CoreWeave's $21B Meta Deal Boosts Backlog to $66.8B, Debt at $21B

CoreWeave secured a $21 billion expanded agreement with Meta Platforms through December 2032, lifting its total backlog to $66.8 billion and bringing cumulative Meta commitments to $35.2 billion. The deal, which includes early deployments of Nvidia's Vera Rubin platform, positions the neocloud provider as a critical infrastructure partner for Meta's AI ambitions, including its Muse Spark model. Yet the headline win masks a precarious balance sheet: CoreWeave carries $21 billion in long-term debt with an average interest rate of approximately 11%, consuming roughly 25% of top-line revenue in interest payments. The company reported a net loss of $1.2 billion in 2025 and $740 million in the first quarter of 2026 alone, even as revenue surged 168% to $5.1 billion last year. With capital expenditures of $7.7 billion in Q1 2026 and a planned $30-$35 billion for the full year, CoreWeave is spending $2.60 in capex for every dollar of new revenue. The company has raised over $20 billion in capital in 2026 through a mix of term loans, convertible notes, and Nvidia's $2 billion investment. This debt-fueled growth model works as long as hyperscaler demand keeps accelerating. The moment any major customer brings AI capacity in-house or cuts spending, the math unravels.

The $21 Billion Debt Stack and Its Cost

CoreWeave's debt load nearly tripled to $21 billion as of the Meta deal announcement, though it stood at $17.3 billion by the end of the first quarter of 2026. The company has raised over $20 billion in capital this year alone, including an $8.5 billion term loan, a $3.1 billion GPU-backed facility, and Nvidia's $2 billion equity investment. In conjunction with the Meta deal, CoreWeave announced a $3.0 billion convertible senior notes offering due 2032, with an option for an additional $450 million, plus a separate $1.25 billion senior unsecured notes offering due 2031. These offerings total $4.25 billion in fresh debt. The average interest rate of 11% means CoreWeave pays roughly $2.3 billion in annual interest, consuming a quarter of its $5.1 billion in 2025 revenue. Interest expense in Q1 2026 alone was $536 million, exceeding the company's $740 million net loss for the quarter. The convertible debt strategy dilutes existing shareholders but provides cheaper capital than the 11% term loans. CEO Michael Intrator is betting that revenue growth, guided above $12 billion for 2026, will eventually outpace interest costs. But with $7.7 billion in quarterly capex and negative free cash flow, the company remains dependent on capital markets to service its debt. The $21 billion figure includes both the new convertible notes and the existing term loan facilities, meaning CoreWeave's interest burden will increase further once the latest offerings are fully drawn. The leverage ratio underscores the fragility. CoreWeave reported $50.8 billion in total liabilities against $4.8 billion in shareholders equity at the end of 2025, a ratio above 10x and well outside the 3x-5x range typical for asset-heavy infrastructure operators. Interest cover, measured as operating income divided by interest expense, is negative because the company has not yet posted operating profit. By contrast, AWS reached positive operating margins within five years of launch and never carried leverage above 1x. The structural difference is timing: CoreWeave must build capacity for contracts that pay out across five to seven years, but the debt to fund that buildout matures inside three years for the term loans and seven years for the convertibles. That mismatch forces the company back to capital markets every twelve to eighteen months. Jane Street, the proprietary trading firm that recently signed a multi-year capacity contract with CoreWeave, joins Meta and Microsoft as a top-five customer, meaning three customers now drive the bulk of contracted revenue. Lenders price that concentration into the 11% coupon. Any rating downgrade from Moodys or S&P would push the next refinancing cost above 13% and accelerate the cash burn that already runs at roughly $3 billion per quarter.

How the $66.8 Billion Backlog Flows Through Revenue

CoreWeave's backlog of $66.8 billion, which later grew to $99.4 billion by March 31, 2026, represents contracted future revenue from customers including Meta, Microsoft, and OpenAI. The $21 billion Meta deal runs from 2027 through 2032, on top of a prior $14.2 billion arrangement through 2031. Microsoft remains the dominant customer, responsible for 70% of CoreWeave's revenue, while OpenAI accounts for over $20 billion of the backlog. The company guided for more than $12 billion in revenue for 2026, implying a 135% increase from 2025's $5.1 billion. However, converting backlog to cash requires massive upfront capital expenditure: CoreWeave spent $7.7 billion in Q1 2026 alone and plans $30-$35 billion for the full year. The capital intensity ratio of $2.60 in capex for every $1 of revenue means the company must raise external funding to fulfill existing contracts. Each new dollar of revenue requires nearly three dollars of investment in Nvidia GPUs, data center power, and networking gear. The backlog acts as a financing tool: CoreWeave uses these contracts to secure debt facilities and GPU-backed loans from banks like Goldman Sachs. But if any major customer defaults or renegotiates, the collateral backing that debt loses value. The $99.4 billion backlog figure, disclosed after the Meta deal closed, includes additional commitments from existing customers that were not previously counted.

The Competitive Reshuffle Among Neoclouds and Hyperscalers

The Meta deal solidifies CoreWeave's position as the leading neocloud provider, pulling away from competitors like Google and smaller players such as Intel and Shield AI in the AI infrastructure race. Meta's decision to double down on CoreWeave, despite building its own data centers, signals that even hyperscalers need external capacity to meet AI demand. Meta plans $115-$135 billion in capex for 2026, nearly double 2025 levels, yet still outsources a portion to CoreWeave. This creates a wedge for neoclouds: they offer faster deployment and specialized GPU clusters that hyperscalers' internal teams cannot match on short timelines. Nvidia's $2 billion investment in CoreWeave and its role as primary GPU supplier, including the Vera Rubin platform, gives CoreWeave preferential access to supply that rivals like Anthropic and Google cannot replicate. However, customer concentration remains extreme: Microsoft at 70% of revenue and OpenAI at over $20 billion of backlog means CoreWeave's fate is tied to two companies. If Microsoft shifts AI capacity to its own Azure infrastructure or OpenAI brings compute in-house through its Guaranteed Capacity offering, CoreWeave loses its revenue base. The company's stock fell 7.84% to $95.63 on May 19, reflecting investor anxiety about this concentration risk. Smaller neoclouds like Lambda and Vultr have struggled to secure similar financing, widening the gap between CoreWeave and the rest of the market.

Downstream Effects on Nvidia, Fabs, and Enterprise Buyers

CoreWeave's $30-$35 billion in planned 2026 capex flows directly to Nvidia, which supplies GPUs including the Vera Rubin platform deployed under the Meta deal. This creates a virtuous cycle for Nvidia: it invests $2 billion in CoreWeave, receives orders for billions in GPUs, and collects revenue from both the hardware sale and ongoing software licensing. The arrangement also pressures TSMC and other fabs to expand advanced packaging capacity for Nvidia's chips, as CoreWeave's demand compounds hyperscaler orders. For enterprise buyers, CoreWeave's debt burden raises the cost of AI cloud services: the company must charge high margins to service its 11% interest payments, making its GPU clusters more expensive than those offered by cash-rich hyperscalers like Google and Microsoft. OpenAI, which expects to burn $115 billion through 2029 before turning a profit, faces particularly acute risk: if it cuts spending, CoreWeave loses over $20 billion in backlog. The capital intensity also affects power markets: CoreWeave's active power crossed 1 gigawatt, with a target of over 8 gigawatts by 2030, straining grid capacity in data center hubs like Northern Virginia and driving up electricity costs for other users. Regulators in states like Virginia and Georgia are already scrutinizing data center power procurement, and CoreWeave's rapid expansion will intensify those debates. The company's power procurement agreements now span multiple regions, each requiring separate grid interconnection approvals that can take years to secure.

What the Deal Signals About AI Infrastructure Strategy

The Meta-CoreWeave deal represents a structural bet that AI compute demand will remain elastic and that outsourcing to specialized providers beats building everything in-house. Meta's $115-$135 billion capex plan for 2026, nearly double 2025, shows that even the largest hyperscalers cannot internalize all AI infrastructure needs. CoreWeave's CEO Mike Intrator positions the company as the "AWS for AI," but the comparison falters on capital efficiency: Amazon Web Services generated positive free cash flow within years of launch, while CoreWeave burns cash faster than it grows. The $3 billion convertible debt offering, combined with $1.25 billion in senior unsecured notes, signals that CoreWeave expects interest rates to remain elevated and prefers equity-linked dilution over floating-rate debt. Nvidia's dual role as investor, supplier, and customer creates a conflict of interest: it profits from CoreWeave's GPU purchases regardless of whether the neocloud succeeds as a business. The broader market reads this deal as a validation of the neocloud model: CoreWeave shares rose 3.5% and Meta shares gained 2.6% on the announcement. But the sustainability depends on continued hyperscaler demand and stable capital markets. If interest rates rise or AI spending cycles turn, CoreWeave's $21 billion debt stack becomes a trap.

The next 12 months will test whether CoreWeave's growth can outrun its debt costs. Revenue guidance above $12 billion for 2026 implies the company needs to nearly triple 2025's top line just to cover interest payments and operating losses. The $99.4 billion backlog provides visibility, but converting those contracts to cash requires another $30-$35 billion in capex this year, which CoreWeave will fund through additional debt and equity offerings. CEO Michael Intrator's target of over 8 gigawatts of active power by 2030 means the company's capital intensity will not ease soon. The most likely exit scenario is acquisition by a hyperscaler: Microsoft, which already accounts for 70% of revenue, could absorb CoreWeave to secure its GPU supply chain and eliminate a dependency on a debt-laden partner. Alternatively, a downturn in AI demand would trigger a forced restructuring of CoreWeave's $21 billion debt stack, wiping out equity holders and converting bondholders into the new owners. For now, the Meta deal buys time and credibility, but it does not solve the fundamental math of spending $2.60 to earn $1.00.