CoreWeave’s share price has fallen almost 25% this month. It’s not because GPUs aren’t selling, but because the company appears unable to service its debt. The company is currently carrying roughly $50 billion of debt, and the unsecured corporate bonds it recently issued carry interest rates above 15%. It’s a model where it purchases graphics processing units with borrowed money and uses the usage fees to repay that debt, but the market has begun to doubt that structure.
If it were only CoreWeave’s issue, it would just be a single company’s problem. Yet even Anthropic, which is slated to go public in November, is seeing its valuation trimmed by some investors from $2 trillion to about $1.5 trillion. Revenue keeps growing, but the valuation is moving in the opposite direction. By the end of the month, Microsoft, Alphabet, and Amazon will each release their Q3 results, and the market is treating those numbers as a litmus test for whether we are in an AI bubble.
1. Trust Built on Debt Crumbles Before the Numbers
I call this scene the ‘trust balance sheet.’ Over the past two to three years, AI infrastructure investments were treated as assets simply by being announced. Statements like “we have secured tens of thousands of GPUs” or “we are building data centers” were enough to lift stock prices. Now those same declarations are being recorded on the liability side. Power constraints are delaying construction, procurement costs are soaring, and investors have realized that the interest accruing on borrowed money is outpacing the returns generated by the equipment purchased with that debt.
2. Declarations of ‘Adoption’ No Longer Earn Trust
The same logic applies to why the upcoming earnings releases are seen as a turning point. The market isn’t asking “how much was spent on AI?” but “where is that money coming back from now?” Analysts note that investors are no longer extending unconditional trust to companies that pour massive sums into AI. Is this a story that only applies to big tech? I don’t think so. Until a few years ago, simply inserting the phrase “AI adoption” into a press release earned a stamp of innovation. Today, if you can’t answer the follow‑up question “what actually changed?” with hard numbers, that same declaration can backfire.
3. Proof Has Become the New Business
So what should companies prepare? I believe that, rather than merely announcing whether they have adopted AI, it is far more critical to have the numbers that show where the investment was actually spent and where the returns are coming from. The due diligence that investors demand from big tech and the scrutiny that boards and customers will demand of our own AI spend are ultimately the same question: proof, not proclamation. When we examine a company’s technology assets, we repeatedly hit this roadblock. The investment exists, but many firms still lack a framework to quantify whether it is an asset or a liability. Our Tech DD '기술자산 가치 환산 + 기술부채' assessment converts redevelopment costs and hidden technical debt into monetary terms, giving you an immediate answer to that question. If you’re curious whether your company’s AI and technology investments are on the asset side or the debt side, feel free to reach out.