CtrlK
CoreWeave, Inc. is an American artificial intelligence cloud computing company headquartered in Livingston, New Jersey. The company specializes in providing cloud-based GPU infrastructure to AI developers and enterprises, and also develops its own chip management software. Originally founded in 2017 as a cryptocurrency mining operation, CoreWeave pivoted to GPU cloud services in 2019 and has since become one of the largest specialized cloud providers for AI workloads. The company went public on the Nasdaq stock exchange on March 28, 2025, under the ticker symbol CRWV.
CoreWeave was founded in 2017 in New Jersey by Michael Intrator, Brian Venturo, and Brannin McBee, along with Peter Salanki. All three primary founders came from backgrounds in commodities trading. Michael Intrator had been working as a portfolio manager and founder of Hudson Ridge Asset Management, an energy industry hedge fund. The founders had built a machine learning model to help them pick investments in the data-heavy energy industry, and Brannin McBee ran the data firm they used.
The company was originally incorporated as Atlantic Crypto Corp. and operated as an Ethereum mining company. The founders used NVIDIA GPUs to verify blockchain transactions, running operations out of a New Jersey data center. At the time, Bitcoin and other cryptocurrencies were rising sharply in price, making GPU-based mining a lucrative business.
Following the cryptocurrency market crash of 2018, Ethereum mining became significantly less profitable. In 2019, the company was renamed CoreWeave, and the founders decided to leverage their large existing inventory of GPUs to provide cloud computing infrastructure to companies. This pivot proved prescient, as the demand for GPU compute was about to explode with the rise of generative AI.
The company initially targeted clients in visual effects (VFX) rendering, scientific computing, and other GPU-intensive workloads. CoreWeave differentiated itself from traditional hyperscale cloud providers like Amazon Web Services, Microsoft Azure, and Google Cloud by offering GPU-specialized infrastructure at lower costs and with faster provisioning times.
As the AI industry accelerated, particularly following the release of ChatGPT in late 2022, demand for GPU compute surged. CoreWeave was well-positioned to capitalize on this trend. In April 2023, the company raised $221 million in a Series B funding round led by Magnetar Capital, with participation from NVIDIA, former GitHub CEO Nat Friedman, and former Apple executive Daniel Gross. This round valued the company at approximately $2 billion.
In January 2023, CoreWeave acquired Conductor Technologies, a cloud-based rendering startup that simplified access to cloud resources at scale. This acquisition enhanced CoreWeave's offerings to VFX studios and brought its headcount to 95 employees. Mac Moore, former CEO of Conductor Technologies, began leading CoreWeave's Media and Entertainment division.
In September 2023, CoreWeave and NVIDIA announced that they had built what they described as the fastest AI supercomputer in the world, located in Plano, Texas. The system, built using over 3,500 NVIDIA H100 GPUs, trained a GPT-3-scale large language model with 175 billion parameters in under 11 minutes, over 29 times faster than competitors in a benchmark test.
In May 2024, CoreWeave raised $1.1 billion in Series C funding led by Coatue Management, with participation from Magnetar Capital, Altimeter Capital, Fidelity Management & Research Co., and Lykos Global Management. This round valued the company at $19 billion.
Also in May 2024, the company secured a $7.5 billion debt financing facility led by Blackstone and Magnetar Capital, one of the largest private debt deals in the technology sector. A further $650 million credit facility was announced in October 2024 to support ongoing growth.
In November 2024, a secondary round was completed with lead investors including Fidelity, Jane Street Capital, Macquarie Capital, and Magnetar Capital, at a valuation of $23 billion.
During 2024, CoreWeave hired key executives to prepare for its public listing. Nitin Agrawal, formerly of Google, joined as Chief Financial Officer. Sachin Jain, previously from Oracle's AI department, was hired as Chief Operating Officer in August 2024.
CoreWeave went public on March 28, 2025, listing on the Nasdaq under the ticker CRWV. The IPO priced at $40.00 per share, with 37.5 million shares offered (36.59 million sold by CoreWeave and 910,000 by selling stockholders). The offering raised approximately $1.5 billion, making it the largest AI-related IPO by amount raised, according to Dealogic.
The IPO launched during a period of broader market volatility related to U.S. tariff announcements. Despite these headwinds, CoreWeave CEO Michael Intrator described the listing as "incredibly successful." By May 2025, the company's market capitalization had doubled to roughly $49 billion. The stock reached a 52-week high of $187.00 before pulling back to trade around $83.67 as of March 2026.
In February 2025, Glenn Hutchins was appointed lead independent director of the board.
| Name | Role | Background |
|---|---|---|
| Michael Intrator | Co-Founder, CEO | Former portfolio manager and founder of Hudson Ridge Asset Management, an energy industry hedge fund |
| Brian Venturo | Co-Founder, Chief Strategy Officer | Former commodities trader specializing in energy futures |
| Brannin McBee | Co-Founder, Chief Development Officer | Ran a data analytics firm used by the other founders |
| Peter Salanki | Co-Founder, CTO | Technical co-founder responsible for engineering and architecture |
| Nitin Agrawal | CFO | Formerly at Google; joined in 2024 |
| Sachin Jain | COO | Formerly at Oracle's AI department; joined August 2024 |
| Glenn Hutchins | Lead Independent Director | Appointed February 2025 |
As of June 2025, CEO Michael Intrator's personal wealth had grown to an estimated $10 billion, making him one of the wealthiest individuals in the AI industry.
CoreWeave operates as a specialized cloud computing provider focused exclusively on GPU-accelerated workloads. Unlike general-purpose hyperscale cloud providers that offer a broad range of computing services, CoreWeave concentrates on delivering high-performance GPU infrastructure optimized for AI training, inference, rendering, and high-performance computing (HPC).
CoreWeave primarily generates revenue through long-term contracts with enterprise customers who reserve GPU capacity. These contracts typically span multiple years, providing the company with predictable, recurring revenue. As of December 2025, CoreWeave reported a revenue backlog of $66.8 billion, more than four times its position at the start of the year.
The company's customer concentration has been a notable feature of its business. In 2024, Microsoft accounted for 62% of CoreWeave's total revenue. By the June 2025 quarter, Microsoft still represented 71% of revenue. However, the company has been actively diversifying its customer base through large deals with OpenAI and Meta.
CoreWeave claims to be up to 80% more cost-effective than major hyperscale cloud providers for GPU workloads, thanks to its purpose-built infrastructure and usage-based billing model. The company's instances can launch up to 35 times faster than traditional virtual machines on legacy cloud platforms.
CoreWeave's cloud platform is built on a Kubernetes-native architecture designed specifically for large-scale, GPU-intensive tasks. The CoreWeave Kubernetes Service (CKS) is a purpose-built orchestration layer that allows AI researchers to deploy thousands of GPUs in seconds.
CKS clusters use bare-metal nodes with NVIDIA BlueField Data Processing Units (DPUs) for offloading node and resource management processes. The platform natively integrates with workload orchestration tools like Slurm, KubeFlow, and KServe.
CoreWeave developed SUNK (Slurm on Kubernetes), an implementation of Slurm deployed on Kubernetes using Helm. SUNK is designed to be a unified training system for demanding AI workloads, delivering production-grade reliability and operational visibility for large, long-running training jobs.
SUNK packages everything needed to run a Slurm cluster in a configurable Helm chart that deploys Kubernetes resources, with base images that pre-install CUDA, NCCL, InfiniBand, environment modules, and Conda. Each Slurm node runs as a Kubernetes Pod with a slurmd container, and SUNK uses declarative node definitions to define and scale the cluster.
CoreWeave Mission Control monitors cluster health end-to-end, detects silent hardware issues and GPU stragglers, and mitigates failures before they stall synchronous training.
CoreWeave's network infrastructure is built on NVIDIA InfiniBand with SHARP (Scalable Hierarchical Aggregation and Reduction Protocol), providing high-bandwidth, low-latency interconnects between GPUs. The company also uses NVIDIA Quantum-2 InfiniBand networking for its latest deployments.
CoreWeave offers a range of NVIDIA GPU accelerators across multiple generations:
| GPU Family | Models Available | Key Use Cases |
|---|---|---|
| NVIDIA Blackwell Ultra | GB300 NVL72 | AI reasoning, agentic workloads, next-generation training |
| NVIDIA Blackwell | GB200 NVL72 | Large-scale AI training, inference at scale |
| NVIDIA Hopper | H100, H200 | Large language model training, HPC |
| NVIDIA Ada Lovelace | Various | Inference, rendering, VFX |
In April 2025, CoreWeave became the first AI cloud provider to make NVIDIA GB200 NVL72 systems generally available. Cohere, IBM, and Mistral AI were among the first customers to gain access. In July 2025, CoreWeave became the first hyperscaler to deploy the NVIDIA GB300 NVL72 platform, which delivers up to a 10x boost in user responsiveness and a 5x improvement in throughput per watt compared to the previous-generation Hopper architecture.
In June 2025, CoreWeave, in collaboration with NVIDIA and IBM, submitted the largest-ever MLPerf Training v5.0 benchmark using nearly 2,500 NVIDIA GB200 Grace Blackwell Superchips. The system trained the Llama 3.1 405B model in 27.3 minutes, achieving a record result on the most complex model in the benchmark suite.
CoreWeave has reported up to 20% higher GPU cluster performance compared to alternative cloud solutions, attributing this to its purpose-built infrastructure and network optimization.
As of 2025, CoreWeave operated 32 data centers with a total of approximately 250,000 GPUs. The company's infrastructure spans over 33 data centers across North America and Europe, available through flexible public or dedicated cloud deployments.
| Location | Details |
|---|---|
| Plano, Texas | ~450,000 sq ft facility housing over 3,500 H100 GPUs; $1.6 billion supercomputer built in partnership with NVIDIA |
| Kenilworth, New Jersey | 280,000 sq ft data center; $1.2 billion investment |
| Douglasville, Georgia | High-density colocation facility operated in collaboration with Flexential |
| Hillsboro, Oregon | Facility operated by Flexential |
| Volo, Illinois | Facility powered by Bloom Energy solid oxide fuel cells |
| Various other U.S. locations | Multiple additional facilities across the United States |
| Location | Details |
|---|---|
| Crawley, United Kingdom | First European data center, operational October 2024; hosts NVIDIA H200 GPUs with Quantum-2 InfiniBand |
| London Docklands, United Kingdom | Second UK facility, operational December 2024 |
| Norway | Planned facility as part of $2.2 billion European expansion; powered by 100% renewable energy |
| Sweden | Planned facility as part of $2.2 billion European expansion; powered by 100% renewable energy |
| Spain | Planned facility as part of $2.2 billion European expansion; powered by 100% renewable energy |
CoreWeave's European expansion represents a total commitment of $3.5 billion, combining the $1.3 billion UK investment with the $2.2 billion continental expansion. The European data centers are designed to meet regional regulatory and operational requirements, providing data-sovereign cloud services with low-latency performance.
| Date | Round | Amount | Lead Investors | Valuation |
|---|---|---|---|---|
| October 2019 | Series A | $2.5 million | - | - |
| November 2021 | Convertible Note | $50 million | Magnetar Capital | - |
| December 2022 | Debt Financing | $100 million | Magnetar Capital | - |
| April 2023 | Series B | $221 million | Magnetar Capital | ~$2 billion |
| 2023 | DDTL 1.0 | Up to $2.3 billion | - | - |
| May 2024 | Series C | $1.1 billion | Coatue Management | $19 billion |
| May 2024 | DDTL 2.0 | $7.5 billion | Blackstone, Magnetar Capital | - |
| October 2024 | Credit Facility | $650 million | - | - |
| November 2024 | Secondary Round | - | Fidelity, Jane Street Capital, Macquarie Capital, Magnetar Capital | $23 billion |
| March 2025 | IPO | $1.5 billion | Public offering on Nasdaq | ~$23 billion (at IPO price) |
| July 2025 | DDTL 3.0 | $2.6 billion | - | - |
| July 2025 | Senior Notes | $1.75 billion | - | - |
| January 2026 | Private Placement | $2.0 billion | NVIDIA | At $87.20/share |
By mid-2025, CoreWeave's total debt had reached approximately $11.17 billion, excluding an additional $3.5 billion in lease liabilities. The company has raised over $25 billion in total through combined equity financing and debt capital commitments.
CoreWeave has attracted 33 investors, including 31 institutional investors. Major investors include:
| Year | Annual Revenue | Year-over-Year Growth |
|---|---|---|
| 2022 | $15.8 million | - |
| 2023 | $229 million | ~1,349% |
| 2024 | $1.92 billion | ~737% |
| 2025 | $5.13 billion | ~168% |
Despite rapid revenue growth, CoreWeave has not yet achieved profitability on a net income basis due to heavy capital expenditures on GPU hardware, data center buildouts, and debt servicing costs.
| Metric | FY 2024 | FY 2025 |
|---|---|---|
| Revenue | $1.92 billion | $5.13 billion |
| Net Loss | $863 million | $1.167 billion |
| Adjusted EBITDA | $1.22 billion | $3.09 billion |
| Adjusted EBITDA Margin | ~64% | ~60% |
| Revenue Backlog | ~$15.1 billion | $66.8 billion |
The company's Q4 2025 results showed revenue of $1.57 billion (up 110% year-over-year), adjusted EBITDA of $898 million (a 57% margin), and a net loss of $452 million.
Microsoft has been CoreWeave's largest customer since the company's early growth phase. In 2024, Microsoft accounted for 62% of CoreWeave's total revenue, and this concentration remained high at 71% of revenue in Q2 2025. Microsoft has announced plans to spend around $10 billion on CoreWeave's services by the end of the decade. CoreWeave has been actively working to reduce its revenue dependence on Microsoft through diversification.
CoreWeave has built a substantial relationship with OpenAI through a series of escalating contracts:
| Date | Contract Value | Details |
|---|---|---|
| March 2025 | Up to $11.9 billion | Initial infrastructure agreement; OpenAI received $350 million in CoreWeave equity |
| May 2025 | Up to $4 billion | Expanded agreement for additional cloud computing capacity |
| September 2025 | Up to $6.5 billion | Further expansion of the partnership |
| Total | Up to $22.4 billion | Combined value across all agreements |
The OpenAI partnership complements OpenAI's existing commercial relationships with Microsoft and Oracle, as well as its joint venture with SoftBank on the Stargate project.
In September 2025, CoreWeave secured a $14.2 billion contract with Meta to provide AI cloud infrastructure. Under the agreement, Meta committed to pay CoreWeave up to approximately $14.2 billion through December 2031, with the option to expand its commitment through 2032. CoreWeave will provide Meta with access to NVIDIA GB300 NVL72 systems, each containing 72 Blackwell AI GPUs. The deal marked a significant step in reducing CoreWeave's revenue dependence on Microsoft.
The relationship between CoreWeave and NVIDIA is one of the most important dynamics in the AI infrastructure industry. NVIDIA is simultaneously a strategic investor, technology partner, and key supplier to CoreWeave.
NVIDIA first invested in CoreWeave during the 2023 Series B funding round. In January 2026, NVIDIA made a $2 billion private placement investment at $87.20 per share, further deepening the relationship.
CoreWeave has consistently been among the first cloud providers to deploy new NVIDIA GPU architectures. The company was the first to make NVIDIA GB200 NVL72 systems generally available (April 2025) and the first hyperscaler to deploy NVIDIA GB300 NVL72 systems (July 2025). CoreWeave is also a participant in NVIDIA's DGX Cloud Lepton marketplace.
In 2023, CoreWeave and NVIDIA collaborated to build what they described as the world's fastest AI supercomputer at CoreWeave's Plano, Texas data center, using over 3,500 H100 GPUs.
CoreWeave uses NVIDIA BlueField DPUs in its data centers for network and resource management, and its network infrastructure relies on NVIDIA InfiniBand and Quantum-2 networking technology.
CoreWeave operates in the rapidly growing GPU cloud infrastructure market alongside several categories of competitors:
The traditional "Big Three" cloud providers (Amazon Web Services, Microsoft Azure, and Google Cloud) all offer GPU instances for AI workloads. However, CoreWeave differentiates itself through GPU specialization, faster provisioning, and competitive pricing. The hyperscalers offer broader service portfolios, while CoreWeave focuses exclusively on GPU-intensive compute.
CoreWeave competes with a growing category of GPU-focused cloud providers, sometimes called "neoclouds":
| Company | Focus | Differentiation |
|---|---|---|
| Lambda Labs | GPU cloud for AI | Developer-friendly experience; strong focus on researchers and ML engineers |
| Together AI | Training and inference | Pay-per-token model; focused on open-source model hosting |
| RunPod | On-demand GPU cloud | Lower price point; popular with individual developers |
| Vast.ai | GPU marketplace | Peer-to-peer model; lowest cost option |
| SambaNova | AI accelerator platform | Custom chips (RDU) rather than NVIDIA GPUs |
| Fireworks AI | Inference optimization | Focused on fast, cost-efficient inference |
CoreWeave is positioned at the enterprise end of the GPU cloud market, requiring multi-year contract commitments for exclusive access to large GPU clusters. Lambda Labs targets more flexible, growth-stage companies, while Together AI focuses on startups with pay-per-token pricing.
CoreWeave has experienced rapid headcount growth. At the time of the Conductor Technologies acquisition in January 2023, the combined company had 95 employees. By the end of 2024, headcount had grown to 881. As of 2025, the company employed approximately 1,450 people. By early 2026, headcount had reached approximately 2,189 employees.
CoreWeave trades on the Nasdaq under the ticker CRWV. Since its IPO at $40.00 per share in March 2025:
Analyst consensus as of early 2026 showed 19 analysts recommending a buy rating, with an average 12-month price target of $120.07 (range: $38 to $295).
CoreWeave's aggressive expansion has been funded primarily through debt. By mid-2025, total debt reached approximately $11.17 billion, with an additional $3.5 billion in lease liabilities. The DDTL 1.0 facility carries a floating interest rate of roughly 15%, while the DDTL 2.0 facility averages around 11%. Servicing this debt requires sustained revenue growth and presents a significant financial risk if demand for GPU compute were to slow.
CoreWeave's heavy reliance on a small number of large customers, particularly Microsoft (which accounted for 71% of Q2 2025 revenue), represents a concentration risk. The company has been actively addressing this through major deals with OpenAI and Meta.
The GPU cloud business is extremely capital-intensive, requiring continuous investment in the latest GPU hardware and data center infrastructure. GPU technology evolves rapidly, and older hardware can become less competitive quickly. CoreWeave must continually invest billions to maintain its technological edge.
The GPU cloud market is attracting significant investment from both hyperscale providers expanding their GPU offerings and new entrants building specialized infrastructure. CoreWeave's ability to maintain its competitive advantages in pricing, performance, and provisioning speed will be tested as the market matures.
CoreWeave's business depends almost entirely on NVIDIA GPUs, creating a single-supplier dependency. Any disruption in NVIDIA's supply chain, changes in NVIDIA's allocation priorities, or the emergence of competitive chip alternatives could affect CoreWeave's operations.