CoreWeave Says Margins Set to Inflect as AI Power Capacity Comes Online

Key Points

• CoreWeave expects margins to improve in the second half of the year as newly added power capacity turns into deployed GPUs and customer revenue. CFO Nitin Agrawal said margins should “inflect” from Q2 to Q3 and reach low double digits by year-end.
• The company said its contract-driven model helps limit cost volatility because it locks in much of its revenue and infrastructure costs when deals are signed. CoreWeave also said higher input costs, such as memory, can be reflected in pricing on new contracts.
• CoreWeave sees strong AI demand, especially for inference workloads, and said customers are expanding into its broader full-stack offerings like storage, networking, CPU and software. It also highlighted long-term growth plans tied to new self-built sites, with more operational control expected to support future margin gains.

CoreWeave NASDAQ: CRWV Chief Financial Officer Nitin Agrawal said the AI cloud company expects margin improvement in the second half of the year as recently added power capacity is converted into deployed GPUs and customer revenue.

Speaking in a fireside-style discussion, Agrawal said CoreWeave structures its business around contribution margin at the contract level and monitors those economics closely as it signs long-term customer commitments across different generations of infrastructure.

“We are reaching an inflection point in terms of that power growth, and then deploying GPUs and delivering it to our end customers,” Agrawal said. He added that the company has confidence in its margin outlook because it knows both the cost structure and revenue structure tied to its contracts.

Agrawal said CoreWeave expects margins to “inflect” from the second quarter to the third quarter and reach “low double digits” as the company exits the year. He also said the company expects to enter 2027 in a stronger position due to the power and deployment dynamics now underway.

Contract Structure Seen Limiting Cost Volatility

Asked about rising input costs, including memory prices and other supply chain pressures, Agrawal said CoreWeave uses a combination of upfront purchasing and pricing discipline. He said the company places purchase orders for infrastructure as it signs customer contracts, which helps lock in both revenue and “a large fraction” of its cost structure alongside data center leases.

“That gives us confidence that the contribution margins that these contracts will achieve once they are actually deployed and running at scale would be very, very static,” Agrawal said.

For new contracts, Agrawal said higher capacity costs may be reflected in pricing to customers. He described those costs as effectively being considered in the next incremental contract based on capacity availability and costs at that time.

CoreWeave Points to AI-First Cloud Architecture

Agrawal said CoreWeave’s differentiation comes from building an “AI-first cloud” from the ground up, rather than adapting traditional cloud infrastructure designed for web applications, websites and data lakes. He said AI workloads are fundamentally different because they are parallelized workloads, requiring different infrastructure, networking and software layers.

Agrawal said CoreWeave’s platform has expanded beyond GPUs. He said the company previously disclosed that its storage product exceeded $100 million in annual recurring revenue and continues to grow. He also said networking, CPU and software are expected to cross $100 million of ARR by the end of the year.

He cited CoreWeave’s history of deploying NVIDIA infrastructure at scale, saying the company was first to deploy H100, H200 and Blackwell systems at scale, and said Vera Rubin systems are “coming just across the corner.” He also pointed to recognition from SemiAnalysis, saying CoreWeave has been the single platinum cloud provider in AI infrastructure for two consecutive years.

Inference Demand Cited as Signal of Durability

Agrawal said one of the strongest signals that AI infrastructure demand is durable is the shift toward inference workloads. He said a majority of demand coming to CoreWeave is now for serving inference workloads, which he characterized as evidence of a sustained economic change.

“Customers are productionalizing AI now, and they’re seeing massive returns on those,” Agrawal said.

He also said average selling prices for A100, H100, H200 and L40 GPUs increased quarter over quarter in CoreWeave’s most recent quarter discussed on earnings. Agrawal said CoreWeave is selling H100s at prices higher than two to three years ago, when they first came out, which he described as a sign of sustained demand for serving workloads.

Agrawal said inference demand is not “one size fits all,” with customers using different models and selecting infrastructure that best fits their needs, including A100s, H100s and L40S systems.

Software and Omni Viewed as Longer-Term Opportunities

Agrawal said CoreWeave does not rely on customer lock-in, instead pointing to platform quality and service levels. He said customers whose initial contracts have expired are returning not only to renew but also to buy additional capacity.

He said customers initially come to CoreWeave for GPU infrastructure but are increasingly adopting storage, CPU, networking and software offerings as part of a broader full-stack AI cloud platform.

Discussing software, Agrawal described two areas: managed software products, including Weave and CoreWeave’s inference platform, and CoreWeave Omni, which monetizes the company’s AI infrastructure stack in GPUs and data centers not owned by CoreWeave. He said Omni could support sovereign, international or enterprise-owned infrastructure using the CoreWeave stack.

Agrawal said these software efforts are not near-term contributors at the scale of the existing business, but he called them “margin accretive” and said they could become meaningful over the coming years. He also said CoreWeave discussed exiting this year at $18 billion to $19 billion ARR and next year north of $30 billion ARR, while revenue backlog approached $100 billion last quarter, with the moderator referencing $99.4 billion.

Power, Supply Chain and Self-Build Strategy

Agrawal said supply constraints in AI infrastructure continue to shift. While chips were the key constraint several years ago, he said power and shell capacity are now the biggest constraints, including shortages in skilled labor such as electricians and plumbers. He also noted memory as a current component shortage.

“This is the only environment we know of,” Agrawal said, adding that CoreWeave was built in an environment where supply constraints were always present.

The moderator said CoreWeave has 3.5 gigawatts of contracted power and a goal of reaching 8 gigawatts of active power by 2030. Agrawal said the company initially leased capacity to move quickly and balance capital expenditures against operating expenses. Looking ahead, he said CoreWeave expects a “healthy mix” of leased and self-built sites, with an increasing percentage of self-built capacity over the next several years.

Agrawal said the company is excited to bring its first self-built site online in the next few quarters this year, with another site expected in 2027. He said greater operational control could help CoreWeave innovate faster while vertical integration could allow it to recapture margin.

On competition, Agrawal said new entrants are unsurprising given demand and supply constraints, but he emphasized the difficulty of delivering performant infrastructure at scale. He said CoreWeave’s technology stack, operational execution and financing engine are key differentiators, adding that the company has used GPU-backed financing for nearly four years.

About CoreWeave NASDAQ: CRWV

CoreWeave is a U.S.-based provider of GPU-accelerated cloud infrastructure designed to support compute-intensive workloads such as artificial intelligence, machine learning, visual effects rendering and other high-performance computing applications. The company supplies access to large fleets of modern GPUs and complementary infrastructure that enable customers to train and deploy large models, run inference at scale, and process graphics-heavy workloads with low latency and high throughput.

CoreWeave’s product offering includes on-demand and dedicated GPU instances, bare-metal servers, private clusters and managed services tailored for enterprise and developer use.

This instant news alert was generated by narrative science technology and financial data from MarketBeat in order to provide readers with the fastest reporting and unbiased coverage. Please send any questions or comments about this story to contact@marketbeat.com.