Tensordyne bets on math to beat Nvidia

Startup relies on logartihmic math to replace multiplication functions with addition wherever possible, and aims to cut power and gain room for memory.

July 20, 2026

Jon Peddie

Tensordyne wants to beat Nvidia at inference by changing the math itself. The Sunnyvale startup, formerly known as Recogni, just taped out Napier, a 3nm chip that replaces most multiplication functions with addition to cut power and free up chip space for memory. The claims are bold: 13 times the throughput of Nvidia’s Blackwell rack, at a fraction of the power. We dig into the architecture, the company behind it, and what’s confirmed versus what’s projection for this YANK[i] [i] YANK – Yet Another Nvidia Killer Tensordyne completed tape-out of its Napier AI inference chip in June. Built on TSMC’s

...

Enjoy full access with a TechWatch subscription!

TechWatch is the front line of JPR information gathering service, comprising current stories of interest to the graphics industry spanning the core areas of graphics hardware and software, workstations, gaming, and design.

A subscription to TechWatch includes 4 hours of consulting time to be used over the course of the subscription.

Already a subscriber? Login below

Tensordyne bets on math to beat Nvidia

Enjoy full access with a TechWatch subscription!

This content is restricted

Login