Amp Raises $1.3 Billion for Shared AI Compute Pool
Leading artificial intelligence firms spend hundreds of billions of dollars on data centers to power their systems. Companies such as Amazon, Google, Anthropic, and OpenAI hold the resources and relationships needed to secure massive computing capacity. Smaller groups often lack such access.
Anjney Midha, a serial technology entrepreneur and former partner at venture capital firm Andreessen Horowitz, started Amp to address this issue. His company, located in Menlo Park, California, purchases surplus computing power from data center operators across the United States and other nations. Amp then distributes this resource to those in need.
Amp's Goal: A Pool for Excluded Organizations
Amp seeks to assemble a worldwide collection of advanced computer chips tailored for AI tasks. These chips support the intense computing demands of training top-tier AI models. Startups, universities, and similar entities gain entry to this pool, bypassing barriers faced when dealing directly with dominant players.
"Some companies just can't get the computing power they need," Mr. Midha said. "The world's wealthiest and most powerful companies are hoarding the infrastructure for themselves."
The startup has attracted over $1.3 billion in funding. Backers include Andreessen Horowitz, the accelerator Y Combinator, and several cloud computing providers. Andreessen Horowitz, often called a16z, invests in early-stage tech companies and has backed numerous AI ventures. Y Combinator supports promising founders through its batch programs, helping them scale quickly.
Key Partners Join the Coalition
Prominent startups have committed to using and contributing to Amp's computing pool. Periodic Labs focuses on AI applications for scientific breakthroughs. ElevenLabs develops systems that generate realistic voices using artificial intelligence.
Stay ahead of the AI curve
The most important updates, news, and content — delivered weekly.
No spam. Unsubscribe anytime.
This initiative fits into larger trends of shared AI resources. For instance, chip giant Nvidia partnered with French startup Mistral earlier this year. Together, they offer pooled computing for European businesses and governments, aiming to lessen reliance on American tech leaders. Nvidia dominates the market for GPUs essential to AI training, while Mistral builds open-source large language models competitive with those from bigger rivals.
Model Inspired by Electricity Grids
Mr. Midha draws a parallel between Amp and an electricity grid. Electricity providers generate power centrally, then distribute it to homes and businesses via a shared network. Similarly, Amp aggregates computing capacity for broad use among AI developers.
Investors supply capital for Amp to acquire power from data centers. Participating AI startups access this capacity for model training. In exchange, they provide funds, data for training, finished models, or even collaborate on joint projects.
Benefits of Collective Negotiations
The coalition's strength lies in group purchasing power, according to Liam Fedus, chief executive of Periodic Labs. A single small startup faces challenges securing enough compute on its own. Amp, representing many, negotiates better terms with suppliers.
"When you pool your demand, you can have far more serious conversation about buying computing power," Mr. Fedus said.
Amp operates amid a computing shortage driven by exploding AI demand. Training advanced models requires thousands of high-end chips running for weeks or months. Tech giants secure long-term contracts, leaving scraps for others. Menlo Park, in Silicon Valley, hosts many such innovators, including nearby firms like OpenAI.
This funding round positions Amp to scale rapidly. By May 12, 2026, it had already lined up major supporters, signaling strong industry interest in democratizing AI infrastructure.