The tightening costs of the DDTL 5.5 facility
The $2.6 billion delayed draw term loan facility, known as DDTL 5.5, which CoreWeave closed around August 10, 2026, shows a significant shift in the cost of capital for AI infrastructure. While the company secured a $2.3 billion debt financing facility led by Magnetar Capital and Blackstone in August 2023, the newer 2026 transactions demand much higher premiums. Lenders pushed the interest spread for the DDTL 5.5 to SOFR plus 550 basis points. This increase follows a July 2026 attempt to launch the same facility with a spread between SOFR plus 425 and 450 basis points. The market required a wider spread and a discount of roughly 96 cents on the dollar to finalize the deal.
These rising costs reflect a changing sentiment among the lenders who provide liquidity to the neocloud sector. Earlier in the year, CoreWeave closed an $8.5 billion DDTL 4.0 facility on March 30, 2026, which featured a lower pricing of SOFR plus 2.25%. The transition from a 2.25% spread to a 5.50% spread demonstrates that the credit market now prices the risk of GPU-backed debt more aggressively. Even though the July 2026 transaction reached oversubscription, the demand appeared only after CoreWeave accepted these higher interest rates and more restrictive covenants. Lenders such as JPMorgan and Mitsubishi UFJ Financial Group, who served as joint lead arrangers for the DDTL 5.5, face a different risk profile now than they did during the initial 2023 financing rounds.
Widening spreads and the cash lockbox
The new debt terms impose heavy restrictions on how CoreWeave manages its incoming revenue. To secure the $2.6 billion, CoreWeave agreed to a cash lockbox arrangement. This structure requires that the cash flowing in from customer contracts goes toward paying down the debt before the company can access those funds. Lenders also demanded a 1.35x debt-service-coverage covenant and full amortization. These requirements highlight a lack of trust in the company’s ability to manage its own cash flow without direct lender oversight.
The market’s reaction to CoreWeave’s creditworthiness remains divided. In late July 2026, the cost to insure CoreWeave’s debt against default, measured by its credit default swap, topped 855 basis points. This level implied a 50% probability of default within five years. While the stock price recovered from a low of $60 to settle in the high $80s after the loan closed, the credit market does not share the same optimism. The bond market writes the risk into the contract through higher yields, whereas the equity market often ignores these signals in favor of growth narratives.
| Facility Type | Interest Rate / Spread | Maturity / Term | Key Condition |
|---|---|---|---|
| DDTL 3.0 | SOFR + 4.00% | Asset-based | Asset-based limits |
| DDTL 4.0 | SOFR + 2.25% | Execution-based | Power access terms |
| DDTL 5.5 | SOFR + 5.50% | 5-year | Cash lockbox / 1.35x coverage |
Concentration of revenue and unrated counterparties
CoreWeave relies on a very small number of clients to support its massive debt load. Fitch noted that the company drew 65% of its first-quarter 2026 revenue from just two customers. The DDTL 5.5 facility specifically depends on revenue from Anthropic and Jane Street. Anthropic, a privately held AI lab, provides roughly 40% of the revenue for this facility. Jane Street, a quantitative trading firm, provides an additional 35%. This means 75% of the revenue backing this specific loan comes from two companies.
This concentration creates a massive risk for lenders because Anthropic lacks a public credit rating. Lenders cannot easily price the credit risk of a counterparty they cannot formally rate. If Anthropic or Jane Street alters its spending or faces financial instability, the entire debt structure faces pressure. The company attempts to mitigate this by diversifying its customer base, noting that no single customer represents more than 35% of its forward backlog. However, the current reliance on a few high-profile AI labs remains a structural vulnerability.
Duration mismatches in AI infrastructure
A fundamental mismatch exists between the lifespan of CoreWeave’s debt and its revenue streams. The DDTL 5.5 facility carries an approximate five-year maturity. In contrast, the customer contracts that secure this debt average only three years in length. CoreWeave also signs data center leases with landlords for terms that reach up to 15 years. This creates a gap where the company must fund long-term physical infrastructure with relatively short-term revenue.
If these three-year contracts do not renew, the company must find new ways to service the five-year debt. This pattern mirrors the asset-liability mismatch that affected other high-growth companies like WeWork. CoreWeave assumes it can renew contracts or re-lease capacity to other customers at the end of the initial terms. This assumption relies on a constant, uninterrupted demand for GPU compute. If the AI boom slows or if customers move to different providers, the company will face a massive funding gap when the debt matures.
The industry debate over GPU useful life
The viability of CoreWeave’s financing model depends on how long its hardware remains useful. CoreWeave uses a six-year useful life for its GPUs in its accounting and depreciation schedules. This differs sharply from other industry players. Nebius uses a four-year schedule, and Scaleway assumes a three-year window because of the 18-month cycle of new GPU generations. When companies disagree this much on the value of their core assets, it suggests that the depreciation math remains a point of contention.
Amazon provides a cautionary example of this shift. AWS extended its server useful life from three to six years over several years, but it reversed course in early 2025. The company moved some equipment back to a five-year schedule due to the rapid pace of AI technology development. If the industry moves toward shorter depreciation timelines, CoreWeave’s reported earnings and asset values could face significant downward pressure. Using a six-year horizon for hardware that might become obsolete in three years creates a risk of sudden write-downs.
Interest coverage and the EBITDA gap
The company’s interest burden creates a heavy drag on its GAAP profitability. In the third quarter of 2025, interest expenses reached $311 million. This amount constituted 23% of the $1.36 billion in revenue recorded during that period. Because the company reported GAAP operating income of only $51.9 million, its interest coverage ratio stood at 0.17. This means the operating income covers only 17 cents for every dollar of interest the company owes.
Bulls point to the adjusted EBITDA of $838 million and a 61% margin to suggest the company remains healthy. This figure provides an interest coverage ratio of 2.7x. However, the gap between the 0.17 GAAP figure and the 2.7x EBITDA figure stems from the $630.5 million in depreciation and amortization charges. If the real economic life of the GPUs is shorter than six years, the depreciation being added back to EBITDA will fail to reflect the true cost of maintaining the fleet. The company’s ability to pay its bills depends entirely on whether its accounting assumptions regarding hardware longevity hold true.
Magnetar Capital’s strategic hedging
Magnetar Capital, a major stakeholder in CoreWeave, uses derivatives to manage the risk of its $4.9 billion position. As of December 31, 2025, Magnetar held 68.2 million shares of Class A common stock. By the third quarter of 2025, this position represented 72% of Magnetar’s reported 13F portfolio. To protect against a potential crash, the fund entered a collar trade in September 2025. This trade included buying 392,632 put options at a $70 strike price and selling calls at strikes of $160 and $175.
The $70 strike price for the puts sits well below the stock price, which traded around $100 during that period. This indicates that Magnetar is paying for protection against a significant crash, perhaps a 30% to 50% decline. By using a collar, Magnetar caps its upside potential in exchange for a floor on its losses. The fund spent more on puts than it received from selling calls, meaning it voluntarily accepted a net debit to secure downside protection. This strategy suggests that the largest investor in the company expects significant volatility and is preparing for a scenario where the stock price drops sharply.
The 2026 debt amortization wall
CoreWeave faces a significant wave of payments in the coming year. The company must handle approximately $4.2 billion in scheduled debt amortization and vendor financing payments in 2026. While the company’s CFO stated that no lump-sum repayments exist until 2028, the cash must still exit the company to meet these self-amortizing obligations. The company expects capital expenditure in 2026 to exceed 2025 levels by more than double, adding further pressure to its cash reserves.
The company relies on its massive revenue growth and a $55.6 billion backlog to meet these obligations. Revenue grew 134% year-over-year in the third quarter of 2025, and the company has secured multi-billion dollar agreements with partners like Meta. However, contracted revenue does not always translate to collected cash. As CoreWeave continues to expand its data center footprint in places like Plano, Texas, it must balance the need for massive capital investment with the necessity of meeting its growing debt service requirements. Will the company’s rapid expansion generate enough liquidity to clear the $4.2 billion hurdle in 2026?
