CoreWeave's origin story is one of the more improbable pivots in modern technology history. The company was founded in 2017 in New Jersey by three commodities traders: Michael Intrator, Brian Venturo, and Brannin McBee. The company was called Atlantic Crypto, and its business was exactly what it sounds like — mining Ethereum using racks of GPU hardware. The founders were not engineers by background. They were traders who recognized an opportunity in the emerging crypto market and had the capital and operational know-how to execute at scale.

Then came the 2018 crypto crash. Ethereum prices collapsed, rendering the mining operation economically marginal. The founders were left with an enormous inventory of GPU hardware and a choice: liquidate or pivot. They pivoted. In 2019, the company was renamed CoreWeave and began repurposing its GPU clusters for cloud computing services, initially targeting graphics rendering and other compute-intensive workloads that needed GPU horsepower.

The real inflection point came in late 2022, when OpenAI released ChatGPT. Almost overnight, every technology company, research lab, and enterprise in the world discovered that it needed massive GPU computing capacity — immediately — to build, train, and run AI models. The major cloud providers (AWS, Azure, Google Cloud) were unprepared for the sudden demand spike. CoreWeave, uniquely positioned with large GPU inventories and a Kubernetes-native infrastructure already purpose-built for that kind of workload, became the go-to alternative. In 2022 alone, the company invested roughly $100 million in Nvidia's new H100 chips — a bet that paid off spectacularly as those chips became the most coveted hardware in the world.

How Fast Is "Fast"? CoreWeave went from $143 million in revenue in 2023 to $1.92 billion in 2024 — a 737% increase in a single year. It then grew to $5.13 billion in 2025, a further 168% increase. CEO Michael Intrator described the company as "the fastest cloud in history to reach $5 billion in annual revenue." That is not marketing language. It is, by most measures, factually accurate.

The company went public on March 28, 2025, in what was described as the largest AI-related IPO by amount raised, according to Dealogic. The IPO had a rocky lead-up — CoreWeave reduced its offering size from $2.7 billion to $1.5 billion and priced at $40 per share, below its initial range of $47-$55, amid reports of prior loan covenant breaches and concerns about customer concentration. Despite the rough debut, the stock subsequently surged over 350% in its first few months of trading, hitting an all-time high of $183.58 in June 2025. It has since pulled back significantly. As of March 27, 2026, it trades at approximately $74.81 — about 87% below the all-time high, but nearly double the IPO price.

CoreWeave is not a generalist cloud provider. It is a specialized AI cloud infrastructure company — what the industry now calls a "neocloud" — built from the ground up to run one category of workload extremely well: GPU-intensive AI compute. Training large language models. Running AI inference at scale. High-performance computing. These are computationally brutal tasks that require enormous numbers of GPUs, extremely low latency networking between those GPUs, and specialized software to orchestrate it all efficiently.

The core product is access to Nvidia GPU clusters — rented to customers on long-term contracts, typically structured as take-or-pay agreements where customers commit to pay for capacity regardless of whether they use it. This structure is critical to understanding CoreWeave's financial model. The take-or-pay structure gives CoreWeave committed, predictable revenue over multi-year periods, and it uses those committed contracts as collateral to finance the capital expenditure required to build the data centers that fulfill them. The company builds. Customers pay. The contracts secure the debt. Rinse and repeat — at an ever-increasing scale.

Beyond raw GPU compute, CoreWeave has been expanding its platform. The company acquired Weights & Biases (the widely used AI developer platform for experiment tracking and model management) in early 2025 for approximately $1.7 billion, and has made additional acquisitions in reinforcement learning tooling and physics AI applications. It also launched an object storage service in Q4 2025, beginning to build a broader cloud services stack that moves it slightly toward competing with the hyperscalers on a fuller product basis.

CoreWeave's data centers are located across the United States and Europe. Its $1.6 billion supercomputer data center for Nvidia in Plano, Texas has been described by Nvidia as the fastest AI supercomputer in the world. The company is also developing Project Horizon — a West Texas campus targeting up to 2 gigawatts of AI compute capacity powered by on-site natural gas generation, with the first 250MW phase expected online by end-2026.

The single most important number in the CoreWeave story is the revenue backlog — and it is a genuinely extraordinary figure. As of December 31, 2025, CoreWeave's contracted revenue backlog stood at $66.8 billion. To put that in context: at the beginning of 2025, the backlog was $15.1 billion. Over the course of a single year, it grew by more than four times. That is not a gradual accumulation. That is an avalanche of committed long-term demand.

What the Backlog Means: This is not a pipeline of leads or a book of potential deals. The backlog represents committed customer contracts — revenue CoreWeave is contractually entitled to collect as it delivers services. A significant portion carries take-or-pay terms, meaning customers must pay whether or not they actually use the capacity. Of the $66.8B total, approximately $28 billion is expected to be recognized within the next 24 months.

The backlog is anchored by some of the most powerful companies in the AI ecosystem:

Management's 2026 and 2027 guidance is grounded in this backlog. The company targets an annualized revenue run rate of $17-19 billion exiting 2026, implying roughly 240-270% revenue growth for the full year. Looking further out, management projects revenue exceeding $30 billion by end-2027. These are not analyst estimates or aspirational targets. They are management's own projections, backed by contracts already signed.

CFO Nitin Agrawal has been explicit on this point: substantially all of the planned 2026 capital expenditure is tied to already-signed customer contracts. The company is not building speculatively. It is building to fulfill committed demand with direct visibility into the long-term margins and cash flows those deployments will generate. That level of revenue predictability is exceptionally rare for a company growing at this velocity.

The Scale of What Is Being Built: CoreWeave plans to more than double its active power capacity from approximately 0.85 gigawatts at end-2025 to over 1.7 gigawatts by end-2026. For context, 1.7 gigawatts of continuous power is roughly equivalent to the output of a mid-sized nuclear power plant, consumed entirely by AI computing infrastructure.

Here is where the story gets genuinely complicated — and where the stock dropped nearly 20% in a single session following the Q4 2025 earnings report on February 26, 2026. Despite beating revenue estimates and delivering the extraordinary backlog growth described above, the market reacted with alarm to three interconnected concerns: the scale of debt already on the balance sheet, the magnitude of planned capital expenditure, and the implications of both for equity holders.

CoreWeave carried $21.37 billion in total debt as of December 31, 2025. For a company with $5.13 billion in annual revenue and a market cap of approximately $39 billion, that is an extraordinary debt burden. To frame it: the debt load is more than four times annual revenue and over half the market capitalization of the company.

This debt is not a reckless accident. It is a deliberate financing model. CoreWeave funds its GPU and data center buildout using asset-backed debt — loans secured by the physical GPU clusters and the committed customer contracts that generate revenue from those clusters. The logic is that the assets are real and the contracts are real, so the debt is serviceable. The company also raised $2.6 billion in convertible senior notes in Q4 2025 alone, and expanded its revolving credit facility to $2.5 billion.

The immediate concern is interest expense. CoreWeave paid approximately $1.2 billion in annual interest expense in 2025. That interest burden is the primary reason the company remains net-loss making despite strong EBITDA margins (~57% adjusted EBITDA). Every dollar of revenue growth that goes toward servicing debt is a dollar not flowing to equity holders. The company has made progress reducing its weighted average interest rate — down 300 basis points in 2025 alone, and 600 basis points from the 2023 peak — which is meaningful. But the absolute quantum of interest paid is still enormous.

The Debt Structure: Management emphasizes that the debt is structured to match the asset and contract lifetimes. Importantly, CoreWeave has no debt maturities before 2029 other than self-amortizing contract-backed debt and vendor financing. This means the near-term liquidity cliff risk that some feared is not the core issue — the issue is the ongoing cost of carrying the debt and what happens if AI demand softens.

The number that truly shocked the market in the February earnings call was the 2026 capital expenditure guidance: $30 to $35 billion. To understand the magnitude, consider that CoreWeave spent $14.9 billion in CapEx in 2025 — already an enormous sum. Management is guiding for more than double that in 2026. In a single year. For a company with $5 billion in trailing revenue.

Management's defense of this number was clear and consistent: virtually all of the planned 2026 CapEx is tied to already-signed customer contracts that will generate revenue as the capacity comes online. The CFO stated: "This is a reflection of the extraordinary amount of contracted demand in front of us." The company is not building on spec. It is fulfilling commitments.

Analysts, however, were not entirely reassured. D.A. Davidson's Alexander Platt noted that the CapEx surge was "more pronounced than expected" and highlighted the gap between when CapEx is incurred (immediately, when leases and hardware are acquired) and when revenue ramps (over subsequent months as customers begin actually using the capacity). This timing mismatch creates a window where the company is spending before it is earning, compressing near-term cash flow even when long-term contracts are in place.

JPMorgan cut its price target from $110 to $90 following the earnings report while maintaining a Neutral rating. Bernstein reiterated an Underperform rating with a $56 target, arguing that CoreWeave's big-tech customers would eventually cannibalize demand by building their own infrastructure in-house. The first-quarter revenue guidance of $1.9-2.0 billion also came in below Wall Street's $2.24 billion estimate — a meaningful miss that compounded the market's anxiety.

Debt at this scale eventually has to be refinanced, repaid, or converted to equity. CoreWeave has $2.6 billion in convertible senior notes outstanding. If those notes convert to equity, existing shareholders face dilution. The company also issued $350 million in common stock to OpenAI as part of their contract arrangement. And the February 2026 debt financing — seeking $8.5 billion in new financing using its Meta contracts as collateral — signals that the fundraising cycle is ongoing and relentless.

The concern among bears is structural: CoreWeave is effectively inventing a new financing model for AI infrastructure — one where the capital cycle is so compressed and so large that the company may need to continuously raise equity or convertible debt at dilutive terms to bridge the gap between CapEx incurred and revenue earned. If the revenue materializes as contracted and on schedule, dilution is manageable. If there are delays, contract renegotiations, or demand softening, the equity cushion erodes quickly.

The Circular Relationships: One concern some analysts have flagged is the interlocking nature of CoreWeave's largest relationships. Nvidia is a hardware supplier, a major customer (take-or-pay contract), and an equity investor. Microsoft is a competitor (Azure), a customer, and a former dominant revenue source. OpenAI is a customer and an equity holder. Meta is a customer and a potential competitor (building its own data centers). These relationships are symbiotic — until incentives diverge.

In 2024, Microsoft alone accounted for 62% of CoreWeave's revenue. That has diversified meaningfully through 2025 as OpenAI and Meta have grown into larger proportions of the backlog, but concentration risk remains real. The top two customers — OpenAI and Meta — together represent over $36 billion of the $66.8 billion backlog, or roughly 54% of total committed revenue.

If either customer renegotiates, reduces their commitment, or decides to accelerate their own in-house infrastructure buildout, the impact on CoreWeave would be material. Satya Nadella has described the original Microsoft-CoreWeave relationship as a temporary gap-fill while Azure caught up — he has since clarified that Microsoft will continue to lease compute capacity through at least 2027-2028, but the long-term trajectory of that relationship bears watching.

CoreWeave is a company where the bull and bear cases are both genuinely compelling, and that tension is exactly why the stock trades with such volatility. It hit $187 and it has traded below $34. Both price points reflect real investor perspectives, not irrational extremes.

The bull frame: CoreWeave is the essential infrastructure layer of the AI era, with $66.8 billion in contracted backlog providing revenue visibility that almost no other hypergrowth company in history has enjoyed. The CapEx is massive but directly tied to committed demand. The debt is expensive but structured to match asset and contract lifecycles. As capacity comes online and revenue ramps, adjusted EBITDA margins of 57-62% suggest the underlying economics are powerful. The question is not whether the business works — the contracts prove it does. The question is how much of the value accrues to equity holders after debt service.

The bear frame: The cumulative financing deficit through 2030 may exceed $100 billion. Interest expense alone is running at $1.2 billion annually and growing. Q1 2026 revenue guidance missed expectations. Insider selling (the CFO disclosed share sales in March 2026) adds to unease. The biggest customers are also the biggest potential in-house competitors. And if AI infrastructure investment cycles turn — as they eventually do — CoreWeave is levered to the downside with $21 billion in debt and rapidly growing obligations.

Both frames are internally consistent. The outcome depends almost entirely on whether AI infrastructure demand remains at the intensity it has demonstrated over the past three years — or whether something changes.

CoreWeave is one of the most fascinating — and genuinely difficult — investment cases in public markets right now. It is simultaneously the company with one of the most compelling revenue visibility stories ($66.8 billion in contracted backlog, a four-times increase in a single year) and one of the most challenging balance sheet profiles ($21.37 billion in debt, $30-35 billion in planned 2026 CapEx) of any publicly traded technology company.

The Q4 2025 earnings report told both sides of that story simultaneously — record backlog growth and a stock down 20% on the same day. That reaction is not irrational. It reflects the genuine tension at the core of the CoreWeave thesis: the demand is real and the contracts are signed, but the cost of meeting that demand is escalating at a rate that puts enormous pressure on the equity, the balance sheet, and management's execution capacity.

What is not in question is the significance of what CoreWeave is building. The world's most powerful AI companies — OpenAI, Meta, Nvidia — have collectively committed over $42 billion to CoreWeave's platform. That is a signal, not noise. These are not naive investors placing speculative bets. They are the companies building the AI systems that will define the next decade, and they have chosen CoreWeave as a critical part of their infrastructure stack.

For investors, the question is straightforward even if the answer is not: can CoreWeave convert that extraordinary contracted demand into equity value for shareholders — after servicing $21 billion in debt, financing $30-35 billion in new CapEx, and navigating the inevitable complexities of the most capital-intensive buildout in the history of cloud computing?

That answer is still being written — one data center at a time.

Sources:

DISCLAIMER

This article is published by Brezco Analytics for informational and educational purposes only. Nothing contained herein constitutes financial, investment, legal, or tax advice. All information is believed to be from reliable sources but is not guaranteed. Past performance is not indicative of future results. Investing in securities involves risk, including the possible loss of principal. The author may or may not hold positions in securities discussed. Always conduct your own due diligence and consult a licensed financial advisor before making investment decisions. This is not a solicitation to buy or sell any security.

© 2026 Brezco Analytics  |  brezcoanalytics.com  |  All Rights Reserved