In a corner of the artificial intelligence market that once seemed immune to doubt, investor sentiment has shifted sharply, exposing growing anxiety about whether the vast sums being poured into AI infrastructure will generate commensurate returns.
Broadcom, CoreWeave, and Oracle — three companies central to the global AI buildout — extended their recent selloff on Monday, deepening losses from last week. The declines have come even as all three firms continue to post strong year-to-date gains, underscoring that the market’s unease is not about the relevance of AI, but about the sustainability of its economics.
The recent weakness reflects a broader reassessment underway on Wall Street. After years of enthusiasm driven by explosive demand for computing power, chips, and data centers, investors are increasingly focused on balance sheets, profit margins, and funding risks. The question now being asked is less about growth and more about how long companies can keep spending at this pace before returns begin to materially justify the investment.
“It definitely requires the ROI to be there to keep funding this AI investment,” Matt Witheiler, head of late-stage growth at Wellington Management, said on CNBC’s “Money Movers” on Monday. He added that, based on what he has seen so far, “that ROI is there.”
Witheiler also outlined the optimistic case still underpinning much of the sector.
“Every single AI company on the planet is saying if you give me more compute I can make more revenue,” he said, a sentiment that has fueled unprecedented capital expenditure across the technology industry.
Yet the market’s reaction to recent earnings suggests that patience may be wearing thin. Both Broadcom and Oracle reported quarterly results last week that beat revenue expectations and reinforced the message that demand for AI-related products and services remains strong. Investors, however, focused less on top-line growth and more on the cost of delivering it.
Oracle’s situation has drawn particular scrutiny. The company has become increasingly reliant on debt markets to finance its aggressive expansion in cloud and data center infrastructure. Management said capital expenditure in the current fiscal year will rise to $50 billion, up from a previous forecast of $35 billion, following new long-term contracts with customers, including Meta and Nvidia.
At the same time, Oracle’s lease obligations have ballooned. As of Nov. 30, the company disclosed $248 billion in lease commitments related to data centers and cloud capacity, with terms stretching between 15 and 19 years. That figure represents a 148% increase from the end of August, highlighting the scale and speed of its expansion. Investors have been left searching for more clarity on how those commitments will be financed and what they mean for future cash flows.
Broadcom’s challenges are rooted less in leverage and more in profitability. Chief executive Hock Tan said AI chip sales are expected to double this quarter from a year earlier to $8.2 billion, driven by demand for both custom AI accelerators and networking semiconductors used in large-scale data centers.
However, producing complete AI server systems requires significant upfront spending. Broadcom’s chief financial officer, Kirsten Spears, cautioned investors that “gross margins will be lower” for certain AI chip systems as the company ramps up production and absorbs higher component costs. That warning reinforced concerns that earnings growth may lag revenue growth in the near term.
The market reaction has been unforgiving. Broadcom shares fell 5.6% on Monday after tumbling 11% on Friday, leaving the stock roughly 18% below the record high it reached just days earlier. Oracle slid another 2.7% on Monday and is now down 17% over the past three trading sessions. The stock has lost about 46% of its value since Sept. 10, when it surged after the company disclosed a massive AI backlog.
Concerns about leverage have further intensified the selloff. Venture capitalist Tomasz Tunguz, founder of Theory Ventures, wrote in a blog post on Monday that Oracle’s recent fundraising has pushed its debt-to-equity ratio to about 500%, a level far above those of its major cloud rivals. He noted that Amazon, Microsoft, Meta, and Google all maintain ratios between 7% and 23%.
CoreWeave, a cloud infrastructure provider built largely around Nvidia’s graphics processing units, is also under pressure. Tunguz pointed out that CoreWeave’s debt-to-equity ratio stands at roughly 120%, making it one of the most highly leveraged players in the AI infrastructure space.
CoreWeave’s stock performance reflects that unease. Shares fell about 8% on Monday after dropping 11% last week, and the company has now lost more than 60% of its value since peaking in June. The decline has occurred even as demand for GPU-powered cloud services remains strong, highlighting how sensitive investors have become to capital intensity and financial risk.
Taken together, the latest market moves suggest the AI trade is entering a more demanding phase. Enthusiasm for the technology itself remains intact, but the tolerance for mounting debt, margin compression, and open-ended spending is fading.
For companies racing to build the backbone of the AI economy, the challenge is no longer just scaling faster than competitors, but proving convincingly that the scale will eventually translate into durable profits.






