It has been a while since my last article on CoreWeave (“COREWEAVE: A FRAUD HIDING IN PLAIN SIGHT“), and considering what has happened in the past few months, plus the latest development last Friday, I believe it is time to write a new chapter in this saga. Before we dive in, please allow me to spend a few words on WeWork because it is important for what comes next.
WeWork pioneered the flexible coworking space by leasing commercial real estate on long-term contracts, renovating it into trendy offices, and subletting desks and offices to members on much shorter, flexible terms. However, this model had a fundamental flaw: it created a dangerous financial mismatch where the company was obligated to pay billions in fixed, long-term lease costs while its income from members was variable and could disappear quickly. These inherent risks were magnified by catastrophic corporate governance under founder Adam Neumann. While a rescue by SoftBank in 2019 after failing to IPO and a later SPAC merger provided temporary relief, the COVID-19 pandemic delivered the final blow by decimating demand for shared office space, leaving WeWork unable to cover its massive fixed lease obligations and ultimately triggering its Chapter 11 bankruptcy filing in 2023.
While WeWork and CoreWeave seem like two very different companies in appearance, the two are, in reality, very similar. While the first one was committed to very long-term fixed leases, the second commits to developing data centers that take years to build, and increasingly does so in partnership with lenders that will be the effective owners of the real estate in most cases, with CoreWeave as a long-term tenant. AI data centers require GPU infrastructure, and CoreWeave purchased that, for the most part, using expensive lending agreements pledging the very same GPUs as collateral. Effectively, similar to WeWork, CoreWeave owns little to nothing, legally speaking (even if, optically, CoreWeave’s GPUs are accounted among its assets).
While WeWork was signing up clients only committed to short-term leases, CoreWeave does sign clients with longer-term agreements up to 5 years, optically avoiding the duration mismatch that flawed WeWork’s business model. Why optically? Not only do CoreWeave clients effectively sign a “pay as you go” contract, but they also provide a limited cash guarantee in case they don’t use the full service agreed upon. In the case of OpenAI, not only did it emerge that the company did not pay any cash upfront to CoreWeave, but it also has the option to pay up to 360 days after using its service (Credit to Ed Zitron for the discovery). Furthermore, let’s not forget that, as part of the agreement, OpenAI even got warrants to invest in CoreWeave shares (Common Stock Issuance Agreement between CoreWeave, Inc. and OpenAI OpCo, LLC), effectively, an upfront discount. As if this wasn’t enough, OpenAI still doesn’t have the cash, and will hardly have the cash, to pay for the full amount of computing power it agreed to purchase from CoreWeave and its other suppliers, currently over $1.4 trillion in total, for the next few years.
This chart, taken from CoreWeave’s own 10-Q, helps to visualize the striking mismatch between commitments and revenues similar to WeWork.
In my opinion, as I wasn’t shy to highlight before, this same chart is also evidence of how CoreWeave’s business model is very similar to a Ponzi scheme.
Let’s add one more piece now, a piece I already highlighted in “NVIDIA EARNINGS: MORE QUESTIONS THAN ANSWERS ON THE STATE OF THE AI BUBBLE“. In its latest earnings, Nvidia disclosed a very odd deal with one of its partners, where it agreed to a lease facility up to $860 million, of which $470 million was placed in escrow. CoreWeave, for its part, reported in its latest financial statements $477 million of “non-current restricted cash and cash equivalents.”
Was this a sign that CoreWeave is now struggling to secure financing to continue the expansion of its data center infrastructure, so Nvidia had to step up and support its client, supplier, and investment? Fast forward to last Friday, January 2nd, 2026, when I believe the company hoped no one was paying attention, CoreWeave filed this 8-K report.
As you can read, CoreWeave agreed with its lenders to significantly renegotiate the terms of one of its borrowing agreements, not only significantly lowering the covenant thresholds that, if breached, would have triggered a default (a disaster for a highly leveraged company with $18.8 billion of debt), but it also agreed to eventually make up for missing payments by issuing equity to its creditors.
Similar to WeWork years ago, signs of financial stress around CoreWeave are undoubtedly mounting, and this is definitely not a good look for a company that already had to cut its capex plans (“CoreWeave growth surges despite capex delay“), meaning it will delay the point in time when it can start delivering computing power to its customers and thus booking revenues from those contracts, and has to repay $9.7 billion of debt in the next 12 months. I don’t think there is a different end to this story than CoreWeave filing for Chapter 11 or, in the best-case scenario, being acquired by a large company that cannot afford to see them go bust, like Nvidia. Ultimately, an end similar to WeWork’s one years ago.
SYNNAX PROMO
Get exclusive intelligence on stocks you can’t find anywhere else – insights that could give you a serious edge. Claim your free Synnax trial month now: https://synnax.app/PROMO/JUSTDARIO