CoreWeave IPO: From Crypto GPUs to $1.5B Valuation

CoreWeave's Initial Public Offering: A Measured Start

Trading commenced for CoreWeave on Friday, but the launch was characterized by cautious investor response rather than enthusiastic demand. The company's initial pricing settled at $40 per share, falling short of the previously projected range of $47 to $50.

Furthermore, the number of shares made available was reduced prior to the offering. Despite this, CoreWeave successfully secured $1.5 billion in funding and achieved a $14 billion market capitalization on its first day of trading.

This outcome contrasts with initial expectations of raising over $3 billion and attaining a significantly higher valuation. Shares debuted at $39 and ultimately concluded trading at $40, indicating a relatively subdued market reception.

Origins and Early Ventures

Despite the modest IPO performance, CoreWeave’s listing represents the largest initial public offering to date within the artificial intelligence sector. It also marks the most substantial U.S. technology IPO since 2021.

Brian Venturo, the company’s chief strategy officer, expressed a sense of gratitude during an interview with TechCrunch. He noted his fortunate position, speaking from a simple conference room while displaying a distinct New Jersey accent.

The company’s genesis can be traced back to a period of downtime experienced by Venturo and his colleagues following the conclusion of a previous investment endeavor.

Venturo previously served as a portfolio manager at Hudson Ridge Asset Management, a hedge fund specializing in the energy sector and founded by CoreWeave’s CEO, Michael Intrator.

During his time at Hudson Ridge, they developed a machine learning model designed to enhance investment decisions within the data-intensive energy industry. It was there they encountered Brannin McBee, the founder of the data firm they utilized.

From Energy to Cryptocurrency

Following the surge in U.S. fracking activity, Hudson Ridge was dissolved, leaving the team with considerable free time, as Venturo explained.

Their attention then turned to the emerging field of cryptocurrency. Before entering the market, they sought a deeper understanding of the underlying infrastructure.

“We wanted to understand from the commodity side, how is this made,” Venturo stated. “So we started doing mining on the pool table in our Manhattan office.”

This initial foray into cryptocurrency mining laid the groundwork for CoreWeave’s future development and specialization in providing infrastructure for computationally intensive tasks.

A Rapidly Expanding GPU Infrastructure

The initial acquisition of GPUs followed a pattern of exponential growth, much like a familiar snack. Starting with a single unit, the number quickly escalated to ten, then to a thousand. The mining operations initially occupied modest spaces, such as a pool table and a closet.

He recounted the progression with humor, noting they eventually found themselves in a predictably stereotypical location: his grandfather’s garage in New Jersey. Further interest from contacts within the financial sector prompted additional purchases.

“For approximately two and a half years, we held the position of the world’s largest Ethereum miner,” he stated. “At its peak, our operation encompassed 50,000 Nvidia consumer GPUs.”

These GPUs were originally designed for gaming on personal computers, not for continuous, demanding operation within “a warehouse lacking both air conditioning and proper ventilation,” he explained. Consequently, the co-founders developed “sophisticated automation and health-monitoring systems” to manage these less robust GPUs in challenging conditions.

The team recognized the potential to repurpose their substantial GPU resources for alternative applications, specifically AI training. However, they also acknowledged the need to acquire the necessary expertise.

A collaboration was established with EleutherAI, an open-source collective focused on developing a large language model. CoreWeave provided GPU access in return for guidance on AI training, and a partnership was formally announced in 2022.

“Our initial expectation was simply to gain an understanding of the infrastructure’s functionality,” Venturo explained. However, EleutherAI’s network included numerous individuals creating AI startups, and this connection “proved to be a pivotal moment for us.”

The positive relationships fostered through the EleutherAI collaboration resulted in these startups becoming paying clients. Venturo described this development as “pure chance [that] initiated our training business.”

CoreWeave’s capabilities came to the attention of Stability AI through their association with EleutherAI, leading to a customer relationship. The founders then sought additional funding to enhance their infrastructure.

During a dinner meeting with investors from Magnetar, Venturo passionately advocated for the future of AI, “literally pounding on the dinner table” to convey his conviction. This resulted in Magnetar providing a $100 million investment.

The Rise of CoreWeave Fueled by Open Source

CoreWeave initially gained recognition through its collaborations within the open source community, which subsequently brought it to the attention of OpenAI.

Subsequently, Microsoft discovered CoreWeave via its relationship with OpenAI.

At that time, Microsoft’s position as CoreWeave’s largest client was a direct result of being OpenAI’s primary investor and exclusive cloud services provider.

Shifting Dynamics and a Major Deal

This dynamic has since evolved.

Recently, OpenAI formalized a substantial agreement with CoreWeave valued at $12 billion.

This new partnership has resulted in Microsoft no longer holding the position of CoreWeave’s largest customer.

Infrastructure and Capabilities

Currently, CoreWeave operates a network of 32 data centers and boasts an impressive inventory of 250,000 GPUs.

This includes access to Nvidia’s highly sought-after Blackwell chips, which the company states are crucial for advanced AI reasoning capabilities.

Addressing Debt Concerns

Venturo recognizes the significant attention given to CoreWeave’s substantial debt of $7.6 billion, with a considerable portion due within the next two years, as reported by the Financial Times.

Considering CoreWeave’s revenue of $1.9 billion – despite having $15 billion in contracted work – this debt has understandably prompted investor caution.

However, Venturo emphasized that each client agreement is structured to specifically cover the debt incurred for the acquisition of the necessary GPUs.

A Fortuitous Journey

He also acknowledged the remarkable series of fortunate events that have propelled the company forward.

“The sheer number of lucky breaks along the way is astonishing,” he stated.

The company’s origins lie with three hedge fund managers who transitioned from crypto mining to establishing a prominent AI training infrastructure.