Original Title: An Interview with CoreWeave Executives: AI Demand Seems to 'Intensify' Every Day
Original Author: Tae Kim
Translation: Peggy, BlockBeats

Editor's Note: This interview provides a window into observing the AI computing power cycle: demand has not cooled off after the last GPU rush, but is instead being driven higher by intelligent agents, reasoning, and enterprise AI applications.

This article interviews CoreWeave co-founder and Chief Development Officer Brannin McBee, as well as Vice President of Business Development and Investor Relations Nick Robbins, discussing the current state of AI demand and the neocloud market. The core message from CoreWeave executives is very direct — AI demand seems to be intensifying in new ways every day, and the real bottlenecks are shifting from "GPU availability" to more complex infrastructure issues: data center power shells, CPUs, storage, electrical work, supply chain execution capabilities, and how much customers are willing to pay for next-generation computing power.

What makes CoreWeave unique is that it sits in the middle of the AI infrastructure chain: serving top clients like OpenAI, Anthropic, Meta, Google, Microsoft, Nvidia, and also directly sensing the demand changes from research labs, enterprise clients, and hyperscale cloud providers. Therefore, what it observes is not just "GPU shortages," but a structural change in AI workloads themselves. With the rise of agentic AI and reasoning models, computing demand is no longer solely centered on GPUs; the importance of CPUs and storage is also increasing. Next-generation data center designs must reserve space for Vera CPUs, Vera Rubin servers, and more storage.

This also explains why competition in AI infrastructure is shifting from mere chip procurement to more comprehensive engineering delivery capabilities. Whoever can more quickly acquire power supply data centers, deploy servers, streamline supply chains, and optimize per-token costs will be closer to the core of this AI capital expenditure cycle. CoreWeave repeatedly emphasizes "customer-driven," but behind that is a larger judgment: AI cloud providers are no longer just selling compute power but are proactively reconstructing the next-generation AI factories based on the roadmaps of cutting-edge clients.

For investors and industry observers, the most noteworthy aspect of this interview is not a single data point but the direction of change in AI infrastructure demand: GPUs remain important, but bottlenecks are spreading; Nvidia is still central, but CPUs, HBM, storage, and data center power capacity are becoming new variables; AI demand continues to grow, but future winners may depend on who can deliver complex infrastructure continuously, stably, and at scale.

Below is the original text:

CoreWeave is regarded as an early market leader with innovative potential in the neocloud (new cloud service) space.

It is the only cloud service provider to receive the highest "Platinum Rating" from AI research institution SemiAnalysis. Founded in 2017, CoreWeave provides large-scale GPU computing power for startups and large enterprises.

Key Context: Recently, we interviewed CoreWeave co-founder and Chief Development Officer Brannin McBee, as well as Vice President of Business Development and Investor Relations Nick Robbins, to discuss the current state of AI demand and the neocloud market.

Below are the key points from this conversation after editing:

AI Demand Continues to Intensify

Tae: When did the wave of intelligent agent AI demand really start to explode?

Brannin: We saw the true beginning in Q4 last year. At that time, we were engaging with clients on engineering-level discussions about products they expected to launch in Q1 this year.

This has always been a very important perspective for us in understanding client needs. We have a deep interconnected engineering relationship with our clients. It’s this relationship that allows us to see trends early, rather than react passively after changes occur.

From the product perspective of the AI market, I would say Q1 was a turning point for reasoning and AI consumer applications, and this acceleration is still ongoing.

Tae: What is the current state of AI demand? Compared to a few months ago, are there signs of slowdown in recent weeks?

Nick: It seems to be intensifying in new ways every day.

Tae: Please discuss the rising trend of CPU demand relative to GPU in the agentic AI wave. Will you deploy Vera CPU racks alongside Nvidia GPU servers?

Brannin: CoreWeave has been running CPUs since 2023. We have always had a complete cloud product. So the question isn’t whether we are just starting to add CPUs, but what exactly clients need. Is this demand increasing in relative terms? The answer is very clear — yes, it is.

As agentic and reasoning capabilities truly emerge in models, storage needs are also rising compared to previous generations. I believe this trend will continue.

Nick: To answer your question, yes. You will definitely see a large deployment of Vera CPUs alongside many Vera Rubin servers. Last year, we fundamentally redesigned our data center plans to leave more space for storage and CPUs, so they can be deployed next to GPUs.

We do this because we are in a very unique position across the ecosystem. We are the only independent cloud provider serving all the most advanced technology users. No other independent AI cloud provider can say that clients like Anthropic, OpenAI, Meta, Google, Microsoft, Nvidia are all our customers.

This creates a beneficial flywheel or positive feedback loop for our business: we understand where clients are taking their technology and plan accordingly.

Bottlenecks Are No Longer Just GPUs

Tae: Will you mainly use Nvidia Vera CPUs in the future?

Nick: It depends on the specific workload. Our actions are driven by client needs. We do expect to be early and significant adopters of Vera CPUs, which we have already disclosed. Currently, our clusters are mainly AMD, but over time this may change based on client demand. Interest in Vera CPUs from clients is very strong.

Brannin: This also reminds us to talk about how our contracts work. As you know, over 98% of our revenue is contract-driven. We are not guessing what infrastructure clients want; they tell us very clearly what configurations they need. Everything is client-driven. Clients define what we build.

Tae: Let’s discuss the competitive landscape. How do you enter and compete in a market with SpaceX, Nebius, Oracle in the neocloud space, and giants like Azure, AWS, Google?

Brannin: Regarding differentiation, I prefer to look at it from a third-party validation perspective. Outside China, nine of the top ten AI labs globally are using our platform. SemiAnalysis consistently ranks us at the top in performance. I don’t think our GPU allocations are because of personal connections with Jensen.

This indicates that vendors have deep confidence in our execution record and engineering capabilities, believing we can best demonstrate their products worldwide.

Nick: Our ability to win large cloud provider clients is because we excel at execution. We can build these systems very quickly and they run very well. We win research lab clients because we offer the most powerful performance versions and the best token efficiency.

We win enterprise clients because our infrastructure performs reliably, and we have built an excellent, best-in-class orchestration layer, which is also recognized by platinum ratings.

But increasingly, among AI cloud providers, we have built the most mature layer of capabilities, covering reasoning and development tools, helping enterprises truly deploy AI into production.

This means we are building and delivering products that ultimately help less mature companies turn data into models, then into deployable intelligent agents, while cross-selling CoreWeave cloud services in the process.

Tae: What are the current bottlenecks? Power-ready data center shells? GPUs? Electrical work?

Brannin: It’s powered shells, meaning data center shells with power supply. More precisely, the components inside these shells. You specifically mentioned electrical work, which is entirely correct. It’s a complex field.

But importantly, we already have 49 such sites online and operational. We’re not relying on just one or two sites. We’ve done 49.

This is a very solid track record of execution.

It also means we’ve accumulated a lot of knowledge about how to handle supply chain issues, which suppliers are suitable for collaboration, and which are not.

Editor’s note: powered shells refer to the data center buildings themselves, excluding the actual hardware servers.

Tae: What can you share about HBM memory costs and shortages? How are you responding? Do clients need to bear the price increases?

Nick: Yes, the answer is yes. Our business model is designed so that when we sign GPU purchase orders and determine how much we will pay, we lock in the GPU prices we charge to clients. Broadly speaking, this also includes server prices, which obviously incorporate HBM costs.

This is how we isolate the impact of daily price fluctuations.

If the component costs in our next transaction increase, we will reflect this in the prices we believe we can charge clients, thus protecting our profit margins. We are well protected in passing these costs to clients. This is something we monitor very closely.

Currently, component acquisition is not the biggest bottleneck. The biggest bottleneck is powered shells. But in the future, this answer may fluctuate.

Tae: How do you expect Vera Rubin deployment to progress? What will the situation look like in the second half of this year?

Nick: We are obviously the first company worldwide to launch and fully validate VR, that is, Vera Rubin cabinets. We did the same with GB200 and GB300 last year. I expect VR to start appearing later this year.

I anticipate a large-scale, very strong deployment ramp-up throughout 2027. The pace will be similar to GB: GB started appearing in 2025, but the real large-scale ramp-up spanned all of 2026. That is, deployments were already happening at the end of last year, but 2027 will be the year of truly large-scale GB deployment.

I expect VR will follow a very similar rhythm over the next 12 to 18 months.

Click to learn about the rhythm of BlockBeats’ recruitment.

Join the official BlockBeats community:

Telegram Subscription Group: https://t.me/theblockbeats

Telegram Discussion Group: https://t.me/BlockBeats_App

Twitter Official Account: https://twitter.com/BlockBeatsAsia