In a recent CNBC interview, Mike Intrator, co-founder and CEO of CoreWeave, delved into the burgeoning demand for artificial intelligence (AI) chips, the state of AI infrastructure, and the company’s ambitious growth outlook. As a New Jersey-based startup, CoreWeave is rapidly expanding its data center presence across the United States and Europe to meet the escalating needs of AI applications.

The Relentless Demand for AI Infrastructure

CoreWeave’s CEO noted the skyrocketing demand for AI infrastructure. “The market that we’re seeing and we are very close to the true demand for this infrastructure is relentless,” Intrator stated. “It’s been in a state of severe disequilibrium for the past 2 1/2 years.” He emphasized that the demand shows no signs of slowing down, with clients consistently planning their needs far into the future.

CoreWeave’s planning cycles reflect this high demand, with some companies looking for a year. “They’re basically pinned because we don’t have enough capacity to be able to train our models or serve them through inference,” Intrator explained. The company’s unique capability to deliver infrastructure at the required scale has positioned it as a key player in the AI landscape. “We are one of the few companies in the world that can bring on this kind of infrastructure at the scale that’s required to actually make a difference,” he added.

Managing Risks in a Rapidly Growing Market

Despite the booming market, Intrator acknowledged the inherent risks of overexpansion. “We’re a small company competing against giants, and the consequences of getting out too far over your skis can be catastrophic,” he noted. To mitigate these risks, CoreWeave focuses on “the long poles” such as building physical infrastructure and securing necessary resources like mid-voltage transformers and power. These elements take considerable time to deliver but are essential for the company’s long-term strategy.

The company’s client-driven approach further limits exposure. By allowing clients to request capacity, CoreWeave ensures that its investments are aligned with actual demand. “Our exposure is limited to the long poles,” Intrator reiterated, highlighting the company’s cautious yet strategic expansion approach. “The way that we handle our clients is that they come in and they request capacity from us,” he explained. “This helps us manage our risks effectively.”

The Challenge of Accessing AI Chips

One of the critical challenges in the AI industry is securing access to the necessary chips. Intrator discussed how timelines for chip access are governed by two main factors: the release of new chip generations and the availability of physical infrastructure to support increased compute capacity. “The timelines for access to the chips are being governed by new generations of chips that are coming out and the capacity to be able to have enough physical infrastructure to bring that compute online,” he said.

When asked about the influence of political shifts, such as comments from former President Trump, Intrator was clear that market dynamics and infrastructure capacity are the primary drivers of chip availability. “Political factors can certainly play a role, but the primary drivers are the release of new chip generations and the capacity to support them,” he emphasized.

CoreWeave’s Commitment to NVIDIA Technology

Intrator confirmed that the company primarily delivers what its clients demand, which is overwhelmingly NVIDIA’s hardware. “We allow our clients to lead us,” he said. We are one of the true resources for the delivery of NVIDIA’s infrastructure.”

While CoreWeave evaluates different types of silicon-based on performance, the client preference for NVIDIA is clear. “Overwhelmingly, our clients want NVIDIA technology,” Intrator noted, emphasizing the company’s commitment to meeting this demand despite exploring other potential silicon options. “Our clients may be self-selecting, knowing that we are good with NVIDIA technology and bringing it to us for their builds,” he added.

The Future of AI Infrastructure and Energy Consumption

As the demand for AI chips and infrastructure continues to grow, the energy required to power these systems has become a crucial consideration. Intrator addressed concerns about the long-term sustainability of AI infrastructure, particularly regarding energy consumption and processing power. “We think about this a lot,” he said. “We try to focus on securing the necessary resources to ensure that we can support the growing demand without overextending ourselves.”

CoreWeave aims to build a resilient and sustainable AI infrastructure by focusing on long-term planning and client-driven capacity requests. “We are taking a strategic approach to ensure that we can continue to meet demand without compromising on our core principles of sustainability and efficiency,” Intrator stated.

Looking Ahead: CoreWeave’s Growth and Industry Impact

CoreWeave’s rapid expansion and strategic planning have positioned it as a formidable competitor in the AI infrastructure market. The company’s focus on client needs and its cautious yet ambitious growth strategy reflects a broader trend in the tech industry towards specialized, scalable solutions for AI applications. “We are not just meeting current demand; we are paving the way for future advancements in AI technology,” Intrator said.

As CoreWeave continues to build data centers across the United States and Europe, the startup is meeting current demand and preparing for future growth. “We are building a foundation to support the next generation of AI applications,” Intrator concluded. His insights underscore the dynamic nature of the AI industry and the critical role that infrastructure providers like CoreWeave play in its evolution.

With AI applications set to become even more integral to various sectors, CoreWeave’s strategic approach and focus on delivering high-performance infrastructure will likely keep it at the forefront of the industry. As Intrator aptly put it, “We are one of the few companies in the world that can bring on this kind of infrastructure at the scale that’s required to actually make a difference.”