Updated
Updated · The Register · Jun 19
Tensordyne Bets on Log Math to Beat Nvidia's Multiply-Heavy AI Systems
Updated
Updated · The Register · Jun 19

Tensordyne Bets on Log Math to Beat Nvidia's Multiply-Heavy AI Systems

2 articles · Updated · The Register · Jun 19

Summary

  • Tensordyne is positioning logarithmic math as the core of its computing systems, arguing it can outperform Nvidia by replacing many compute-intensive multiplications with additions.
  • That approach targets one of AI hardware's biggest cost drivers: multiplication-heavy workloads that consume more compute resources than simpler arithmetic operations.
  • The bet amounts to a direct architectural challenge to Nvidia, whose dominance in AI chips has been built on scaling conventional high-performance compute for training and inference.
  • If Tensordyne can make log-based computing practical at scale, it could offer an alternative path to faster or cheaper AI processing in a market still centered on brute-force acceleration.

Insights

Can a new chip using logarithmic math truly dethrone Nvidia's GPU empire in the escalating AI arms race?
As 'shadow AI' compromises firms and biased algorithms fail asylum-seekers, who is accountable for AI's unintended consequences?
With nations embedding cyber weapons in critical infrastructure, is an all-out digital world war now inevitable?