The world’s leading AI companies are spending hundreds of billions of dollars on the computing data centers needed to create AI.
Tech giants like Amazon and Google and well-funded startups like Anthropic and OpenAI are among the small group of makers with the money and connections to access all the computing power.
This leaves other organizations in the cold.
Anjney Midha, a serial tech entrepreneur who was a partner at venture capital firm Andreessen Horowitz, hopes to change that dynamic with an unusual startup. His young company, Amp, is looking to buy additional computing power from data center operators in the United States and other countries to share this valuable resource with anyone who needs it.
Amp, based in Menlo Park, California, aims to create a global pool of specialized computer chips that can be used by start-ups, universities and other organizations that would otherwise not have access to the vast amounts of computing needed to train the most powerful artificial intelligence models.
“Some companies just can’t get the computing power they need,” Mr. Midha said. “The world’s richest and most powerful companies are hoarding infrastructure for themselves.”
Amp has raised more than $1.3 billion from investors including Andreessen-Horowitz, start-up incubator Y Combinator and various cloud computing providers.
Several high-profile startups have also agreed to help use and share their pool of computing power, including Periodic Labs, an AI startup focused on scientific discovery, and Eleven Labs, which builds AI systems for voice generation.
Amp is part of a broader effort to share AI infrastructure. Chipmaker Nvidia and French start-up Mistral said this year they were sharing computing power to build artificial intelligence systems for European companies and nations, hoping to reduce their reliance on US tech giants.
Mr. Midha likened his start-up to creating an electrical grid. Just as a pool of electricity can be shared between homes and offices, a pool of computing power can be shared between start-ups and universities.
Investors contribute funds that Amp can use to purchase computing power from data center operators. AI startups then join the coalition to use the computing power to build AI models. In return, these start-ups contribute additional funds or perhaps other resources.
Some startups may share digital data needed to train AI models. Or they can share the models themselves and distribute the underlying software code to others in the coalition. Companies can even work together to train models together.
The ultimate value of the coalition is collective bargaining, said Liam Fedus, CEO of Periodic Labs. On its own, his start-up would try to get the necessary computing power. However, if Amp can negotiate with data center operators on behalf of many startups, it may gain additional leverage.
“When you aggregate your demand, you can have much more serious conversations about buying computing power,” Mr. Fedus said.
