Startup Aims to Optimize Hardware to Address GPU Shortage for Generative AI

Are you tired of hearing that AI development is hindered by a shortage of GPUs? Well, CentML has emerged from stealth to help solve this issue. The Toronto-based AI startup optimizes machine learning models for faster performance and lower compute costs. And it’s doing more than that: it’s increasing access to AI compute, which is considered a massive obstacle in AI development. Keep reading to learn more about how CentML is revolutionizing the AI industry.

Hard for Smaller Companies to Access GPUs

CentML raised $3.5 million in a 2022 seed round, which was led by Radical Ventures. The company’s co-founder and CEO, Gennady Pekhimenko, told VentureBeat that access to the hardware and software stack for language models is crucial. He even joked that he invested in Nvidia, acknowledging the company’s grip on the GPU market. But Nvidia primarily sells its most expensive chips, making it difficult for smaller companies to access affordable compute. CentML solves this issue by optimizing less expensive chips that aren’t fully utilized. By doing so, CentML is serving a larger market.

Race to Access AI Chips Has Become Like Game of Thrones

David Katz, partner at Radical Ventures, likens the competition for access to AI chips to Game of Thrones, minus the blood. He also acknowledges the insatiable demand for compute needed to run large models, creating a bottleneck in access. CentML’s solution “creates a little bit more efficiency” in the market and demonstrates that legacy hardware can run complex billion-plus-parameter models. And the best part? It’s increasing the supply of chips in the market.

In conclusion, CentML is not only optimizing machine learning and lowering compute costs, but it’s also increasing access to AI compute. Imagine the possibilities for AI development when compute is no longer a roadblock. Stay informed on the latest AI developments by following VentureBeat’s ongoing generative AI coverage.

Leave a comment

Your email address will not be published. Required fields are marked *