CoreWeave's $104 Billion AI Contracts Face Financing Pressures

Before Michael Intrator became the chief executive of CoreWeave, one of America's largest specialized AI cloud companies, he was trading natural gas and managing investments in energy markets.

Today, the company he helped establish has approximately $104 billion in contracted revenue backlog, supplies computing infrastructure to major AI developers and trades publicly on Nasdaq. Yet its rapid expansion has also created billions in borrowing obligations and a growing debate over the economics of artificial intelligence.

Intrator's career illustrates an unusual transformation: from commodities trading to cryptocurrency mining, and eventually to the infrastructure powering some of the world's most advanced AI models.

Intrator began his financial career at Natsource Asset Management in 1998, working with environmental markets and energy-related investments. In 2013, he co-founded Hudson Ridge Asset Management, a hedge fund specializing in natural gas.

His background was finance rather than computer engineering, but that experience shaped his next business opportunity.

In 2017, Intrator joined Brian Venturo and Brannin McBee in establishing Atlantic Crypto, initially focused on mining cryptocurrencies with graphics processing units, or GPUs.

The operation emerged during a period of significant cryptocurrency speculation, when powerful graphics cards could generate revenue by processing blockchain transactions.

After the 2018 cryptocurrency downturn, the founders began shifting their attention toward selling computing capacity rather than relying exclusively on mining returns.

Their business became CoreWeave in 2019.

CoreWeave's transformation accelerated as demand emerged for GPU-powered rendering, machine learning and artificial intelligence.

Unlike conventional cloud providers offering broad computing services, CoreWeave concentrated on specialized GPU infrastructure.

That positioned the company to benefit when generative AI applications began requiring enormous clusters of Nvidia processors.

CoreWeave went public in March 2025, turning Intrator's earlier cryptocurrency venture into a publicly traded AI infrastructure company.

According to its second-quarter results, revenue reached $2.58 billion during the three months ended June 30, 2026, up 112% year over year.

Its revenue backlog stood at approximately $104 billion, excluding more than $25 billion in additional customer commitments announced early in the following quarter.

The business model relies on securing long-term customer agreements and financing the GPUs and data centers needed to fulfill them.

As Coinpaper's explanation of AI contracts illustrates, promised future revenue does not mean the customer has already paid the entire contracted amount.

CoreWeave's growing backlog has not eliminated its financing challenges.

The company reported approximately $35.6 billion in indebtedness as of June 30, while quarterly net interest expense reached $640 million.

Its borrowing model reflects the enormous cost of developing AI infrastructure, where equipment can be financed through GPU-backed loans and other secured arrangements.

Those pressures became more visible in October, when CoreWeave shares fell more than 22% during a week of rising bond yields.

Investors increasingly questioned whether the company's rapid revenue growth could keep pace with debt service and capital expenditure.

Meanwhile, Reuters reported on October 16 that Intrator and related entities had sold nearly $1 billion in CoreWeave shares since its IPO, predominantly through prearranged trading plans.

The transactions do not establish that Intrator expects the business to fail, but they have attracted attention during a period of heightened scrutiny of AI financing.

The same challenges extend beyond CoreWeave. Competing infrastructure companies such as Nebius and IREN are also attempting to outgrow financing costs while expanding computing capacity.

For Intrator, the transition from natural gas to AI has created a business with extraordinary contracted demand. Its next test is converting that demand into sustainable cash flow — a central question for the wider AI debt boom.