Investors Hated CoreWeave’s Earnings. It Might Be Even Worse Than They Realize

Artificial intelligence (AI)-focused cloud infrastructure provider CoreWeave (NASDAQ:CRWV) released its fourth quarter earnings on Thursday, and the market’s reaction was swift and brutal. Shares plummeted 18.5% on Friday, wiping out billions in market value as investors recoiled from a wider-than-expected loss and announcements of escalating capital expenditures. 

The company reported quarterly revenue of $1.57 billion, slightly beating estimates, but posted an adjusted loss per share of $0.56, far exceeding the anticipated $0.49 per share deficit. Full-year 2025 revenue hit $5.1 billion, up 168% year-over-year, yet the emphasis on ramping up spending to fuel AI infrastructure growth sparked fears of overspending in a competitive landscape. 

Guidance for first-quarter revenue of $1.9 billion to $2 billion fell short of the $2.29 billion consensus, exacerbating concerns. The problems, though, may be much deeper and far worse than investors suspect.

Growth Meets Mounting Losses

CoreWeave is a company in hyper-growth mode, but one grappling with profitability headwinds. Q4 adjusted EBITDA stood at $898 million with a 57% margin, yet the net loss ballooned to $452 million due to soaring interest expenses and accelerated deployments. 

CoreWeave’s contracted backlog swelled to a record $66.8 billion, with an average contract length of five years, signaling robust demand from hyperscalers, AI firms, and enterprises. However, debt tripled year-over-year, share count more than doubled, and expenses in sales, marketing, and infrastructure spiked, driving the deeper losses. 

Management forecasts margins bottoming in Q1 before climbing to low double digits by year-end, targeting 25% to 30% long-term, but skeptics see red flags everywhere.

A Stark Divide on the Business Model

In a scathing X thread, famed short-seller Jim Chanos lambasted the results as “disastrous,” arguing that even under a generous 10-year depreciation schedule for GPUs — far longer than the industry-standard six years — CoreWeave would still be unprofitable after accounting for all economic costs. Interest expenses would also overwhelm operations, driving pre-tax income negative starting in Q2 2024 and plunging to $811 million by Q4. 

Chanos also highlighted CoreWeave’s liquidity woes, which are at “distressed” levels signaling potential cash crunches. He predicts CoreWeave needing to issue new stock very soon beyond the sale of any Nvidia (NASDAQ:NVDA | NVDA Price Prediction) holdings, further diluting shareholders. 

In his view, CoreWeave is essentially renting GPUs “below cost,” relying instead on others’ data centers, and is overvalued at 9x revenues for such meager returns.

Rittenhouse Research, however, countered with a more optimistic take, emphasizing that CoreWeave isn’t prioritizing short-term profits amid explosive growth. Gross margins are compressing not from inefficiency but from upfront costs like data center leases and $73 million in tenant fit-outs for 100 megawatts of capacity, incurred before revenue kicks in. Operating expenses rose modestly from $303 million to $331 million in 2025, dropping as a percentage of revenue from 31% to 21%. 

With $17 billion in 2026 annual recurring revenue already booked, Rittenhouse projects $17 billion in total revenue at 75% gross margins, $1.2 billion in opex, and depreciation at 50% of revenue, yielding 17% EBIT margins — close to 2024’s 20%. 

Chanos rebutted that even this implies under $3 billion in EBIT on $60 billion to $70 billion in capital employed, equating to mid-single-digit pre-tax returns on invested capital (ROIC) — unimpressive for an equipment lessor.

Echoes of Michael Burry’s Warning

If Chanos is correct, the issues are exacerbated by depreciation realities. Echoing billionaire investor Michael Burry’s warnings from last year, the standard six-year GPU lifespan may be overly optimistic. Burry argued that Nvidia’s rapid chip cycles — new generations emerging every one to two years that are two to three times more efficient — render older hardware economically obsolete in just two to three years: not physically broken but uncompetitive for premium AI workloads. Stretching depreciation inflates earnings by understating expenses, potentially by $176 billion industry-wide through 2028. 

For CoreWeave — which extended its depreciation schedule to six years in 2023 — a shorter two- to three-year life would balloon depreciation, erasing any EBIT coverage of interest and turning the business model unsustainable. This hidden “fraud” of aggressive accounting could mean CoreWeave’s losses are vastly understated, amplifying liquidity risks and valuation concerns.

Key Takeaways

For CoreWeave investors, this spells significant risk: The stock’s high multiple assumes flawless execution in the AI boom, but if demand softens or competition intensifies, heavy debt and capex could lead to distress. The company itself faces execution risks — scaling infrastructure without profitability could force restructurings. 

Yet, Rittenhouse’s defense has merit: Like Amazon (NASDAQ:AMZN) in the 1990s, CoreWeave’s upfront investments could pay off if the $66.8 billion backlog translates to sustained hypergrowth and margin recovery. 

Still, investors may soon face heavy dilution as CoreWeave raises capital to grow out of this hole. It may not be a stock to buy aggressively, even after Friday’s collapse.