The rapid advancement of AI has made computational power the cornerstone of this technological revolution. Training large-scale language models with billions of parameters or deploying complex AI applications demands unprecedented computing resources.So, where does this immense computational power come from?Traditionally, giants like Amazon, AWS, Google Cloud, and Microsoft come to mind. However, as AI development accelerates, the market is quietly shifting. A new wave of AI infrastructure companies is emerging, challenging traditional cloud giants with unique business models and technologies.These newcomers, dubbed "Neocloud" by chip research firms, are the new force in AI compute supply. Unlike Amazon or Google, which aim to be all-encompassing, Neoclouds focus solely on delivering high-powered compute for AI models. Companies like CoreWeave exemplify this approach, targeting the massive training demands of AI with GPU clusters and rapid order fulfillment. Unlike the slow-moving giants, these nimble players act as AI’s “special forces,” backed by NVIDIA, which benefits from increased chip sales and a tighter ecosystem.Why are Neoclouds breaking through the dominance of Amazon, Google, and Microsoft? The AI infrastructure market is booming, driven by the need for robust compute to support large-scale model training and inference. This includes GPU clusters, custom servers, high-density data centers, and advanced power and cooling systems. IDC predicts that by 2028, over 75% of AI infrastructure spending will go to accelerator-chip-equipped servers, with a 42% annual growth rate.Beyond infrastructure, the broader AI hardware and software market—encompassing chips, cloud platforms, models, applications, APIs, and development platforms—is projected to reach $780–990 billion by 2027, with a 40–55% annual growth rate, according to Bain & Company. Neoclouds bridge the gap between infrastructure and application ecosystems, anchoring themselves at the core of the AI value chain.Governments and capital markets are reinforcing this trend, treating AI infrastructure as a national strategic asset. The U.S. has launched the Stargate Project, investing $500 billion over four years in national compute centers. The EU is crafting cloud infrastructure and digital sovereignty policies, while Saudi Arabia partners with NVIDIA to build a universal cloud and AI supercomputer. These policies demand customized, efficient, and flexible compute solutions—areas where traditional cloud providers struggle but Neoclouds excel.Capital markets reflect this shift. CoreWeave, for example, secured $2.3 billion in debt financing from Blackstone and others for data center expansion and GPU purchases, followed by a $1 billion Series C in May 2024, valuing it at $19 billion. By March 2025, it raised $1.5 billion through an IPO, reaching a $23 billion valuation. Its stock surged over fivefold to above $187, cementing its status as an AI compute unicorn. Other players, like xAI, raised $3.4 billion last year and secured a $750 million credit line in June 2025 to build a supercomputing center with 100,000 AI chips, blending edge energy with sustainable computing.This isn’t just about individual successes—it’s a systemic trend. AI infrastructure is becoming a core asset, driven by governments, capital, and enterprise demand. Among the rising stars is Nbis, a company we’ve tracked since last December when its stock was $28. Recently, we doubled down on it, and here’s why: Nbis is a full-stack AI infrastructure player with a vertically integrated model, from hardware to platforms, designed specifically for AI workloads. Its cloud platform includes a native AI development studio, enabling developers to train and deploy open-source models seamlessly, cutting costs by up to 50% compared to traditional clouds.Nbis strategically builds data centers in locations like Finland and Paris, leveraging natural cooling to reduce energy costs. Its self-developed high-temperature hardware further optimizes efficiency. With a high-speed communication protocol and integrated toolchain, Nbis shortens deployment cycles and boosts R&D efficiency. Industry projections estimate the specialized cloud and AI infrastructure market will grow from $33 billion in 2023 to $260 billion by 2030, with the broader cloud market reaching $2 trillion. Even capturing 1% of this market could yield Nbis $20 billion in annual revenue.Capital markets back Nbis strongly, with a $700 million funding round last December, including NVIDIA as an investor. With over $2 billion in cash reserves, Nbis plans to invest $1 billion in data center expansion by the end of 2025. Its revenue soared from $21 million in 2023 to $117 million in 2024, a nearly sixfold increase, with $200 million in annual recurring revenue already secured for Q1 2025. Management projects $700 million in revenue for 2025, five times 2023’s figure.Valuation-wise, Nbis is a bargain. Assuming $700 million in 2025 revenue and a 40% growth rate, it could hit $2.5 billion by 2028. At a 20x P/E or 8–10x P/S, its market cap could exceed $20 billion. With 70% resource utilization, revenue could reach $7 billion, implying a $35 billion valuation at a 5x P/S ratio. At 90% utilization, revenue could climb to $9–10 billion, far surpassing current expectations.However, risks loom. Nbis’s capital-intensive model relies on external financing, unlike cash-rich giants like Oracle, which leverages its software cash flow for AI infrastructure and serves as a key supplier for OpenAI’s Stargate Project. Customer concentration is another concern—CoreWeave’s $11.9 billion contract with OpenAI through 2030 is at risk if supply falters, as OpenAI can terminate the deal. Similarly, much of CoreWeave’s revenue comes from Microsoft, which may reduce reliance as its own infrastructure scales. HSBC’s “sell” rating on CoreWeave, with a $32 target (77% below current price), cites overreliance on a few clients like Microsoft, OpenAI, and NVIDIA, who treat it as a temporary compute provider.Upstream, Neoclouds depend heavily on NVIDIA’s GPUs, with high-cost chips like the GB200 posing risks if updates render existing hardware obsolete. Traditional giants like Amazon, Google, and Microsoft are developing their own AI chips and leveraging global infrastructure and enterprise relationships. To survive, Neoclouds must maintain efficiency, customization, and supply chain diversification to avoid being squeezed by giants or tethered to NVIDIA.
@TigerObserver @Tiger_comments @TigerStars @Daily_Discussion @TigerPM
Comments