In San Francisco, leading artificial intelligence companies are investing heavily in data centers essential for AI development. Tech giants like Amazon and Google, along with startups such as Anthropic and OpenAI, dominate access to this computing power, leaving others at a disadvantage. Anjney Midha, a seasoned tech entrepreneur and former partner at venture capital firm Andreessen Horowitz, seeks to change this with his startup, Amp. Based in Menlo Park, California, Amp aims to purchase surplus computing power from data centers in the U.S. and abroad, creating a global pool of specialized computer chips accessible to startups, universities, and organizations that lack the resources to train advanced AI models.
Midha highlights the challenge faced by companies unable to secure necessary computing power, as tech conglomerates monopolize infrastructure. Amp has secured over $1.3 billion from investors like Andreessen Horowitz, Y Combinator, and cloud computing providers. Notable startups, including Periodic Labs and Eleven Labs, have committed to utilizing and sharing Amp’s computing resources. This initiative is part of a broader movement to consolidate AI infrastructure, with companies like Nvidia and Mistral pooling resources for European AI development to reduce reliance on U.S. tech giants.
Midha likens Amp to an electricity grid, where computing power is shared among startups and universities. Investors provide funds for Amp to acquire computing power, which AI startups can then use to develop models. In exchange, these startups may contribute funds, data, or even the AI models themselves to the coalition. The collective bargaining power of the coalition is a significant advantage, as emphasized by Liam Fedus, CEO of Periodic Labs, who notes that joining forces with Amp enables more effective negotiations with data center operators.
