The original article, titled An Interview with CoreWeave Executives: AI Demand Seems to Intensify Every Day, was written by Tae Kim and translated for BlockBeats by Peggy. The interview features Brannin McBee, CoreWeave co-founder and chief development officer, and Nick Robbins, vice president of corporate development and investor relations. Their discussion centers on the current state of AI demand, the neocloud market, and the infrastructure requirements behind the next wave of AI workloads.
CoreWeave was founded in 2017 and provides large-scale GPU compute to startups and large enterprises. It is described in the source as an innovative early market leader in the neocloud category and as the only cloud provider to receive the highest platinum rating from AI research firm SemiAnalysis. The company sits in the middle of the AI infrastructure chain: it serves leading customers including OpenAI, Anthropic, Meta, Google, Microsoft, and Nvidia, while also seeing demand changes from research labs, enterprises, and hyperscale cloud providers.
Demand accelerated from the fourth quarter into inference and agents
Asked when the wave of agentic AI demand began to break out, Brannin McBee said CoreWeave saw the real beginning in the fourth quarter of last year. At that time, the company was already in engineering-level conversations with customers about products they expected to bring to market in the first quarter of this year. McBee said this relationship is important to how CoreWeave evaluates customer demand, because the company has a deeply interconnected engineering relationship with customers that allows it to see trends ahead of time rather than respond after the fact.
From an AI product market perspective, McBee said the first quarter marked a major inflection point for inference and AI consumption, and that acceleration has continued. When Tae Kim asked whether AI demand had slowed at all compared with a few months earlier, Nick Robbins answered that it seems to intensify every day in new ways. The message from the interview is direct: the demand cycle did not cool after the earlier rush for GPUs. Agentic AI, reasoning models, and enterprise AI applications are continuing to raise infrastructure requirements.
CPU and storage are becoming part of next-generation AI factory design
The discussion then moved to the rising need for CPUs relative to GPUs in the agentic AI wave. McBee said CoreWeave has been running CPUs since 2023 and has always had a complete cloud product. In his view, the question is not whether the company has only just started adding CPUs, but what customers actually need and whether that need is rising in relative terms. His answer was clear: yes. As agentic capabilities and reasoning capabilities truly emerge in models, storage demand is also rising compared with prior generations, and he said that trend will continue.
Robbins added that customers will absolutely see large numbers of Vera CPUs deployed beside large numbers of Vera Rubin servers. He said CoreWeave fundamentally redesigned its base data center plan last year to leave room for more storage and more CPUs next to GPUs. Robbins tied that decision to CoreWeave's position in the ecosystem. He said the company is the only independent cloud provider serving all the most advanced technology users, and that no other independent AI cloud provider can say Anthropic, OpenAI, Meta, Google, Microsoft, and Nvidia are all its customers. That customer base gives CoreWeave a useful flywheel, or positive feedback loop: it can understand where customers are taking the technology and plan accordingly.
When asked whether CoreWeave would mainly use Nvidia Vera CPUs in the future, Robbins said it depends on the workload. CoreWeave is customer-demand driven. He said the company has disclosed that it expects to be an early and significant adopter of Vera CPUs. At present, its fleet is mainly AMD, but that will change over time according to customer demand, and customer interest in Vera CPU is very strong. McBee used the point to explain how the company's contracts work: more than 98% of CoreWeave revenue is contract-driven. The company is not guessing what infrastructure customers want. Customers tell CoreWeave very specifically what configurations they need, and customers define what the company builds.
Competition with neocloud firms and hyperscalers
The interview also covered competition with SpaceX, Nebius, and Oracle in neocloud, as well as Azure, AWS, and Google among hyperscale cloud providers. On differentiation, McBee said he prefers to look at third-party validation. Excluding China, nine of the world's top ten AI labs use CoreWeave's platform. SemiAnalysis has consistently placed CoreWeave alone at the highest level for performance. McBee said he does not believe CoreWeave receives its GPU allocation because of a personal relationship with Jensen. Instead, he said it reflects deep supplier confidence in the company's execution record and engineering capability, as well as confidence that CoreWeave can best demonstrate the capabilities of those products globally.
Robbins broke down why CoreWeave wins different types of customers. The company wins hyperscale cloud customers because it is very good at execution, can build these systems at extreme speed, and the systems run very well. It wins research lab customers because it offers the strongest technical versions and performs best on per-token efficiency. It wins enterprise customers because the infrastructure runs well and because CoreWeave has built a best-in-class orchestration layer, which is one source of recognitions such as the platinum rating.
Robbins said an increasingly important point is that CoreWeave has built one of the most mature capability layers among AI cloud providers across inference and developer tools, helping enterprises actually put AI into production. According to his description, CoreWeave is building and delivering products that help less technically mature enterprises turn data into models and then into agents that can run internally, while also cross-selling CoreWeave cloud services during that process.
The current bottleneck is powered shells and execution
Asked what the current bottleneck is, whether powered data center shells, GPUs, or electricians, McBee identified powered shells as the constraint. More precisely, he said the constraint is the components inside those shells. When Tae Kim specifically mentioned electricians, McBee said that was exactly right and described it as a complex area. He emphasized that CoreWeave already has 49 such sites online and operating. The company is not placing its hopes on one or two sites; it has done this 49 times, creating a deep execution record.
McBee said that execution record also means CoreWeave has accumulated a significant amount of knowledge about how to handle supply chain issues. The company understands which suppliers in that chain are suitable partners and which are not. In the framing of the interview, the bottleneck has moved from the simple question of whether GPUs are available to a broader infrastructure problem involving powered shells, internal components, electricians, CPU, storage, and supply chain execution.
On HBM memory costs and shortages, Robbins said customers do bear the increased cost. CoreWeave's business model is designed so that when the company signs a GPU purchase order and determines how much it will pay, it also locks in the GPU price charged to customers. More broadly, that is the server price, and the server price includes HBM costs. Robbins said that at present, obtaining components is not the biggest bottleneck; the biggest bottleneck is the powered shell.
Vera Rubin ramp expected to follow a GB-like pattern
The interview closed with the expected deployment ramp for Vera Rubin. Robbins said CoreWeave was clearly the first company in the world to start and fully validate VR, meaning Vera Rubin racks. He said the company did the same last year with GB200 and GB300. Robbins expects VR to begin appearing later this year, with the truly large-scale and very strong deployment ramp running throughout 2027.
Robbins compared that timing with the GB cycle. GB began appearing in 2025, but the real large-scale ramp runs through 2026. He said a fair amount had already been deployed by the end of last year, but this year is the true large-scale deployment year for GB. He expects VR to show a very similar rhythm over the next 12 to 18 months. Across the interview, CoreWeave repeatedly framed its infrastructure decisions as customer-driven: customers define the configurations, the company builds against those requirements, and the next AI data center design is being shaped around the roadmaps of the most advanced users.