CoreWeave says it has signed a multiyear deal with Anthropic, including a variety of Nvidia chips at data centers in the US; it now has 43 active data centers
AnthropicPBC agreed to rent data center capacity from CoreWeave Inc. as part of efforts to handle increasing demand for its artificial intelligence services.
BloombergBrody Ford
Context & Ripple Effects
CoreWeave’s Anthropic agreement extends a customer roster already built through large capacity commitments with OpenAI and Meta. It also comes after Nvidia agreed to backstop unsold CoreWeave cloud capacity, reducing the provider’s exposure to gaps between infrastructure buildout and customer demand.
The company’s reported 43 active data centers show that CoreWeave is moving beyond a narrowly scoped GPU-rental model toward operating capacity at a scale that can serve several frontier-model developers concurrently. Its earlier expanded OpenAI capacity agreement established the same long-duration, customer-reserved model now being applied to Anthropic.
First-order effects
Anthropic gains multiyear access to US-based data-center capacity equipped with Nvidia chips, giving it another dedicated channel for meeting demand for its AI services.
CoreWeave adds Anthropic as a major customer and can place more of its expanding data-center footprint under contracted use rather than relying on spot demand.
Second-order effects
The deal strengthens CoreWeave’s position when negotiating for chips, power, and data-center capacity, because a broader set of large customers supports continued infrastructure deployment.
Other AI developers and cloud-capacity providers face added pressure to secure longer-term compute arrangements, while Nvidia benefits from another route for its chips to reach large AI workloads.
Third-order effects
If similar agreements continue, AI compute may be organized less around general-purpose public cloud purchasing and more around long-duration reservations from specialized infrastructure operators.
The combination of customer contracts and supplier backstops could make capacity providers more central intermediaries between chipmakers and model developers, while also concentrating execution risk in a smaller group of operators.
The trend: This is another step in the shift toward specialized AI infrastructure platforms financing large data-center fleets through long-term, bilateral capacity commitments.
1. CoreWeave has successfully brought on about 300mw of capacity in the last year — 2. CoreWeave does not have the capacity to serve this contract before 2027 (at the very very earliest) — 3. Anthropic does not have the money to pay for it — 4. nobody seems to bother talkin…
It's almost as if there is overwhelming demand for AI compute. “CoreWeave Announces Multi-Year Agreement With Anthropic” “With the addition of Anthropic, nine of the leading ten AI model providers now leverage CoreWeave's platform, reflecting the growing demand for [image]
$CRWV sees its best day in about two months after announcing a multiyear agreement with Anthropic. 📈 CoreWeave will support the development and deployment of Anthropic's Claude AI models. [image]
$CRWV Signs Anthropic Love to see it that Anthropic is entering Neocloud market as a customer. They don't need to buy from HS. Both $NBIS $IREN should take note. Anthropic is the growth customer of the future.
$GLXY took a big risk betting on $CRWV in the early days. That bet now seems to be paying off. With every new deal CoreWeave signs, their credit rating improves and Galaxy's tenant risk decreases.
Most people still think AI competition is about models. But the real race just became obvious: Whoever controls the GPU clouds that run the models will quietly control enterprise AI. Today's @CoreWeave-@AnthropicAI deal is a big reminder...
Big news: CoreWeave just signed a multi-year agreement with @AnthropicAI to support the development and deployment of their Claude AI models. 9 of the top 10 AI model providers now run on CoreWeave. When scale and reliability are non-negotiable, CoreWeave is the platform that