Research Finder
Find by Keyword
CoreWeave Becomes Fastest Growing Cloud Provider Ever
Explosive 110% revenue growth and a $67 billion backlog collide with a steep EPS miss and $8.2 billion in quarterly capital spending.
02/27/2026
By the Numbers
- Revenue: $1.6 billion (up 110% YoY)
- FY2025 Revenue: $5.1 billion (up 168% YoY)
- Diluted EPS: -$0.89 (vs. -$0.21 consensus estimate)
- Contracted Backlog: $66.8 billion
- Q4 CapEx: $8.2 billion
Key Highlights
- CoreWeave is now the fastest cloud provider in history to scale from zero to $5 billion in annual revenue.
- The massive $66.8 billion backlog provides multi-year visibility but requires eye-watering capital to fulfill.
- Profitability remains a distant target as infrastructure investments and interest expenses on debt continue to balloon.
- Strategic wins with Meta and federal sector expansion aim to diversify the customer base beyond AI startups.
The News
CoreWeave recently reported its fourth quarter and full-year 2025 financial results, highlighting a period of hypergrowth tempered by significant bottom-line pressure. The company exceeded revenue expectations but posted a wider-than-anticipated loss of $0.89 per share due to aggressive infrastructure build-outs. Management also provided ambitious 2026 guidance, targeting revenue between $12 billion and $13 billion. To find out more, visit the CoreWeave Investor Relations page.
Analyst Take
We have spent the last few days digging into CoreWeave’s latest numbers, and I also went on Bloomberg to give my predictions and what i was looking for ahead of the Earnings print, and the picture is as complex as the GPU clusters they build. On one hand, you have a company that is essentially the "canary in the coal mine" for the AI era; if they are growing, the AI revolution is alive and well. On the other hand, the financial model looks increasingly like a high-stakes construction project where the cranes never stop moving. My analysis suggests that while the top-line growth is staggering, the "cost of admission" to the elite tier of cloud providers is rising faster than many anticipated.
What was Announced
CoreWeave used this earnings cycle to showcase several major technical and strategic shifts designed to move the company beyond being just a "landlord for GPUs." They officially launched CoreWeave Federal, a dedicated environment architected to meet the security and compliance needs of government agencies. This is a smart move to reduce reliance on volatile VC-backed AI startups. They also achieved NVIDIA Exemplar Cloud status for the GB200 NVL72 systems, which is effectively a seal of approval for their ability to handle the next generation of Blackwell chips. Perhaps most interestingly, they introduced CoreWeave ARENA, a proprietary orchestration layer that aims to deliver better resource management for massive distributed training jobs. This software play is clearly intended to create "stickiness" so customers don't just flee to the lowest-cost provider.
Looking at the numbers from SeekingAlpha and Barrons, it is clear the market is wrestling with the sheer scale of the spending. A $1.6 billion revenue quarter is impressive, but when you spend $8.2 billion in CapEx in those same three months, you are essentially betting the entire house on future demand. The interest expense alone reached $388 million this quarter. That is a heavy anchor to carry. However, the $66.8 billion backlog is the ultimate counter-argument. If that backlog is durable, then the spending isn't just vanity; it is the necessary plumbing for a contracted future.
The divergence between the GAAP results and management's adjusted figures is worth noting. While Adjusted EBITDA margins look healthy at 57%, the GAAP net loss widened to $452 million. This tells me that the "real" cost of running this business, including the massive depreciation of hardware that might become obsolete in three years, is starting to bite. we are observing a trend where the "hyperscaler" title is no longer just for Amazon or Google. CoreWeave is operating at that same level of capital intensity, but without the benefit of a high-margin advertising or retail business to subsidize the build-out.
The secondary market for AI services is shifting toward specialization. CoreWeave’s focus on being "purpose-built" for AI is their primary defense. They aren't trying to host your email or your corporate website; they are architected to run LLMs at scale. This specialization allowed them to beat the big three clouds to the punch on certain NVIDIA deployments. However, as the giants catch up, CoreWeave's lead will have to depend more on that ARENA software layer and less on just having the most chips.
Customer Concentration Concerns Remain
CoreWeave’s customer concentration has historically been its greatest strategic vulnerability, as Microsoft and OpenAI accounted for a staggering 77% of total revenue as recently as late 2024. While the partnership with Microsoft provided the initial multi-billion-dollar bedrock for CoreWeave’s existence, relying on a single hyperscaler for nearly two-thirds of sales created a precarious "single point of failure" risk. To mitigate this, CoreWeave has aggressively diversified its $66.8 billion backlog, securing a massive $14.2 billion capacity deal with Meta and an expansive $22.4 billion commitment from OpenAI that extends through the end of the decade. The recent $4.8 billion sovereign AI agreement with Saudi Arabia’s Level42 marks a critical shift toward international diversification, tapping into low-cost energy and regional state-backed funding to balance its US-centric book of business. This Level42 deal aims to deliver the Kingdom’s first independent AI cloud, effectively positioning CoreWeave as a "sovereign utility" rather than just a supplemental provider to American big tech. Despite these wins, concentration remains high in the near term, as no single client currently represents more than 35% of the backlog, yet the top five customers still command the vast majority of future revenue.
We remain wary that any shift in capital spending by these few "whales" could lead to a catastrophic revenue shortfall, especially given the company's massive $30 billion capital expenditure plan for 2026. Ultimately, while CoreWeave is architected to serve the largest AI labs on Earth, its long-term stability depends on converting this narrow concentration into a broader, more resilient enterprise customer base.
Looking Ahead
Based on what we are observing, CoreWeave is currently in a "sprint to scale" that leaves very little room for error. The key trend we will be tracking is the conversion rate of the $66.8 billion backlog into recognized revenue. If deployments lag or if customers try to renegotiate contracts as chip supply eases, the debt load could become problematic. Based on our analysis of the market, our perspective is that CoreWeave is successfully transitioning from a niche provider to a legitimate fourth pillar in the cloud landscape.
Going forward, we will be looking for enterprise adoption for inference workloads, rather than just frontier labs training huge LLMs, as another key indicator. If CoreWeave can demonstrate customer and workload diversification, then we expect the growth to continue and concerns to dissipate. The proof will be in the pudding.
We are going to be looking for how the company performs on margin stabilization. When you look at the market as a whole, AI infrastructure build-out is not slowing down, despite some whispers of "AI fatigue" on Wall Street. HyperFRAME will be closely monitoring how the company handles the upcoming Blackwell ramp in future quarters. If they can maintain their pricing power while scaling to the 5GW capacity they’ve promised by 2030, they may well justify this unprecedented burn rate.
Steven Dickens | CEO HyperFRAME Research
Regarded as a luminary at the intersection of technology and business transformation, Steven Dickens is the CEO and Principal Analyst at HyperFRAME Research.
Ranked consistently among the Top 10 Analysts by AR Insights and a contributor to Forbes, Steven's expert perspectives are sought after by tier one media outlets such as The Wall Street Journal and CNBC, and he is a regular on TV networks including the Schwab Network and Bloomberg.