Vaire's Reversible Computing Chip Could Solve The Biggest Problem in AI Hardware - Energy Consumption

Founded by Rodolfo Rosini(CEO), Dr. Hannah Earley(CTO), & Andrew Sloss, Vaire's Reversible Computing could solve the biggest challenge in AI Compute, i.e., Energy Consumption. Their initial test showed around a 50% reduction in the Energy Consumption. This single solution can tackle 3 big challenges in running the data centres viz water consumption for cooling, CO2 emission, & an increasing requirement of energy.

Funded by 7percent Ventures, Lifeline Ventures, SeedCamp, and Jude Gomila, Vaire is betting big on developing reversible computing.

Why Reversible Computing? Most of the heat in any circuit is generated when prior information is overwritten with new one. If this information can be reversed to its original state, this could save tons of energy required for massive parallel computing. This was an extremely difficult engineering challenge to implement on the existing CMOS chips.

Up until now, this could be solved by shrinking the transistor density on a chip, but shrinking any further is not a viable solution for modern compute power requirements. Even increasing the transistor cycles is not a solution, as it produces too much heat, and moving to faster cycles could alter the behaviour of the chip itself.

Designing Reversible Computing: Reversible computing is originally slower than traditional computing, but combining this with GPU design i.e., parallel processors, can be promising.