By Tae Kim

This article is from the free weekly Barron's Tech email newsletter. Sign up here to get it delivered directly to your inbox.

Battleground Stock. Hi everyone. It's been a wild ride for CoreWeave since the company priced its initial public offering at $40 last week. The pricing came in below the expected range, due to lower-than-expected demand. By Wednesday afternoon, though, the stock was trading at $57, more than 40% above its offering price.

Ultimately, it's a pivotal test of investors' appetite for fast-growing, high-flying artificial intelligence businesses.

CoreWeave is a polarizing stock for investors. Founded in 2017, it provides large-scale access to graphics processing units via the cloud. With physical GPUs expensive and in hot demand, it's a much needed service for both start-ups and more mature enterprises. CoreWeave has a close partnership with Nvidia, which invested in the start-up when it was private and as part of the IPO last week.

Bears question future demand for CoreWeave's cloud, its highly leveraged balance sheet, and the business model itself. Conversely, the bulls see a best-in-class company atop the burgeoning GPU cloud market. Last week, CoreWeave was the only vendor to attain a top Platinum tier rating from semiconductor and AI research firm SemiAnalysis -- above Amazon Web Services, Microsoft Azure, and Google Cloud.

Barron's Tech spoke to CoreWeave co-founder and chief strategy officer Brian Venturo just before the stock began trading on Friday. We discussed major trends in AI, the state of GPU demand, and how CoreWeave is competing with larger cloud computing vendors.

I first spoke with Venturo back in 2023 about the early demand for GPUs. Since then, CoreWeave's sales have soared, with revenue up 737% in 2024, to $1.9 billion.

Here are edited highlights from our latest conversation:

Barron's: How does CoreWeave differentiate itself versus Amazon Web Services, Microsoft Azure, and Google Cloud?

Venturo: The differentiation starts at the lowest level in the data center, in the hardware choices and the configuration that we build, all the way up to the software stack.

We are purpose-built for AI. Whether it's the way that we're building our network architectures to the throughput requirements for storage, we're designing specifically for the AI use cases. The legacy hyperscalers may have to retrofit existing facilities and existing regions to do so.

What are some of the major AI trends you're seeing among your customers?

You still see that the largest AI foundation model builders need access to very large, contiguous compute capacity for training. Our customers are able to pair data centers together now, which helps alleviate some of the constraints of the power grid and the resources around the data centers. But while that's still growing, the inference side of the house is growing more rapidly from a customer perspective because there are so many companies that are now bringing applications to market.

While there may be 15 to 30 of the foundation model builders that need that large contiguous capacity, there's now tens of thousands of customers that are bringing AI applications to market that need inference capacity at the metro edge for their customers.

What categories of AI applications are getting traction?

It's everything from the chatbots to the high-frequency trading firms that are using it for both simulation, as well as live trading strategies. The way that they operate is they identify an edge, and they immediately buy as much as they can until the edge is gone.

The introduction of the OpenAI deep research model where you're able to send out AI to do what's effectively a batch research project is changing the development of intellectual property and of scientific research.

Beyond that, we're starting to see use cases in health sciences and services including AI transcription for medical records and the augmentation of diagnoses for doctors.

What's happening with GPU demand?

The crazy thing is, you see a big dislocation between what the new cycle narrative is and what the narrative on the ground is. The narrative on the ground has been constant for the last several years, which is, "I need more infrastructure, I need bigger infrastructure, and I need it faster." We have customers all the time that are asking for more, larger, faster, and that has not stopped.

There's a major product transition happening at Nvidia. How does CoreWeave manage that transition, given that the new products are more powerful and cost efficient?

From a CoreWeave moat perspective, the faster Nvidia moves, the more pronounced our advantage is.

We were first to market with H100; we were first to market with H200; we were first to market with GB200, right? The faster that goes, the better our ability to be nimble and flexible and bring our expertise to bear does for our customers. That's something that I don't know if the market understands yet.

What's your primary message to investors and employees after the IPO?

We're really proud to get here. It's been a long journey building CoreWeave to the growth levels that we've had. But, as we're telling employees and our investors, it's just a day. While it's great to celebrate, we have to continue executing. We now have a larger portfolio of investors that we get to convince that we're building the next AWS, but for AI. We're excited to prove that business model out over the next several years and convert people into believers one at a time.

Thanks for your time, Brian.

This Week in Barron's Tech

   -- How Google Could Go the Way of Kodak 
 
   -- Apple Stock Nears Citi's Worst-Case. It Remains a Buy Even After Siri 
      Delay. 
 
   -- Roblox Stock Rises on Google Ad Collaboration. Gen Z Users Are Key. 
 
   -- Intel's New CEO Says Turnaround 'Won't Be Easy' 
 
   -- Spotify Stock Continues to Shine. Buy It to Play Defense. 

Write to Tae Kim at tae.kim@barrons.com or follow him on X at @firstadopter.

This content was created by Barron's, which is operated by Dow Jones & Co. Barron's is published independently from Dow Jones Newswires and The Wall Street Journal.

(END) Dow Jones Newswires

April 02, 2025 15:37 ET (19:37 GMT)

Copyright (c) 2025 Dow Jones & Company, Inc.