As organizations rapidly adopt AI technologies, the demand on cloud providers is intensifying. Major players like Amazon Web Services, Microsoft, Google, and Oracle are facing an extraordinary surge in demand for the server chips necessary to train and deploy AI-driven applications, leading them to limit chip availability for their customers. Microsoft has openly acknowledged these challenges, cautioning that service interruptions could occur if it cannot secure enough AI chips for its data centers.
Startups are also feeling the strain, with companies like CoreWeave, a cloud computing provider specializing in GPUs, under pressure. CoreWeave recently raised $221 million in a Series B funding round in April and an additional $200 million extension in May, bringing its total debt financing to an impressive $2.3 billion.
This credit facility, led by prominent investors such as Blackstone and Magnetar Capital and supported by Coatue, DigitalBridge Credit, and funds managed by Pimco and Carlyle, comes on the heels of CoreWeave's announcement to construct a $1.6 billion data center in Plano, Texas. In a blog, CoreWeave’s co-founder and CEO, Michael Intrator, noted that the debt facility will offer “financial headroom and flexibility” to reach its ambitious goal of establishing 14 data centers by year-end.
“We’ll allocate the loan entirely toward purchasing hardware for contracts we’ve already secured with clients and continue attracting top industry talent," Intrator stated. "The demand for GPU compute has far exceeded expectations, but our strategic investments in capacity are proving successful — we’re delivering solutions where others fall short."
Founded in 2017 by Intrator, Brian Venturo, and Brannin McBee, CoreWeave was established to fill a perceived gap in the cloud services marketplace. Venturo, a passionate Ethereum miner, strategically acquired GPUs from bankrupt cryptocurrency mining operations, primarily opting for Nvidia hardware due to its superior memory capabilities.
Initially dedicated to cryptocurrency applications, CoreWeave pivoted in recent years to encompass general-purpose computing and generative AI technologies, such as AI models for text generation. The company’s robust GPU infrastructure is perfectly suited for the intensive requirements of training today’s advanced models. Nvidia has also seen tremendous growth, briefly reaching a market capitalization of $1 trillion.
Today, CoreWeave offers access to a diverse range of Nvidia GPUs in the cloud, including H100s, A100s, A40s, and RTX A6000s, catering to AI, machine learning, visual effects, rendering, batch processing, and pixel streaming applications. CoreWeave is further leveraging its infrastructure for unique projects, including an “AI supercomputer” consisting of over 3,500 H100s launched in collaboration with Nvidia last month.
Competing against established cloud giants like AWS, Google, and Microsoft is no small feat. For context, AWS generated $80.1 billion in revenue last year, while Google Cloud and Azure brought in $75.3 billion and $26.28 billion, respectively. These figures significantly overshadow CoreWeave’s valuation of approximately $2 billion, let alone its available funds of $576.5 million.
To highlight the competitive landscape, a Statista report from Q4 2022 indicated that AWS holds a 32% market share, Azure commands a 23% share, and Google Cloud accounts for 10%. Adding to the challenges, many companies are exploring cost-cutting measures in the face of economic uncertainty. In 2023, the cloud infrastructure market experienced a growth slowdown to 21%, a stark decline from 36% growth the previous year.
This environment has prompted consolidation within the industry. In July, DigitalOcean acquired Paperspace, a New York-based cloud computing and AI startup, for $111 million in cash. However, opportunities for smaller players do exist, as evidenced by companies like Scaleway, Clever Cloud, and Vultr. CoreWeave has also intensified its focus on infrastructure to support the booming generative AI sector. According to CoreWeave’s CTO, Brian Venturo, the company’s new data centers can house approximately 20,000 GPUs at a single site, surpassing the traditional offerings of other cloud providers.
"The escalating demand for computation driven by generative AI necessitates significant investment in specialized GPU cloud infrastructure, an area in which CoreWeave is clearly positioned as a leader in fostering innovation," stated Jasvinder Khaira, a senior managing director at Blackstone.
While tech giants invest heavily in developing in-house supercomputers and AI chips, smaller companies are increasingly turning to cloud providers like CoreWeave. According to PitchBook data, approximately $1.7 billion was raised across 46 deals for generative AI startups in Q1 2023, with an additional $10.68 billion in deals announced during the quarter awaiting completion.
One such startup is Inflection AI, co-founded by DeepMind veteran Mustafa Suleyman, which trained its AI assistant, Pi, using CoreWeave's infrastructure. Inflection is now collaborating with Nvidia and CoreWeave to develop an AI training cluster composed of 22,000 H100 GPUs.
In addition to its robust infrastructure, CoreWeave seeks to distinguish itself through initiatives like its accelerator program, launched in late October. This program operates on an ongoing basis, offering companies compute credits, discounts, and other hardware resources within the CoreWeave cloud ecosystem.
Currently, CoreWeave employs just over 115 individuals, a growth spurt of 150% in the past year, in part due to its acquisition of cloud rendering platform Conductor Technologies in January. Intrator emphasized a commitment to continued hiring throughout the year, bolstered by the recent debt financing.
Ranking
