A chip startup called Tensordyne is wagering that an unconventional approach to arithmetic could help it challenge Nvidia, the company that currently dominates the market for AI hardware. According to The Register, Tensordyne is making "a big bet on log math" — logarithmic mathematics — as the foundation of its effort to beat Nvidia.
The idea, in plain terms, is to rethink one of the most basic things a computer chip does: crunch numbers. Most AI chips today, including Nvidia's, handle the enormous volume of calculations behind modern AI using conventional number formats. A logarithmic approach represents numbers differently, which can change how efficiently certain operations — especially the multiplications that dominate AI workloads — are carried out. Tensordyne's pitch, as framed by The Register, is that this different mathematical foundation is its path to outperforming the incumbent.
The Register's report is the source for these details; the specifics of Tensordyne's hardware, performance, and timeline are described in that coverage.
Why it matters: Nvidia's grip on AI computing has made it one of the most valuable companies in the world, and challengers keep searching for an angle that lets them compete rather than copy — so a bet on rethinking the math itself is a reminder that the race for AI chips may be won on engineering ideas, not just manufacturing scale.