BTC Signals Pro | Bitcoin Trading Alerts & Market Insights
BTC Signals Pro | Bitcoin Trading Alerts & Market Insights
The sprawling data center in Plano, Texas, a metallic behemoth spanning over 450,000 square feet and costing a cool $1.6 billion, stands as a gleaming monument to the AI boom. It hums with 30 megawatts of raw computing power, the kind of infrastructure that fuels the insatiable appetite of AI models. It’s a physical manifestation of an economic wave that has lifted markets and driven significant U.S. growth. Yet, the company operating it, CoreWeave, isn’t just riding this wave; it might be the canary in the coal mine for a potential AI financial meltdown.
CoreWeave sits squarely at the epicenter of the AI economy, providing crucial computing capacity to titans like Microsoft and OpenAI. But here’s where the numbers start to tell a different story than the narrative of limitless growth: CoreWeave doesn’t own most of these vast data hubs. It’s built, by its own admission, on a staggering mountain of debt, obligations piled high in a frantic race to build out server farms for its high-profile customers. This mountain, I’d argue, looms significantly larger than the cash it’s actually pulling in. When CoreWeave announces earnings on Monday, everyone, from the most fervent bull to the most cynical bear, will be scrutinizing whether this company can actually outrun its liabilities. This isn't just about CoreWeave; it's a bellwether for the entire AI infrastructure gamble.
Let’s unpack those liabilities. CoreWeave carries $7.6 billion in current liabilities—that’s money due within the next 12 months—and a total of $11 billion in debt on its balance sheet, as per its August quarterly report. For a company like Google or Microsoft, these figures wouldn't even register. But CoreWeave's 2024 revenues were a mere $1.9 billion. The CEO projected 2025 revenues to land between $5.15 billion and $5.35 billion. That’s impressive growth, no doubt. But then you look at the capital expenditure (capex) for the same year: an estimated $20 billion to $23 billion. This kind of spending is a bold bet, to put it mildly.
And those numbers are just the start. Lurking off the balance sheet is a truly eye-watering obligation: $34 billion in scheduled lease payments set to kick in between now and 2028. Many of these stretch over a decade or more, but a substantial portion is for data centers still under construction, not yet generating a dime of revenue. This creates a precarious vulnerability. What happens if CoreWeave's numerous, often unprofitable, startup customers can't meet their contractual obligations? Or if construction delays mean CoreWeave can't deliver capacity on time, allowing customers to walk? This isn't just a minor operational hiccup; it's a fundamental risk to the entire edifice. I've spent years sifting through balance sheets, and this particular leverage profile, especially for a company in a nascent, capital-intensive sector, is what I find genuinely unsettling. It's like building a skyscraper on a foundation of IOU notes, hoping the tenants will pay up before the next construction bill arrives. One has to question the methodology behind projecting such rapid revenue conversion from RPOs when the underlying customer base, while blue-chip, is also heavily investing in its own infrastructure.
CoreWeave’s bulls point to its "remaining performance obligations" (RPOs) as the salvation. These are booked revenues, not yet paid, and as of June 30, they stood at a little over $30 billion, with about $15 billion expected within the next two years. The company has announced massive new deals, like a $14.2 billion agreement with Meta and a pact with Poolside for 40,000 Nvidia GPUs, which will undoubtedly boost this figure significantly. Wall Street forecasts for 2026 revenue range from $10.9 billion to $14.9 billion. The argument is simple: the revenue will come, in great quantity, and scale will solve everything, turning CoreWeave into the next Amazon Web Services—the "picks and shovels" of the AI gold rush.
But let’s talk about margins. CoreWeave reported razor-thin operating margins of 1.6% in the last quarter. After accounting for hefty interest expenses on its debt (much of it carrying rates between 9% and 15%, with a weighted average of 11%—to be more exact, far above the 5.5-6% for investment-grade corporate debt), those margins plunge into negative territory. The company lost over $600 million on $2.2 billion in revenue in the first six months of 2025. If the company hits, say, $12 billion in revenue next year and those 1.6% operating margins hold, that translates to a mere $192 million in operating income. That’s nowhere near enough to cover its interest expenses, let alone make principal repayments on its $11 billion in debt. Cantor Fitzgerald's Thomas Blakey says, "A lot has to go right." Indeed. A lot of margin expansion has to go right, not just revenue growth.
Then there’s the customer concentration. In Q2 2025, a staggering 71% of CoreWeave’s revenue came from Microsoft. While Microsoft’s credit rating is stellar, the company is also spending tens of billions on building its own data centers. As analyst Gil Luria points out, Microsoft will honor its contracts, but the likelihood of renewal is "much less guaranteed." And CoreWeave’s other major customer, OpenAI, has made commitments far exceeding its current cash flow. If OpenAI faces financial headwinds, CoreWeave might find itself at the back of a very long line of creditors, behind giants like Microsoft, Amazon, and Oracle. Can CoreWeave truly count on its customers' long-term loyalty and solvency given their own strategic shifts and financial commitments? It's a question that keeps me up at night when looking at these sorts of filings.
CoreWeave’s upcoming earnings report isn't just another quarterly update; it’s a high-stakes moment for the entire AI infrastructure narrative. The company’s pivot from crypto mining to AI, fueled by billions in debt and equity, has been meteoric. Its stock price surged a staggering 160% post-IPO—to be more exact, from a debut of $40 to $104 by Friday's close, despite a June peak. The bulls see this as evidence of an unstoppable force, a company poised to dominate a new technological frontier.
But the bear case, articulated by firms like Kerrisdale Capital, is stark: CoreWeave is "a debt-fueled GPU rental business with no moat, dressed up as innovation," facing a potential 90% downside. The math, as Luria says, is simple. Is CoreWeave getting a sufficient return on its investment to justify the exorbitant interest on its debt? The current data suggests a significant, widening gap. While creditors have shown some willingness to renegotiate terms, like delaying principal payments on DDTL 2.0, that only kicks the can down the road. Without an almost unimaginable surge in revenue and a dramatic improvement in operating margins, CoreWeave will need to keep borrowing, or renegotiate, or face an uncomfortable reckoning. The Plano data center, for all its power, might just be a monument to a bubble waiting to burst.
The allure of AI is undeniable, and CoreWeave has positioned itself at the heart of it. But the financial architecture supporting this ambition looks increasingly precarious. The numbers don't lie: a company bleeding cash, with razor-thin margins, colossal debt, and future obligations that dwarf current revenues, is making a bet so audacious it borders on reckless. It’s a classic case of growth at any cost, where the cost might just become unsustainable. When the music stops, the real question won't be how much computing power CoreWeave has, but how much cash it actually generates. And right now, that equation looks deeply flawed.