Tensordyne Revives Logarithmic Math In A Bid To Cut AI Power Use

Article excerpt

Tensordyne is reviving a decades-old mathematical approach, logarithmic computing, to slash the energy demands of AI systems. The startup claims logarithmic math could dramatically reduce inference costs and power consumption compared to conventional chip designs, potentially offering a path around the energy crisis plaguing data centers running large language models. The technique replaces traditional binary arithmetic with logarithmic representations, reducing the transistors needed per calculation. If the approach works at scale, it could reshape how AI hardware is built, though skeptics question whether the speed-to-efficiency trade-offs will prove practical for real-world deployments.