When choosing where to allocate capital in the AI infrastructure super-cycle, the temptation to back a rising star like CoreWeave is undeniable. On the surface, it looks like the quintessential high-growth disruptor—born out of crypto mining, pivoted into GPU cloud services, and now positioned as the “AI-first cloud” with deep Nvidia backing. Yet when I weighed my options, I chose to buy Nvidia stock rather than CoreWeave. The reasoning lies in durability, scale, and risk-adjusted return, and a close look at CoreWeave’s strengths, weaknesses, opportunities, and threats only reinforced that decision.
CoreWeave’s strengths are real. The company is laser-focused on GPU-intensive workloads, which gives it an edge over hyperscalers like AWS or Azure that still juggle broad portfolios of services. Its close partnership with Nvidia ensures priority access to cutting-edge hardware like H100 and Blackwell GPUs, and its leaner, purpose-built infrastructure allows it to provision compute faster and cheaper for AI developers. In the middle of a GPU shortage, this agility has made CoreWeave a lifeline for startups and enterprises desperate for compute. For a company still relatively young, the fact that it has carved out this position against trillion-dollar hyperscalers is impressive.
But the weaknesses cannot be ignored, and they ultimately tipped my hand toward Nvidia. CoreWeave is dependent on capital-intensive expansion—leasing GPUs, building data centers, and raising money to finance it all. Its product scope is narrow: customers may rent GPUs from CoreWeave, but they still rely on AWS, Microsoft, or Google for the rest of their workloads. That creates dependency risk and leaves CoreWeave vulnerable to being squeezed into a specialized niche. Most critically, CoreWeave’s fate is deeply tied to Nvidia’s pricing power and supply pipeline. Nvidia controls the hardware roadmap, sets margins, and has the leverage to capture most of the value in the stack. By buying Nvidia directly, I skip the dependency and own the very company dictating the terms of CoreWeave’s existence.
Opportunities for CoreWeave are vast—it could become the “fourth cloud,” expand internationally, or branch into inference hosting and managed AI services. The demand for compute is not a passing fad, and CoreWeave could ride this wave to scale rapidly. But here again, Nvidia stands to capture the broader upside. Whether CoreWeave or another challenger wins market share, every rack of GPUs they deploy is built on Nvidia silicon. It doesn’t matter whether the customer rents from CoreWeave, Azure, or a research consortium in Germany—Nvidia benefits. CoreWeave is a levered bet on one execution path, Nvidia is the toll booth on all of them.
The threats to CoreWeave are sharper. Hyperscalers are racing to build their own GPU fleets, while also investing in custom silicon like AWS Trainium and Google TPU to lessen reliance on Nvidia. If these alternatives gain traction, CoreWeave could be outcompeted both on cost and scope of services. Its debt-heavy model makes it sensitive to higher rates, and its narrow dependence on Nvidia could become a choke point if regulation, antitrust, or export controls shift the supply chain. Nvidia, by contrast, is the bottleneck itself. It is not hostage to any single customer and maintains leverage across the ecosystem.
This is why my capital went to Nvidia. It is the foundation of the AI factory build-out, selling the picks and shovels that every player—CoreWeave included—depends on. CoreWeave may grow faster in percentage terms if it executes flawlessly, but Nvidia captures the super-cycle no matter which cloud or startup wins. In a world of uncertain execution, customer concentration, and financing risks, Nvidia offers not just growth but durability. CoreWeave is a fascinating bet on specialization; Nvidia is the meta-bet on the entire AI era.