technical evaluation, proof-of-concept development and understanding the complete hardware and software integration process.

Edge 130p

We would like to evaluate this for higher-performance edge computing, intelligent security systems and host platforms with suitable PCIe connectivity.

Server 150p

This may become relevant for larger inference deployments and specialized edge-server applications, depending on supported workloads, economics and system requirements.

We would appreciate your team's recommendations regarding the most appropriate evaluation hardware for these application categories rather than assuming that one accelerator is suitable for all of them.

  1. VOYAGER SDK, SOFTWARE INTEGRATION AND NATHAN AI

Thank you for directing us to the public Voyager SDK, GitHub repository, documentation, compatibility matrix and model zoo.

We will review these resources as part of our technical evaluation.

An important aspect of our strategy is maintaining ownership and control of KOPO-developed applications and software.

We therefore want to understand how Voyager can operate within an independently developed software environment.

In particular, we are interested in:

  • Supported Linux distributions, kernel requirements and host architectures.

  • Compatibility with ARM-based and x86-based host processors.

  • Supported AI frameworks, model formats and conversion workflows.

  • Deployment of custom-trained models.

  • Access to inference outputs for integration with KOPO applications and Nathan AI.

  • Multi-camera and real-time video-processing capabilities.

  • SDK redistribution and commercial deployment rights.

  • Firmware, runtime and driver maintenance.

  • Long-term compatibility and software support.

We also understand from your public materials that Voyager currently emphasizes computer-vision workloads, including detection, classification, pose estimation and segmentation.

Because Nathan AI is intended to support broader forms of intelligence, we would appreciate clarification regarding the boundaries of Axelera's current accelerator capabilities.

Specifically, we would like to understand whether your existing or planned hardware and software platforms support language models, multimodal inference, vision-language models or other generative AI workloads.

We are not assuming that these capabilities are currently available. We simply want to understand Axelera's technology roadmap and identify where its accelerators may complement other computing platforms.

  1. LONG-TERM SOFTWARE AND PLATFORM INDEPENDENCE

A fundamental principle of our product strategy is that KOPO should retain control of its software ecosystem and customer relationships.

We are comfortable integrating proprietary accelerator hardware and supported software runtimes where necessary.

However, we want our own applications, AI services, device interfaces and business logic to remain independently developed and controlled by KOPO.

For this reason, we would be interested in understanding whether commercial deployments can operate without mandatory dependence on Axelera-hosted cloud services, whether offline inference is supported and what licensing arrangements apply to applications distributed with Axelera-powered hardware.

We would also appreciate clarity on how SDK updates, hardware revisions and product lifecycle changes are managed so that commercial products remain supportable over time.

  1. OEM, ODM AND COMMERCIAL PARTNERSHIP OPPORTUNITIES

Beyond technical evaluation, we would like to explore whether Axelera supports long-term OEM relationships with companies integrating its accelerators into their own branded products.

The principal commercial areas of interest include:

  • Evaluation hardware and development kits.

  • Engineering and integration assistance.

  • OEM supply arrangements.

  • Production availability and lead times.

  • Minimum order quantities and volume pricing.

  • Hardware lifecycle and long-term availability.

  • Technical support and escalation arrangements.

  • Commercial SDK and runtime licensing.

  • Product roadmap alignment.

  • Opportunities for collaborative reference designs or system integration.

At this stage, we are not proposing fixed purchasing volumes or committing to a particular hardware architecture.

Our intention is to identify technically suitable products, complete meaningful evaluation and then develop a commercially practical relationship based on validated requirements.

  1. PROPOSED EVALUATION APPROACH

We believe a phased approach would be most productive.

First, our technical team would review Voyager SDK documentation, supported models and host compatibility.

Second, we would select a limited number of appropriate evaluation products, potentially including an embedded module, a development platform and a PCIe accelerator.

Third, we would test representative computer-vision workloads and their integration with KOPO-developed software.

Fourth, we would evaluate performance, latency, power consumption, thermal characteristics, deployment requirements and total system cost.

Finally, if the technical and commercial results are satisfactory, we would explore production integration and a longer-term supply relationship.

This approach would allow both teams to evaluate the opportunity realistically without making premature assumptions about product suitability or manufacturing commitments.

  1. REQUEST FOR INTRODUCTION TO YOUR BUSINESS AND ENGINEERING TEAMS

We appreciate your offer to flag our interest internally.

Since our requirements span both technical evaluation and potential OEM cooperation, an introduction to the appropriate Axelera business development or strategic partnerships representative would be particularly helpful.

We will also use your official contact channel as recommended and reference this community discussion.

If possible, we would appreciate guidance on:

  1. Which Axelera product your engineering team would recommend for our initial embedded computer-vision evaluation.

  2. Whether evaluation hardware is currently available for prospective OEM partners.

  3. The appropriate contact for discussing commercial integration, licensing and longer-term supply arrangements.

  4. Whether Axelera offers direct engineering engagement during early proof-of-concept development.

  5. Whether a technical discussion can be arranged after we have reviewed the SDK and established our initial evaluation requirements.

We would be happy to execute an NDA before sharing confidential product architectures, detailed development plans or commercially sensitive information.

CLOSING

Our broader ambition is to develop a family of intelligent products that share a common KOPO software and AI experience while using the most suitable hardware technologies for each application.

We believe the future of AI will involve a combination of cloud intelligence, local computing and specialized edge acceleration.

Our interest in Axelera is therefore strategic: we want to understand whether your technology can become one of the important building blocks supporting KOPO's intelligent-device ecosystem over the long term.

Thank you again for the helpful response and for offering to direct our inquiry internally.

We look forward to continuing the technical discussion and exploring potential cooperation with your business development team.

Best regards,

Nathan