CoreWeave Raises Yield on $2.6 Billion Loan to Attract Investors
Miles Bennett
CoreWeave raised the spread on a $2.6 billion loan by up to 1.25 percentage points to 5.5 points over the benchmark, signaling that investor appetite for AI-infrastructure debt is cooling systemically.
What changed in the loan terms?
The spread rose by up to 1.25 percentage points, reaching benchmark + 5.5 points.
The issue price dropped from 99 cents to 97 cents on the dollar. This means → buyers get a bigger discount upfront, pushing the effective yield even higher.
CoreWeave also tightened the loan documents with stronger creditor protections. JP Morgan is leading the deal; investors submit orders by Thursday.
Why sweeten the terms mid-deal?
CoreWeave's five-year credit-default swap — a market gauge of perceived default risk — has surged more than 50% this month, hitting the highest level since December.
This reflects broad unease over whether the AI-infrastructure borrowing wave is sustainable. Investors now demand higher returns before committing.
In plain terms = the last round of AI-themed debt sold on excitement alone; this round needs a price hike.
Where does the borrowed money go?
The loan is a delayed-draw term facility — drawn in tranches after signing — earmarked for purchasing and installing GPUs and related components.
Clients include Anthropic, Jane Street Group, and Hudson River Trading, under take-or-pay contracts (payment is due whether capacity is used or not).
This means → CoreWeave locks in revenue, but the upfront hardware spend is entirely debt-funded.
Why do high-frequency traders need cloud compute?
Jane Street and Hudson River Trading rely on cloud providers for the compute that powers trading algorithms capable of reacting to market moves in microseconds.
In plain terms = more compute = faster data processing + faster trade execution. In high-frequency trading, one microsecond faster is real money.
Is the AI-infrastructure financing tide turning?
CoreWeave expects to spend more than $34 billion on AI infrastructure this year, leaning heavily on the junk-debt market.
Earlier, the company raised $3.1 billion through a first-of-its-kind GPU-backed deal that drew $19 billion in orders — demand was overwhelming.
This signals a pivot: from "oversubscribed" to "must sweeten to sell." The pricing floor for AI-infrastructure financing is shifting structurally higher.
Content is for reference only, not financial advice.