In the rapidly evolving landscape of artificial intelligence, companies like \u00a0Coreweave\u00a0, \u00a0Lambda Labs\u00a0, \u00a0Crusoe\u00a0, and \u00a0Nebius\u00a0 have emerged as prominent players. These organizations represent the most booming yet fragile segment of the AI value chain: Neoclouds<\/em>. They have successfully raised tens of billions in capital and debt to construct sprawling data centers equipped with \u00a0NVIDIA GPUs\u00a0. However, their business model hinges on an increasingly questionable assumption: the belief that the demand for computing capacity will continue to skyrocket exponentially.<\/p>\n The concern extends beyond mere financial losses. The entire infrastructure relies on a \u00a0vicious cycle\u00a0:<\/p>\n Neoclouds promised rapid GPU infrastructure deployment\u2014often within months rather than years\u2014with prices up to \u00a066% lower\u00a0 than \u00a0AWS\u00a0, \u00a0Azure\u00a0, or \u00a0Google Cloud\u00a0. As companies scrambled for GPU resources, especially as traditional \u00a0Hyperscalers\u00a0 failed to meet demand, the market initially responded with enthusiasm. For instance, \u00a0Coreweave\u00a0 soared from a mere \u00a0$16 million\u00a0 in revenue in 2022 to an astonishing \u00a0$5.35 billion\u00a0 last year. Similarly, \u00a0Nebius\u00a0\u2014partly rooted in \u00a0Yandex\u00a0\u2014jumped in quarterly earnings from \u00a0$5 million\u00a0 to \u00a0$105 million\u00a0. Overall, the Neocloud segment has experienced astonishing growth, averaging \u00a082% annually\u00a0 over the past four years.<\/p>\n Despite the impressive numbers, the \u00a0revenue stream\u00a0 for many Neoclouds is unusually concentrated. For example, \u00a0Coreweave\u00a0 derived a staggering \u00a060%\u00a0 of its anticipated 2024 revenues from Microsoft, primarily through its partnership with \u00a0OpenAI\u00a0. If you subtract the massive contributions from Microsoft’s projects and similar payouts from \u00a0NVIDIA\u00a0 (an additional \u00a015%\u00a0), the combined revenue across the main Neoclouds dwindles down to approximately \u00a0$1 billion\u00a0. This suggests severe dependency on a limited clientele, as analysts, including \u00a0Ed Zitron\u00a0, have pointed out.<\/p>\n Another significant concern is the financial health of these companies. \u00a0Coreweave\u00a0 reportedly has \u00a0$25 billion\u00a0 in debt and generated revenues of only \u00a0$5.35 billion\u00a0\u2014a staggering \u00a0debt-to-asset ratio\u00a0 of \u00a085.4%\u00a0. This figure is akin to having a mortgage that’s double your annual income, exacerbated by the rapid depreciation of GPUs, unlike stable property values.<\/p>\n Meanwhile, \u00a0Nebius\u00a0 has secured a \u00a0$4.2 billion round\u00a0 to bolster its infrastructure supporting its lucrative contract with Microsoft. Similarly, both Lambda Labs and Crusoe have raised substantial funding through various channels. Their operational blueprint generally follows this pattern:<\/p>\n The challenges arise when primary clients like Microsoft decide to scale back their demands or when Neoclouds struggle to build their infrastructure quickly enough.<\/p>\n \u00a0NVIDIA\u00a0 not only invests directly in several Neoclouds but is also a pivotal player in their supply chain and client base. Just recently, \u00a0Coreweave\u00a0 inked a \u00a0$6.3 billion\u00a0 deal with NVIDIA to guarantee more capacity until 2032. This arrangement creates a particularly intricate scenario, making NVIDIA essential for sustaining growth across many Neoclouds, which in turn rely on NVIDIA for fulfilling client contracts.<\/p>\n Establishing a \u00a0Gigawatt Data Center\u00a0 is financially demanding, costing anywhere between \u00a0$32.5 billion and $50 billion\u00a0. Companies like Oracle and Crusoe took about \u00a02.5 years\u00a0 to complete such a project for OpenAI. Nebius, however, has promised to deliver multiple gigawatts relatively quickly, albeit under increasingly unrealistic conditions.<\/p>\n Recent reports indicate that Coreweave has experienced significant operating losses while simultaneously reporting explosive growth in income. Moreover, Nebius forecasts reaching \u00a0$1.1 billion\u00a0 in annual recurring revenues by the end of 2025, significantly influenced by its deal with Microsoft. Analysts are left to ponder critical questions: What happens if Microsoft decides to invest in its capacity? What if OpenAI, reliant on such immense operations, faces significant setbacks?<\/p>\n Consolidation is already in motion. Recently, Coreweave acquired Core Scientific for \u00a0$9 billion\u00a0 in shares. Industry experts anticipate that only a few formidable players will emerge unscathed from this turbulent environment. As the inevitable adjustment approaches, the pressing question remains: how devastated will these companies be when the reality of GPU capacity demand inevitably clashes with their heavily leveraged financial structures?<\/p>\nThe Importance of Understanding Neoclouds<\/h2>\n
\n
Understanding the Business Model<\/h2>\n
The Risks of Dependency on a Single Client<\/h2>\n
Tracing the Financial Pathway<\/h2>\n
\n
Complications in the Neocloud Landscape<\/h2>\n
NVIDIA’s Central Role<\/h2>\n
The Cost of Building Infrastructure<\/h2>\n
Signs of Instability<\/h2>\n
The Decisive Moments Ahead<\/h2>\n