CoreWeave closes $8.5 billion GPU-backed loan, first to achieve investment-grade rating

CoreWeave Secures Historic GPU-Backed Financing

CoreWeave has closed an $8.5 billion loan from a consortium of banks and investors, establishing the largest chip-backed debt deal ever completed and the first GPU-backed financing to receive an investment-grade rating 1

1

. The delayed-draw term loan facility allows the Livingston, New Jersey-based company to expand its cloud computing capacity as surging demand for AI continues to reshape the technology landscape 2

2

.

The deal is secured by graphics processing units (GPUs) and customer contracts, most notably a contract with Meta Platforms Inc. worth at least $19 billion 1

1

. Mitsubishi UFJ Financial Group Inc. and Morgan Stanley led the financing, with Goldman Sachs Group Inc. and JPMorgan Chase & Co. also participating. Blackstone Inc. served as an anchor investor in the transaction 1

1

.

Investment-Grade Rating Transforms Borrowing Costs

The term loan facility received an A3 rating from Moody's, placing it firmly in investment-grade territory and marking a watershed moment for AI infrastructure financing 3

3

. This achievement opened access to insurance firms and other institutional investors with stricter capital requirements, contributing to the deal being oversubscribed 5

5

.

The investment-grade rating delivered substantial cost savings for CoreWeave. The facility features two portions: a floating-rate tranche with a margin of 2.25 percentage points over the Secured Overnight Financing Rate, and a fixed-rate tranche financed at approximately 5.9% 1

1

. According to Nick Robbins, vice president of corporate development, the new loan's borrowing costs are roughly 7.5 percentage points cheaper than CoreWeave's first GPU loan in 2023 1

1

. This stands in stark contrast to the company's junk bonds sold last year, which currently trade with yields around 10% 1

1

.

Aggressive Expansion Amid AI Infrastructure Boom

CoreWeave can initially borrow up to $7.5 billion through the delayed-draw structure, with the capacity expanding to $8.5 billion once the chips are operational 1

1

. This brings total equity and debt financing commitments secured by the company over the past 12 months to approximately $28 billion 2

2

. The company expects 2026 capital expenditures of $30-$35 billion, more than doubling 2025 levels 3

3

.

"There is just so much demand for this infrastructure, it's truly not stopping," said Brannin McBee, chief development officer and CoreWeave co-founder. "This type of financing that we are doing is the most sophisticated and scalable way to finance the build-out of artificial intelligence infrastructure" 1

1

.

The company had already accumulated a $21.6 billion debt pile as of the end of last year, along with an additional $3.7 billion of untapped borrowing capacity 1

1

. This aggressive, debt-funded expansion reflects the broader industry trend, with conservative estimates pegging AI spending at $3 trillion or more 1

1

.

Market Response and Performance Benchmarks

Investor confidence surged following the announcement, with CoreWeave's stock jumping 12% on the last trading day 3

3

. The company, which completed its public listing on Nasdaq in March 2025, now carries a market capitalization of $40.7 billion 4

4

.

CoreWeave recently reported results from the MLPerf Inference v6.0 benchmark suite using NVIDIA GB200 NVL72 and GB300 NVL72 infrastructure 4

4

. The company's GB200 NVL72 configuration led performance for DeepSeek-R1 in server and offline modes, while the GB300 NVL72 system delivered results that were twice CoreWeave's MLPerf 5.1 results on the same hardware footprint 4

4

. Eight of the top 10 model providers now use the CoreWeave AI cloud platform 4

4

.

Risks and Strategic Considerations

Despite impressive revenue growth of 168% over the last twelve months, CoreWeave faces several headwinds 4

4

. The company operates with significant debt and is burning through cash rapidly, with long-term debt reaching $14.7 million as of December 31, 2025, compared to $5.5 million a year earlier 3

3

.

Customer concentration among a few large AI and hyperscale clients, including Meta Platforms Inc., creates dependency risks 3

3

. The company also relies heavily on NVIDIA for GPU supply and faces execution risks in scaling data centers, including power constraints, construction delays, and operational complexity 3

3

. The timing gap between upfront capital expenditures and gradual revenues from new capacity will likely pressure near-term profits 3

3

.

CoreWeave specializes in cloud infrastructure for compute-intensive workloads, operating a GPU-first architecture optimized for training and inference of generative AI models 5

5

. With its own data centers located in the United States and Europe, the company maintains full control over its infrastructure, enabling high-performance computing (HPC) capabilities and low latency 5

5

. The loan matures in March 2032, providing long-term capital for the company's ambitious expansion plans 1

1

.