Last month, U.S. President Donald Trump brought together 33 of Silicon Valley’s most influential leaders to address a burgeoning challenge: How to meet insatiable demand for data-processing power in the age of artificial intelligence?
It’s a question worthy of such a summit. Even cloud service giants like Amazon, Google and Microsoft are hitting data-processing bottlenecks. While those so-called hyper-scalers continue to invest heavily in custom chips, power systems and next-generation data-center architecture, no single player can meet the scale and complexity of AI’s demands alone. Instead, tech’s titans are now looking to forge strategic partnerships with a new breed of “neo-cloud” providers.
Unlike the hyper-scalers, which offer a broad range of cloud solutions, neo-cloud providers simply rent raw computing power for AI and advanced applications—a business model known as “GPUs-as-a-service.” Instead of relying on massive, centralized data centers, these scrappy operators deploy their data-crunching power across thousands of smaller centers or regional hubs, increasing the flexibility and resilience of their networks—not unlike high-frequency traders who aim to shave off precious milliseconds by placing their servers closer to securities exchanges.
By partnering with neo-cloud newcomers, such as Nebius and CoreWeave, behemoths like Microsoft aim to handle daunting AI workloads at increasing scale. For example, Microsoft recently signed a deal with Nebius—worth up to $20 billion over five years—under which Microsoft will lease computing power from Nebius’ new data center in Vineland, New Jersey.1
Meanwhile, leading AI chipmaker Nvidia has taken a significant equity stake in CoreWeave, worth roughly $3 billion at the end of June, and has also agreed to purchase up to $6.3 billion in unsold data-processing capacity from the company through the spring of 2032.2 (Nvidia is a shareholder in Nebius as well.)
For smaller enterprises, we believe neo-clouds have the potential to lower the barriers to entry for high-performance AI and advanced analytics. Thanks to their streamlined operations, neo-cloud providers can rent their GPUs at an hourly rate between two and seven times lower than what hyper-scalers typically charge,3 potentially allowing young businesses to get on their feet without having to make large upfront investments or long-term commitments in cutting-edge technology.
Furthermore, neo-clouds can offer industry-specific data-processing power, potentially giving them an additional edge over the hyper-scalers. For example, Nebius offers cloud services expressly designed to support the e-commerce industry—a big reason it recently landed Shopify as a client.4
Yet the benefits of neo-clouds extend well beyond technical agility, in our view. Amid broader trends toward de-globalization and regional data sovereignty, we believe newer cloud-service players may be better positioned than the hyper-scalers—and perhaps more trusted—to comply with local data-governance and environmental mandates, including Europe’s General Data Protection Regulation, Artificial Intelligence Act, and Digital Operation Resilience Act. CoreWeave, for one, aims to build advanced AI “factories” in the UK powered by renewable energy and advanced cooling technology to limit water consumption.5
Over the next five years, the total addressable GPU-as-a-service market is expected to grow at 35% a year, reaching $260 billion by 2030.6 We will be keeping an eye on those providers with strong relationships with chipmakers and hyper-scalers, as well as those with differentiated software stacks catering to specific industries. For thematic tech investors, we believe the forecast calls for a strong chance of neo-clouds.
