Companies Look to Squeeze More Power Out of AI Chips
According to experts, training generative AI models requires tens of thousands of graphics chips – processing power is key for training models or producing inference from AI models. Even as tech giants and large businesses snap up the advanced computer chips that power AI, other companies are looking to new techniques to squeeze more out of the AI chips at their disposal.
In this article, Anant Adya, EVP – Infosys Cobalt, explains how there is a growing demand for strategies aimed at enhancing the performance of advanced processors, driven by the constraints imposed by limited chip supplies, which necessitate innovative approaches by companies.
“If you go down the path of training AI models from scratch, it will require a tremendous amount of [computing power],” said Anant Adya, executive vice president of IT consulting and outsourcing services firm Infosys. “Even if you manage to find compute, it will be cost prohibitive.”