Is there any interest in making a Furiosa NPU dongle for consumers, to connect to existing computers? Currently people are being priced out of buying new computers due to the RAM/SSD/CPU shortages and most PCs are coming with insufficient RAM. If there were a way to add fast LLM inference capability to an existing computer, it could alleviate a pain point for people who want to run LLMs locally.
Thank you for the suggestion. We definitely understand the interest in affordable local LLM inference on existing PCs.
At the moment, however, our focus is on AI accelerators for data center deployments, where we believe we can deliver the greatest value in terms of performance, power efficiency, and scalability. As a result, a consumer-oriented NPU dongle is not part of our current product roadmap.
We appreciate your feedback, and we’ll keep it in mind as we continue to evaluate future market opportunities.