CoreWeave has greater scale and a solid track record, but its heavy reliance on debt makes its growth more expensive and risky.
Nebius is less proven, but its infrastructure-partnership model could let it scale up its AI compute footprint with less capital.
I’d favor Nebius as a long-term investment because it has a more capital-efficient path to capturing demand for AI infrastructure.
The artificial intelligence (AI) trade has moved well beyond processor chips. Someone still has to install the graphics processing units (GPUs) and central processing units (CPUs), connect them via high-speed networks, cool them, store data, and keep thousands of accelerators running when customers need them.
That need has led to the creation of a new class of AI infrastructure companies, and CoreWeave (NASDAQ: CRWV) and Nebius Group (NASDAQ: NBIS) are two of the more interesting public-market names. Both are building AI clouds around Nvidia (NASDAQ: NVDA) hardware, but their strategies look quite different.
Missed Nvidia in 2009? This Rare Signal Is Flashing Again. In 2009, a "Double Down" signal flashed for a little-known chipmaker called Nvidia. For the first time in years, that same "Total Conviction" signal is flashing for a company 1/100th the size of Nvidia. Continue »
That difference matters to investors because AI compute is a capital-intensive business.
CoreWeave has taken the straightforward approach: Acquire infrastructure and the electricity to power it, load it with cutting-edge Nvidia processors, and sell the resulting compute capacity to AI companies.
The company had about 1.5 gigawatts (GW) of active power capacity and roughly 3.7 GW of contracted power as of June. It also became the first AI cloud provider to bring up and validate Nvidia's new Vera Rubin NVL72 system. NVL72 is a rack-scale system in which 72 GPUs work together with high-bandwidth networking and other components, making the data center architecture itself part of the product.
CoreWeave is also moving deeper into the software layer. Its SUNK platform is designed to simplify the deployment and management of large AI clusters, while its newer cross-cloud products let customers move workloads and data between CoreWeave and other cloud environments. That is important because the long-term value in AI infrastructure may not come from simply renting GPU hours but from managing the complicated systems around those GPUs.
Image source: Getty Images.
The biggest concern for investors is the capital required to keep doing this. CoreWeave has raised billions through infrastructure-backed financing facilities, including an $8.5 billion facility in March and another $2.6 billion facility in August. Taking on more debt can allow an infrastructure business to grow faster, but investors need to think about how those debt loads impact the company's overall financial picture.
Nebius Group is taking a more flexible approach. It's still investing in its own AI factories and deploying Nvidia's latest systems. It plans to offer Vera Rubin NVL72 capacity in the U.S. and Europe, while its U.K. expansion includes three new NVIDIA-powered deployments expected to reach 65 megawatts (MW) at full capacity.
But the move that stands out is Nebius' infrastructure-partnership model. Under the structure announced in July, its infrastructure partners will finance, own, and operate the physical data centers while Nebius supplies the systems architecture, hardware design, software, and services stack, then sells the resulting capacity to customers. Nebius says this business model can create a higher-margin revenue stream with less incremental capital required from the company itself.
That could become an important advantage. Nebius does not necessarily need to own every building in which its software and AI cloud operate. It could become a provider of the layer connecting infrastructure capital to AI demand.
The company already has a major customer relationship to build around. In March, Meta Platforms (NASDAQ: META) agreed to a five-year, $12 billion dedicated-capacity arrangement, with another commitment that could bring its total compute purchases to $15 billion over five years. Nvidia has also agreed to invest $2 billion in Nebius, with the companies targeting more than 5 GW of Nvidia systems by the end of 2030.
For me, Nebius is the more interesting stock, though it's also the less proven one. CoreWeave has demonstrated that customers will pay for its infrastructure, and its greater scale gives it a major head start. The company is also securing enormous amounts of power and bringing new Nvidia architectures online before many of its rivals do the same.
Nebius, however, is attempting something that could matter more over a longer period: separating the AI cloud platform from the capital required to build every piece of physical infrastructure. I would rather own a piece of a company that is figuring out how to scale up AI compute infrastructure without all of the capital costs sitting on its own balance sheet.
Before you buy stock in Nebius Group, consider this:
The Motley Fool Stock Advisor analyst team just identified what they believe are the 10 best stocks for investors to buy now… and Nebius Group wasn’t one of them. The 10 stocks that made the cut could produce monster returns in the coming years.
Consider when Netflix made this list on December 17, 2004... if you invested $1,000 at the time of our recommendation, you’d have $446,157!* Or when Nvidia made this list on April 15, 2005... if you invested $1,000 at the time of our recommendation, you’d have $1,377,357!*
Now, it’s worth noting Stock Advisor’s total average return is 983% — a market-crushing outperformance compared to 212% for the S&P 500. Don't miss the latest top 10 list, available with Stock Advisor, and join an investing community built by individual investors for individual investors.
See the 10 stocks »
*Stock Advisor returns as of September 4, 2026.
Micah Zimmerman has no position in any of the stocks mentioned. The Motley Fool has positions in and recommends Meta Platforms and Nvidia. The Motley Fool has a disclosure policy.