CoreWeave的668亿美元订单，可能是AI泡沫的第一道裂缝

Core Viewpoint - The article discusses the risks associated with companies like CoreWeave that rely on debt to fuel growth in the AI computing market, highlighting that rapid revenue growth can mask underlying financial vulnerabilities [2][4]. Financial Performance - CoreWeave achieved the fastest-ever $5 billion in annual revenue in 2025, surpassing early-stage AWS and Azure, with a backlog of remaining performance obligations (RPO) reaching $66.8 billion, more than tripling from the beginning of the year [4]. - However, the financial report reveals alarming figures: Q4 earnings per share (EPS) loss of $0.89, a 1.6 times year-over-year increase, and an operating loss of $89 million, with a net loss of $452 million, nearly nine times that of the previous year [6][7]. Capital Expenditure and Debt - CoreWeave anticipates capital expenditures (CapEx) of at least $30 billion in 2026, three times that of 2025, indicating a reliance on future cash flow to finance current GPU and data center investments [7]. - The business model involves borrowing to purchase GPUs and build data centers, betting on future demand, which raises concerns about sustainability given the high levels of debt and losses [7][9]. Market Structure and Risks - The AI computing market mirrors the telecom bubble of the early 2000s, where companies over-leveraged based on perceived unlimited demand, leading to systemic risks when actual demand fell short [9][10]. - CoreWeave's customer base is highly concentrated among major tech companies, which possess strong bargaining power. If these companies increase in-house computing capabilities or demand lower prices, CoreWeave's profit margins could be severely impacted [10]. Cash Flow and Valuation Concerns - The backlog of $66.8 billion in orders does not equate to cash on hand, and the $30 billion CapEx represents real cash outflow, creating potential cash flow issues if customer deployments are delayed or actual usage rates fall [10][11]. - The risk of asset depreciation is significant, as GPUs are fast-depreciating assets. If CoreWeave incurs high debt to purchase GPUs at peak prices, a decline in rental prices could lead to substantial asset write-downs [11][14]. Debt as a Risk Amplifier - The article identifies three scenarios that could exacerbate CoreWeave's debt issues: a slowdown in computing demand, a drop in GPU prices leading to asset impairment, and pressure from major clients to lower prices or delay payments [14]. - The financial structure of companies like CoreWeave serves as a barometer for the overall health of the AI infrastructure sector, indicating that high leverage could lead to significant vulnerabilities in the event of market shifts [13][14]. Conclusion - The article concludes that while the AI boom is real, the tolerance for high leverage and rapid expansion is cyclical. Companies relying on borrowed capital for growth may face severe challenges during periods of tightening liquidity [16].