CoreWeave said its third‑quarter earnings show a new financing model. The company will ask more customers to pay for AI compute capacity up front and will focus on contracts that last a few months instead of years. By doing this, CoreWeave hopes to fund the data‑center build‑out it needs without tapping the unsecured debt market.
In the quarter, CoreWeave posted $3.66 billion in revenue, a 168 % jump from a year earlier, but it also recorded a $709 million net loss and $846 million in interest expense. CEO Michael Intrator told investors the firm will rely on loans that are backed by specific customer contracts rather than borrowing at market rates.
How the new model works
The company expects about $9 billion in customer pre‑payments and deposits over the next three months. More than $3 billion of that is tied to new contracts for capacity that is already built, so no extra hardware spending is required. CoreWeave also closed a $2.6 billion loan in August, called DDTL 5.5, which lets it draw money in installments as short‑term contracts are signed.
Short‑term deals let CoreWeave charge higher rates. It says it can sell capacity at over $40 million per megawatt per year for short contracts, versus $20‑$25 million per megawatt for four‑ to five‑year deals. Prices were raised 25 % in July and another 10 % later in the quarter, and the company says the higher rates more than offset its cost increases.
Risks and market reaction
The strategy hinges on continued strong demand for AI compute. Upfront payments lock the company into pricing that could look high now but may be less competitive if demand spikes further. Lenders still prefer longer‑term contracts because they provide predictable cash flow.
Wall Street gave the earnings a positive bump, but bond investors are wary. Yields on some CoreWeave bonds rose above 15 % last month, reflecting concerns about the debt needed for new data‑center construction before those assets generate returns.
Why it matters
For AI startups and developers, CoreWeave’s push for upfront payments could mean higher compute costs, especially if they need short‑term bursts of power. The reliance on pre‑payments also signals that the AI infrastructure market is leaning more on guaranteed cash flow from big tech customers, which may affect pricing stability for smaller players.
