CoreWeave and Anthropic Multi-Year Cloud Infrastructure Partnership

On April 9, Meta committed an additional $21 billion to CoreWeave, expanding their relationship to approximately $35 billion through 2032. Less than 24 hours later, Anthropic signed a $6.8 billion multi-year deal. Together, these two announcements added roughly $28 billion in contracted revenue to CoreWeave’s books in a single 48-hour window — pushing its total contracted backlog past $66.8 billion. For a company that generated $5.13 billion in full-year 2025 revenue, this backlog represents over a decade of revenue visibility at current run rates.

The timing was almost certainly deliberate. Landing the Meta expansion first gave CoreWeave momentum heading into the Anthropic announcement, creating a narrative of unstoppable demand that drove shares up 11-13% to around $102 — more than doubling from its $40 IPO price just one year earlier. With OpenAI’s existing $22.4 billion commitment, CoreWeave now holds massive long-term contracts with three of the most important AI companies in the world. The company’s 2026 revenue guidance of $12-13 billion would represent another 130-150% year-over-year growth, suggesting these contracts are translating into real near-term revenue, not just paper promises.

Behind the headline-grabbing contract wins sits a balance sheet that would make traditional infrastructure investors uneasy. CoreWeave carries $21 billion in long-term debt, plans $30-35 billion in capital expenditures, and simultaneously announced a $3.5 billion convertible debt offering alongside the Anthropic deal. The company is not yet profitable. Its 47.77% gross margin is healthy for an infrastructure business, but the sheer scale of capital deployment required to build out GPU clusters, secure power, and construct data centers means profitability remains a future milestone rather than a present reality.

Motley Fool analyst David Jagielski captured the tension: the long-term upside is real, but only for investors willing to stomach the risk. CoreWeave’s model essentially involves borrowing billions to build infrastructure that is then locked into multi-year contracts. If AI demand continues its current trajectory, this works spectacularly — the contracted backlog dwarfs the debt. But the strategy leaves almost no margin for error. A slowdown in AI spending, a shift in customer preferences, or a sustained rise in interest rates could turn the leverage from an accelerant into an anchor. The company’s market capitalization of approximately $54 billion reflects optimism, but the debt-to-equity ratio tells a more complicated story.

Anthropic’s decision to commit $6.8 billion to CoreWeave rather than build proprietary infrastructure reveals a strategic calculation about where an AI company should allocate capital. With its annual run rate surging from $9 billion to over $30 billion in roughly four months, Anthropic faces enormous and rapidly growing compute demands. Building data centers from scratch would require years of lead time, massive upfront capital, and operational expertise that is peripheral to Anthropic’s core mission of building safe AI systems. CoreWeave offers a shortcut: specialized GPU clusters, already operational or coming online later in 2026, without the capital expenditure burden.

The technical argument is equally compelling. Anthropic expects a 20-30% performance improvement in inference workloads by using CoreWeave’s bare-metal architecture compared to traditional virtualized cloud environments offered by hyperscalers like AWS and Azure. For a company serving Claude at production scale to a rapidly growing customer base, that performance gap translates directly into cost savings and better user experience. The bare-metal approach eliminates the overhead of virtualization layers, giving Anthropic’s models more direct access to GPU compute. This positions CoreWeave not as a commodity cloud provider but as a performance-differentiated platform — a distinction that matters enormously when inference costs are one of the largest line items for any AI model provider.

CoreWeave’s origin story adds an improbable layer to its current dominance. The company began as a cryptocurrency mining operation before pivoting to AI infrastructure around 2019. That pivot, driven by the recognition that GPU compute had applications far beyond hashing algorithms, turned out to be one of the most prescient strategic bets in recent tech history. The same GPU hardware that powered crypto mining could be repurposed and expanded for AI training and inference — and the operational expertise in managing large-scale GPU deployments transferred directly.

The crypto connection resurfaces in an unexpected way. Market analyst Ran Neuner pointed out that AI and crypto mining now compete for the same fundamental resource: electricity. CoreWeave operates over 250,000 GPUs across 32+ data centers with 1.3 gigawatts of contracted power. As AI companies demonstrate willingness to pay premium prices for energy — locking in multi-year, multi-billion-dollar commitments — crypto miners face increasing pressure. The irony is stark: the industry that originally justified CoreWeave’s GPU purchases is now being outbid for power by the industry CoreWeave pivoted toward. With Microsoft accounting for 67% of 2025 revenue and that concentration now rapidly diversifying through the Anthropic, Meta, and OpenAI deals, CoreWeave has completed its transformation from a niche crypto operation to the infrastructure layer that nine of the ten leading AI model providers depend on.