CoreWeave, a company once hailed as the poster child for AI infrastructure growth, finds itself at the center of one of the most hotly debated stories in technology investing. The tale is as much about ambition as it is about risk, pitting the company’s breakneck expansion and headline-grabbing partnerships against the cold realities of mounting debt and a fiercely competitive landscape.
Just last June, CoreWeave’s stock soared to an all-time high of $187 per share, a figure that seemed to confirm its status as a darling of the AI revolution. Fast forward to March 2026, and the mood has shifted dramatically: shares now trade near $81, a plunge of over 56%, according to Bassyonni. Some sources peg the decline at closer to 60%. Either way, the drop is steep, and it’s arrived at a moment when the company’s revenue numbers are, paradoxically, surging.
CoreWeave’s business model is simple in concept but complex in execution. The company builds out massive data centers, fills them with Nvidia’s latest graphics processing units (GPUs), and rents out that computing power to clients running artificial intelligence workloads. This setup allows customers to access cutting-edge hardware without the massive upfront investment, a compelling proposition as demand for AI compute capacity explodes. But there’s a catch: staying competitive means constantly refreshing hardware, which requires continuous—and enormous—capital investment.
Unlike established cloud giants with diversified revenue streams, CoreWeave doesn’t have other profitable business lines to help subsidize its expansion. Instead, it relies heavily on external funding to fuel its growth. That reliance has become a flashpoint for investors, especially as the company’s losses have widened and its free cash flow remains deeply negative. In the fourth quarter of 2025, CoreWeave’s net losses increased sharply, and its capital expenditures are projected to surge to between $30 billion and $35 billion in 2026, up from $14.9 billion the previous year. Interest expenses alone are expected to land between $510 million and $590 million in just the first quarter of 2026.
Despite these daunting figures, CoreWeave keeps notching strategic wins that would make most tech CEOs envious. In late January, Nvidia deepened its partnership with a $2 billion investment in CoreWeave’s Class A shares at $87.20 each. The move not only shored up CoreWeave’s balance sheet but also cemented a technology alliance that will see CoreWeave deploy Nvidia’s Rubin platform, Vera CPUs, and future architectures. The goal? To build out more than 5 gigawatts of AI-optimized infrastructure by 2030—a target that, if met, would put CoreWeave in rare company.
CoreWeave CEO Michael Intrator was quick to dismiss suggestions that this partnership amounted to circular financing, calling such claims “ridiculous.” Nvidia CEO Jensen Huang also pushed back against critics who argued that the company was simply propping up its own chip demand. Instead, both leaders framed the deal as a validation of CoreWeave’s business model and a way to future-proof the platform against competition from tech giants like Microsoft and Meta.
Not long after the Nvidia deal, CoreWeave announced a multi-year strategic agreement to power Perplexity’s AI inference workloads using Nvidia GB200 NVL72 clusters. This partnership is particularly significant because inference—the process of running AI models to generate answers for users—represents a recurring revenue stream, as opposed to the one-off nature of model training. CoreWeave’s bare-metal, GPU-first architecture gives it an edge here, eliminating virtualization overhead and allowing for direct hardware access. The Perplexity deal isn’t just a feather in CoreWeave’s cap; it’s a signal that the company is pivoting toward what many see as the “new gold rush” in AI infrastructure.
But for every bullish data point, there’s a bearish counterpoint. Critics argue that CoreWeave’s growth engine is running on borrowed momentum. Nvidia releases new generations of hardware on roughly an annual cycle, meaning that equipment installed today could feel outdated to some clients within a year. GPUs used under intensive workloads may wear out within one to three years, requiring replacement sooner than traditional depreciation schedules would suggest. This relentless refresh cycle only compounds the capital demands of the business.
Then there’s the question of profitability. CoreWeave’s fourth-quarter 2025 revenue rose an eye-popping 110% year over year, hitting $1.6 billion. The company’s revenue backlog climbed to nearly $70 billion—a 342% jump compared to the same period a year earlier. Approximately 42% of that backlog is expected to convert to actual revenue within the next two years, giving the growth narrative some near-term substance. Management projects an annualized revenue run rate of $17 billion to $19 billion by the end of 2026, more than tripling the current base. These numbers are impressive, but they don’t automatically equate to profitability.
What the market is really waiting for is clarity on a deceptively simple question: Can CoreWeave’s annual capital expenditures stay below its operating profits as revenue scales? If the answer is yes, the company could emerge as a durable player in the AI cloud infrastructure space. If not, concerns about sustainability and valuation will only deepen.
Lost amid the handwringing over debt and capex is CoreWeave’s enormous contracted revenue backlog—$66.8 billion as of March 2026, secured across roughly five-year average contract lengths. These pre-sold commitments, anchored by landmark deals with OpenAI, Meta, Microsoft, and others, essentially de-risk much of the company’s aggressive spending. Every dollar invested today is already matched to future cash flows from long-term customers who need those GPUs now and for years to come.
Still, the debate rages on. Some investors see the recent share price decline as a rare buying opportunity, arguing that the market is missing the forest for the trees. The same growth drivers—Nvidia’s partnership, inference deals like Perplexity, and a fortress-like backlog—are exactly what could service and ultimately sustain CoreWeave’s investments as revenue scales. Others remain cautious, pointing out that a backlog is a measure of demand, not profitability, and that the economics of delivering AI compute at scale remain unproven for newcomers like CoreWeave.
For now, CoreWeave sits at a crossroads. The company is flush with demand and strategic partnerships, but it’s also navigating a minefield of financial risks and operational challenges. The next year or two will be critical in determining whether CoreWeave can convert its backlog and momentum into lasting profitability—or if it will become another cautionary tale in the annals of tech investing.
With billions already committed and the AI infrastructure race heating up, CoreWeave’s journey is far from over. Investors and industry watchers alike will be keeping a close eye on whether this high-wire act can stick the landing.