We use cookies to understand how you use our site and to improve your experience.
This includes personalizing content and advertising.
By pressing "Accept All" or closing out of this banner, you consent to the use of all cookies and similar technologies and the sharing of information they collect with third parties.
You can reject marketing cookies by pressing "Deny Optional," but we still use essential, performance, and functional cookies.
In addition, whether you "Accept All," Deny Optional," click the X or otherwise continue to use the site, you accept our Privacy Policy and Terms of Service, revised from time to time.
You are being directed to ZacksTrade, a division of LBMZ Securities and licensed broker-dealer. ZacksTrade and Zacks.com are separate companies. The web link between the two companies is not a solicitation or offer to invest in a particular security or type of security. ZacksTrade does not endorse or adopt any particular investment strategy, any analyst opinion/rating/report or any approach to evaluating individual securities.
If you wish to go to ZacksTrade, click OK. If you do not, click Cancel.
CRWV's Vera Rubin Push: Can AI Infrastructure Fuel Its Growth Engine?
Read MoreHide Full Article
Key Takeaways
CoreWeave is adding NVIDIA Vera CPUs to support the rising demands of agentic AI workloads.
CRWV says Vera Rubin NVL72 delivered up to 4.8x higher token throughput for SWE-2 inference.
CoreWeave faces risks from NVIDIA dependence, heavy capital needs and competition from major AI clouds.
CoreWeave, Inc. (CRWV - Free Report) is positioning itself as a broader infrastructure platform for the next phase of AI. Its latest announcements around NVIDIA (NVDA - Free Report) Vera CPU and NVIDIA Vera Rubin NVL72 highlight a shift toward supporting agentic AI, reinforcement learning and increasingly complex inference workloads.
Agentic AI systems operate differently from conventional inference workloads. An AI agent may repeatedly reason, call tools, execute code, interact with data and evaluate its own output. Each of these steps can create substantial CPU demand alongside GPU usage. NVIDIA Vera CPU is designed specifically for this emerging workload profile. Its platform will offer Vera on bare metal using the same operating model and economics as the rest of its infrastructure.
A Vera rack can contain 128 CPUs and 11,264 cores, theoretically supporting more than 11,000 concurrent isolated environments. CoreWeave also reported that testing showed more than three times faster agent sandbox startup times compared with an x86 CPU. CoreWeave's Sandboxes product is intended to make these environments easier to provision while providing hardware isolation. The broader opportunity for CoreWeave is to monetize not only accelerator hours but also the CPU-intensive infrastructure surrounding every AI workload.
NVDA Vera Rubin Could Strengthen CRWV’s GPU Advantage
The second announcement is likely to have greater implications for CoreWeave's competitive positioning. The company has announced the availability of the NVIDIA Vera Rubin NVL72, with Cognition as its first production customer. Cognition uses CoreWeave for training, reinforcement learning and inference. According to Cognition's testing on CoreWeave, the Vera Rubin NVL72 delivered up to 4.8x higher total token throughput for SWE-2 inference workloads versus a GB200 NVL72 baseline, 3.8x higher output-token throughput for reinforcement-learning workloads, faster research and development cycles, and potentially lower cost per AI session. These numbers highlight why rapid access to new NVIDIA architectures could be commercially important for CoreWeave.
If agentic applications become a major AI growth category, customers such as Cognition could provide an important demonstration of CoreWeave's ability to support workloads that require tightly integrated compute resources. The company also emphasizes its full-stack platform, including CoreWeave Kubernetes Service, SUNK, Mission Control, CoreWeave Sandboxes and serverless inference.
However, this strategy carries risks. CoreWeave remains heavily dependent on NVIDIA's technology roadmap and must continually invest in expensive infrastructure. AI hardware evolves rapidly, while data-center construction, power procurement and financing require substantial capital. Competition is another consideration. Hyperscalers and other specialized GPU clouds such as Microsoft Azure (MSFT - Free Report) and Nebius Group N.V. (NBIS - Free Report) are also investing aggressively in AI infrastructure.
CRWV vs. Peers: Comparing AI Infrastructure Strategies
Nebius continues to see strong demand for its AI cloud infrastructure, with the company closing four landmark deals during the quarter with an average value of more than $1 billion each. These agreements carry yields of $20-$25 million per MW, while upfront payments cover 50-60% of the associated capital expenditures. Like CRWV, NBIS continues to deepen ties with NVDA. In June, it announced plans to invest approximately £1.7 billion in expanding AI compute capacity across the U.K.
The investment includes three new deployments of advanced NVIDIA-powered infrastructure. Nebius launched its first U.K. deployment of NVIDIA Blackwell Ultra infrastructure in late 2025. Building on that foundation, it plans to establish three additional sites across the U.K., deploying the latest generations of NVIDIA’s full-stack AI factory platform technology. When fully operational in 2027, these deployments are expected to deliver 65 MW of AI computing capacity.
Microsoft is capitalizing on AI business momentum and Copilot adoption while accelerating Azure cloud infrastructure expansion. Its AI investments are converting into measurable commercial traction across its stack. Multi-model flexibility, paired with continued access to OpenAI's frontier models under an IP arrangement extending to 2032, allows customers to optimize cost and performance while keeping Microsoft central to their AI infrastructure decisions. However, Microsoft’s August 2026 receipt of NVIDIA’s first production Vera Rubin GPUs and its June 2026 completion of the first Mount Pleasant data center facility show that infrastructure expansion is progressing, but they also illustrate the scale of investment needed to support demand.
View the Zacks #1 Rank List
Image: Shutterstock
CRWV's Vera Rubin Push: Can AI Infrastructure Fuel Its Growth Engine?
Key Takeaways
CoreWeave, Inc. (CRWV - Free Report) is positioning itself as a broader infrastructure platform for the next phase of AI. Its latest announcements around NVIDIA (NVDA - Free Report) Vera CPU and NVIDIA Vera Rubin NVL72 highlight a shift toward supporting agentic AI, reinforcement learning and increasingly complex inference workloads.
Agentic AI systems operate differently from conventional inference workloads. An AI agent may repeatedly reason, call tools, execute code, interact with data and evaluate its own output. Each of these steps can create substantial CPU demand alongside GPU usage. NVIDIA Vera CPU is designed specifically for this emerging workload profile. Its platform will offer Vera on bare metal using the same operating model and economics as the rest of its infrastructure.
A Vera rack can contain 128 CPUs and 11,264 cores, theoretically supporting more than 11,000 concurrent isolated environments. CoreWeave also reported that testing showed more than three times faster agent sandbox startup times compared with an x86 CPU. CoreWeave's Sandboxes product is intended to make these environments easier to provision while providing hardware isolation. The broader opportunity for CoreWeave is to monetize not only accelerator hours but also the CPU-intensive infrastructure surrounding every AI workload.
NVDA Vera Rubin Could Strengthen CRWV’s GPU Advantage
The second announcement is likely to have greater implications for CoreWeave's competitive positioning. The company has announced the availability of the NVIDIA Vera Rubin NVL72, with Cognition as its first production customer. Cognition uses CoreWeave for training, reinforcement learning and inference. According to Cognition's testing on CoreWeave, the Vera Rubin NVL72 delivered up to 4.8x higher total token throughput for SWE-2 inference workloads versus a GB200 NVL72 baseline, 3.8x higher output-token throughput for reinforcement-learning workloads, faster research and development cycles, and potentially lower cost per AI session. These numbers highlight why rapid access to new NVIDIA architectures could be commercially important for CoreWeave.
If agentic applications become a major AI growth category, customers such as Cognition could provide an important demonstration of CoreWeave's ability to support workloads that require tightly integrated compute resources. The company also emphasizes its full-stack platform, including CoreWeave Kubernetes Service, SUNK, Mission Control, CoreWeave Sandboxes and serverless inference.
However, this strategy carries risks. CoreWeave remains heavily dependent on NVIDIA's technology roadmap and must continually invest in expensive infrastructure. AI hardware evolves rapidly, while data-center construction, power procurement and financing require substantial capital. Competition is another consideration. Hyperscalers and other specialized GPU clouds such as Microsoft Azure (MSFT - Free Report) and Nebius Group N.V. (NBIS - Free Report) are also investing aggressively in AI infrastructure.
CRWV vs. Peers: Comparing AI Infrastructure Strategies
Nebius continues to see strong demand for its AI cloud infrastructure, with the company closing four landmark deals during the quarter with an average value of more than $1 billion each. These agreements carry yields of $20-$25 million per MW, while upfront payments cover 50-60% of the associated capital expenditures. Like CRWV, NBIS continues to deepen ties with NVDA. In June, it announced plans to invest approximately £1.7 billion in expanding AI compute capacity across the U.K.
The investment includes three new deployments of advanced NVIDIA-powered infrastructure. Nebius launched its first U.K. deployment of NVIDIA Blackwell Ultra infrastructure in late 2025. Building on that foundation, it plans to establish three additional sites across the U.K., deploying the latest generations of NVIDIA’s full-stack AI factory platform technology. When fully operational in 2027, these deployments are expected to deliver 65 MW of AI computing capacity.
Microsoft is capitalizing on AI business momentum and Copilot adoption while accelerating Azure cloud infrastructure expansion. Its AI investments are converting into measurable commercial traction across its stack. Multi-model flexibility, paired with continued access to OpenAI's frontier models under an IP arrangement extending to 2032, allows customers to optimize cost and performance while keeping Microsoft central to their AI infrastructure decisions. However, Microsoft’s August 2026 receipt of NVIDIA’s first production Vera Rubin GPUs and its June 2026 completion of the first Mount Pleasant data center facility show that infrastructure expansion is progressing, but they also illustrate the scale of investment needed to support demand.