# Intel Nestor Ni1000 **Entity class:** Neural-network recognition accelerator ## Definition and history The Intel/Nestor Ni1000 Recognition Accelerator was an early-1990s neural-network chip with an integrated controller and on-chip learning capability, described in [[articles/Intelligence Inside|Intelligence Inside]]. The article's source ledger attributes development funding to ARPA and the Office of Naval Research and names handwriting, fingerprint, and military-target recognition as intended applications. Its published specifications include approximately 1,024 artificial neurons and 3.7 million transistors. The original ONR-hosted document lacks a stable public URL in the article's bibliography, so exact figures retain that provenance qualification. ## Mechanism and limits The Ni1000 belongs to a product line for classification at hardware speed rather than to the later conversational interface. An integrated controller, stored learned patterns, and a learning path made recognition an embodied component of a deployable device. The record supports a funded recognition accelerator and specific applications; it does not by itself document each operational deployment. ## Relationships **Manufacturer and collaborator:** [[wiki/Intel|Intel]] and Nestor. **Predecessor comparison:** [[wiki/Intel 80170NX ETANN|Intel 80170NX ETANN]]. **Funding:** [[wiki/DARPA|ARPA]] and [[wiki/Office of Naval Research|Office of Naval Research]]. **Article route:** [[articles/Intelligence Inside|Intelligence Inside]]. ## Sources [Computer History Museum, ETANN and Ni1000 catalog](https://www.computerhistory.org/collections/search/?s=Ni1000); [[articles/Intelligence Inside|Intelligence Inside]].