[Wafer: response was truncated before the model finished its internal reasoning. Increase max_tokens, or disable thinking on this model (e.g. chat_template_kwargs.enable_thinking=false), then retry.]

[Wafer: response was truncated before the model finished its internal reasoning. Increase max_tokens, or disable thinking on this model (e.g. chat_template_kwargs.enable_thinking=false), then retry.]

Published on: Aug 09, 2026
[Wafer: response was truncated before the model finished its internal reasoning. Increase max_tokens, or disable thinking on this model (e.g. chat_template_kwargs.enable_thinking=false), then retry.]

CoreWeave leases most of its data-center capacity instead of owning it, and the cost of that strategy is now quantifiable. Capitalizing the company's about $165 per kW per month rent at a 9.5% stabilized yield implies the landlord has deployed roughly $20.8 million per MW - about 2.5 times the facility cost of efficient owners such as xAI.

CoreWeave rents powered shells from partners including Core Scientific, Applied Digital, and Galaxy/Helios, then installs its own GPUs and networking. On its own balance sheet, the company mainly shows GPU capital expenditures. The leased facilities stay off the books, shifting more of the cost structure into operating expenses rather than capital expenses.

The rent math

At $165 per kW per month, one megawatt costs $165,000 per month, or $1.98 million per year. Capitalizing that annual rent at a 9.5% stabilized yield produces an implied capital value of about $20.8 million per MW. That is the amount the landlord effectively deployed to generate the rent stream, or the market value of the leased facility.

Per gigawatt, that works out to $20.8 billion. For comparison, typical AI-ready powered-shell construction costs for owned facilities run $10-15 million per MW. First-principles builds that repurpose existing buildings and use behind-the-meter electricity, such as xAI's Colossus sites, have been reported as low as $2.7-8 million per MW for the facility portion.

At $20.8 million versus roughly $8 million, CoreWeave's effective facility cost is about 2.5 times the cheapest owned builds. Among its peer set, it is among the highest, not the lowest.

What leasing does to the financials

Leasing moves cost from the balance sheet to the income statement. An owner books a large CapEx number and depreciates it over many years, since buildings last far longer than GPUs. CoreWeave keeps most of that CapEx off its books; rent hits operating expenses every month instead.

The result is that CoreWeave's reported CapEx looks low relative to revenue growth and capacity added, making the business appear less capital-hungry than a fully owned model. But the trade-off is a permanent drag on operating margin. Rent is a recurring cash cost that never goes away, while an owner eventually finishes depreciating the building and keeps more of the economics.

CoreWeave's long-term adjusted operating-margin target of 25-30% already embeds this landlord tax. Contribution margins stabilize in the mid-20s once deployments mature, partly because of these lease costs.

The ownership pivot

CoreWeave has acknowledged the issue. It has begun selective self-builds and has explored ownership routes, including past discussions around Core Scientific, to reduce reliance on third-party landlords over time.

The comparison is deliberately apples-to-apples. Leasing accelerates scale and keeps GPU CapEx as the dominant line item, but it embeds a higher effective facility cost and a permanent margin drag relative to vertically integrated or ultra-efficient owners such as xAI.


Get Daily AI News

Your membership also unlocks:

700+ AI Courses
700+ Certifications
Personalized AI Learning Plan
6500+ AI Tools (no Ads)
Daily AI News by job industry (no Ads)