CoreWeave’s Next Phase: From GPU Scarcity to a Durable AI Cloud

CoreWeave Inc. is transforming its business model from capitalizing on GPU scarcity to establishing a long-term AI cloud infrastructure. A recent research study, commissioned by CoreWeave and conducted by Qualitate, sheds light on the company’s competitive advantages, customer adoption drivers, and the challenges it faces in a market dominated by hyperscalers. The findings reveal how enterprises are balancing GPU spending between CoreWeave and traditional cloud providers while navigating performance, cost, and compliance trade-offs.

Research Highlights CoreWeave’s Competitive Edge and Challenges

According to research conducted by Qualitate and cited by SiliconANGLE, CoreWeave Inc. is transitioning from leveraging GPU scarcity to building a durable artificial intelligence (AI) cloud infrastructure. The study, commissioned by CoreWeave, involved 13 in-depth interviews with enterprise buyers and evaluators across industries such as healthcare, technology, financial services, professional services, and retail. These interviews, totaling over seven hours, explored CoreWeave’s competitive advantages, customer adoption drivers, and challenges like GPU scarcity, performance, cost, and competition with hyperscalers.

The research compared CoreWeave’s financial disclosures with public statements from CEO Michael Intrator and CFO Nitin Agrawal. CoreWeave reported Q2 revenue of $2.6 billion, a 112% year-over-year increase, and a revenue backlog of $104 billion, with an additional $25 billion added in the first month of the current quarter.

GPU Scarcity Drives Adoption, but Performance Retains Customers

GPU scarcity emerged as a key factor driving enterprises to CoreWeave, but the research found that performance, cost, and operational experience are the primary reasons customers stay. Ten of the 13 respondents cited challenges with hyperscaler GPU capacity or product access as part of their decision to adopt CoreWeave. For example, a healthcare IT director reported receiving GPU capacity from CoreWeave in weeks, compared to months with AWS.

CoreWeave’s enterprise support, reliable multi-node performance, Kubernetes compatibility, and GPU workload matching were highlighted as critical factors in customer purchasing decisions. The research also noted that not every proof of concept with CoreWeave converts into a signed customer, as hyperscalers remain deeply embedded in enterprise application estates.

Inference Workloads Grow Alongside Training

The research revealed that inference workloads are expanding alongside training workloads, rather than replacing them. All seven current CoreWeave users reported running inference workloads alongside training, fine-tuning, or model development activities. A healthcare customer’s training-to-inference mix shifted from 90-10 to 60-40, with expectations of reaching a 50-50 split within 12 months. This trend indicates that inference workload growth does not come at the expense of training volumes.

Hyperscalers Remain in the Mix, But CoreWeave Captures GPU Spending

While all seven current CoreWeave users retain hyperscalers like AWS and Azure for broader cloud services, they allocate a significant portion of their GPU spending to CoreWeave. The research found that CoreWeave competes on specialized AI performance and operating experience, while hyperscalers maintain advantages in application services, data platforms, and governance. Current CoreWeave spending ranges from below $100,000 to over $5 million annually, reflecting its growing role in enterprise AI infrastructure.

On-Premises Costs and Compliance Challenges

The research highlighted economic challenges for enterprises considering on-premises AI infrastructure. Breakeven requires sustained GPU utilization of 50-80%, but most early adopters struggle to achieve even 50% utilization. Compliance also influences workload placement, though it is not a blanket veto for CoreWeave adoption. A healthcare director noted that CoreWeave has SOC 2 Type 2 certification and a Business Associate Agreement (BAA) but lacks granular logging and real-time compliance dashboards.

A technology customer chose CoreWeave for regional inference to meet data-locality requirements where hyperscalers could not provide sufficient GPU capacity.

Financial Growth and Risks

CoreWeave’s rapid growth is accompanied by financial risks. The company reported a $10.5 billion gap between operating cash flow ($3.7 billion in the first six months of the year) and cash purchases of property and equipment ($14.1 billion). Three customers—rumored to be Microsoft, OpenAI, and Meta—accounted for 72% of CoreWeave’s Q2 revenue, underscoring concentration risks.

CoreWeave’s balance sheet is not considered a "fortress," with debt and lease obligations growing faster than equity. The company warned in its 10Q filing that a shift toward pay-as-you-go or consumption-based models could impact cash-flow predictability and margins. Customers use a mix of committed contracts, hybrid models, and on-demand capacity to manage spending.

Customer Research Methodology

The research study involved 13 enterprise buyers and evaluators from the United States, Canada, the Netherlands, the United Kingdom, and Germany. Respondents included decision-makers such as field chief officers, vice presidents, directors, and senior managers. Industries represented spanned healthcare, technology, financial services, professional services, retail, and others. The interviews aimed to provide qualitative insights into buyer decisions and infrastructure usage patterns.

What this means

LazyFounders analysis — our interpretation, not reported fact.

CoreWeave’s shift from capitalizing on GPU scarcity to building a durable AI cloud infrastructure reflects a broader maturation of the AI infrastructure market. The research highlights that while scarcity was the initial hook, customers are sticking around for performance, cost efficiency, and operational reliability—factors that suggest CoreWeave is carving out a niche beyond just being a stopgap for GPU shortages.

For founders and operators, this signals a critical insight: infrastructure startups can’t rely solely on scarcity-driven demand. The market is evolving, and long-term success hinges on delivering tangible operational advantages, such as multi-node performance, Kubernetes compatibility, and enterprise-grade support. CoreWeave’s reliance on a handful of major customers (rumored to be Microsoft, OpenAI, and Meta) also underscores the risks of concentration, particularly when aggressive expansion strains cash flow.

The growth of inference workloads alongside training workloads is another noteworthy trend. It challenges the assumption that training is the primary driver of AI infrastructure demand. For startups in this space, this means designing solutions that can flexibly handle both workloads without sacrificing performance or cost efficiency. Finally, the research underscores the importance of compliance and security in enterprise adoption—areas where CoreWeave still has room to improve.

Key takeaways

• CoreWeave’s Q2 revenue grew 112% year-over-year to $2.6 billion, with a $104 billion revenue backlog.
• GPU scarcity drives initial adoption, but performance, cost, and operational experience are key to customer retention.
• Inference workloads are growing alongside training workloads, not at their expense.
• CoreWeave’s rapid growth is accompanied by financial risks, including a $10.5 billion gap between operating cash flow and infrastructure spending.
• Enterprises retain hyperscalers for broader services but allocate significant GPU spending to CoreWeave.

FAQ

What drove CoreWeave’s Q2 revenue growth?

CoreWeave’s Q2 revenue grew 112% year-over-year to $2.6 billion, driven by enterprise adoption of its AI cloud infrastructure. GPU scarcity initially attracted customers, but performance, cost efficiency, and operational experience are key to retention.

How are inference workloads evolving in AI infrastructure?

The research found that inference workloads are growing alongside training workloads, not at their expense. Enterprises are running both types of workloads on CoreWeave’s platform, with some expecting inference to account for 50% of their AI workloads within a year.

Why do enterprises retain hyperscalers while using CoreWeave?

Enterprises retain hyperscalers like AWS and Azure for broader cloud services, such as application hosting, data platforms, and governance. However, they allocate significant GPU spending to CoreWeave for specialized AI performance and operational advantages.

What are the financial risks for CoreWeave?

CoreWeave’s rapid infrastructure expansion has created a $10.5 billion gap between operating cash flow and capital expenditures. Additionally, 72% of its Q2 revenue came from just three customers, raising concentration risks.

How does compliance influence CoreWeave adoption?

Compliance influences workload placement but is not a dealbreaker for CoreWeave adoption. The company has SOC 2 Type 2 certification and a BAA, but some customers note gaps in granular logging and real-time compliance dashboards.

Related on LazyFounders

• Microsoft
• OpenAI
• Meta
• More ai news