The three major dilemmas of AI venture capital: returns, fundraising, and network effects

After Oracle recently announced an aggressive capital expenditure plan, its stock price plummeted sharply, sparking market discussion. This isn’t just simple price fluctuation but a symbolic shift—investor attitudes toward AI venture capital are being rewritten. Once filled with imagination, capital stories now face cold, hard questions: Can this money really be recouped? Will funding continue? Will these tech giants all run into trouble together?

We observe that the market is shifting from blind optimism to rational assessment. Supporting this change are three deep-rooted venture capital dilemmas: uncertain returns, worsening financing conditions, and amplified interconnected risks among enterprises.

Money Burning vs. Returns: The ROI Dilemma of AI Investment

The most notable feature of this AI wave is that tech companies are collectively entering a “money-burning mode.” According to FactSet, the five major hyperscalers have cumulatively spent $357.2 billion on AI-related capital expenditures over the past four quarters. Even more exaggerated, Bloomberg forecasts this figure will continue to swell to about $500 billion by 2026.

How outrageous is this spending? From a cash flow perspective, these five companies allocate an average of 60% of their free cash flow to AI capital expenditures. Oracle is the most aggressive—its capital expenditure accounts for 582% of its operating cash flow. What does this mean? It means Oracle is not only investing all its free cash flow into AI but also borrowing more to fill the gaps.

But the real question is: Can such heavy spending be profitable?

So far, the commercialization path of AI remains unclear, and profit models are still being explored. No one can accurately predict how much real profit this huge investment will generate in the end. More troubling, basic economic principles suggest that as investment scales up, marginal returns should decline. Yet, the costs of AI investments haven’t decreased—in fact, since 2023, prices for computers and information processing equipment have continued to rise. This sharply contrasts with the 1990s internet boom, when capital goods prices steadily declined. This indicates that current AI investments haven’t achieved economies of scale but are instead caught in a “diseconomies of scale” trap.

Market reactions are straightforward: investors are re-pricing. If expectations for returns were overly optimistic before, they now need significant correction. Oracle’s stock price decline is a direct reflection of this adjustment. Previously, a company mentioning “increased capital expenditure” could send its stock soaring. Now, investors demand real, tangible returns, not endless stories of investment.

This cyclical adjustment of expectations is common in capital markets. Historically, every wave of technological revolution has gone through a “first optimism, then disappointment” phase. Over a long cycle of more than ten years, investment activity tends to fluctuate in stages, with multiple expansions and contractions spaced 3-5 years apart. Stock prices, as leading indicators, also experience synchronized or even more volatile swings. As Keynes observed, stock buyers often lack a clear understanding of what they buy; speculators care more about how market sentiment will shift next than about rational estimates of future capital returns. When overly optimistic bubbles burst, market prices turn downward.

Worsening Financing Environment: Funding Pressure on Venture Capital

Large-scale investments require time, but the costs supporting them must be paid immediately. This creates a timing mismatch: companies need to spend real money before generating returns. These funds come either from internal cash flows or external financing. Whether they can raise money—and at what cost—depends entirely on financial market confidence in the company’s debt repayment ability—what we call credit conditions. Once confidence falters, both the difficulty and cost of financing rise rapidly.

Take Oracle as an example: its ambitious AI investment plan largely relies on external financing. According to its Q3 report, Oracle’s cash burn increased further, with free cash flow turning negative at -$10 billion. The balance sheet looks even more alarming: by fiscal year 2025, Oracle’s earnings are only $28.9 billion, but net debt stands at $97.7 billion—this huge gap indicates rising debt risk.

The market is now re-evaluating Oracle’s debt risk. The most direct indicator is the credit default swap (CDS) spread—reflecting investors’ perception of default risk. Oracle’s CDS spread has been rising over the past few months, surpassing 140 basis points and reaching the highest level since 2008. What does this mean? It indicates that creditors are increasingly wary of Oracle’s creditworthiness, demanding higher risk premiums to lend. In other words, Oracle’s future financing difficulty and costs are likely to rise significantly.

Oracle is not an isolated case. Other AI-related companies are facing similar issues: revenue growth falling short of expectations, yet financing needs increasing. For example, CoreWeave, which focuses on high-performance computing infrastructure, lowered its full-year revenue guidance in November due to delays in some customer contracts, causing its stock to fall. The company then announced plans to raise $2 billion via private convertible bonds, and its stock price tumbled again. Since November, CoreWeave’s stock has fallen 37%. Meanwhile, its bond CDS spread surged from below 400 basis points to about 773 basis points, indicating market concerns over its credit status and refinancing ability have intensified.

This creates a vicious cycle: revenue shortfalls → increased financing pressure → higher financing costs → squeezed operational space → further difficulty in refinancing. In such an environment, AI venture capital projects overly dependent on external funding face mounting survival challenges.

Chain Risks of “Linked” Enterprises: Interconnection Crisis

Another unique phenomenon of this AI wave is that tech giants have taken over roles traditionally held by venture capital. Companies like NVIDIA, OpenAI, and Oracle are not just market participants but have become primary financiers and industry leaders for startups. This model has advantages—strengthening industry synergy, reducing information asymmetry, and improving overall efficiency.

But problems also arise: this close web of interests creates new risks. When companies form complex investment and financing relationships, risks from a single enterprise can quickly spread along supply and capital chains, making the entire ecosystem fragile.

Currently, NVIDIA, OpenAI, Oracle, and others have built a dense network of connections across investment, cloud services, chip deployment, and joint R&D. Specifically: NVIDIA has committed to investing up to $100 billion in OpenAI, supplies $6.3 billion worth of cloud services to CoreWeave, invests $5 billion in Intel, and plans joint chip R&D. OpenAI has signed a $300 billion cloud computing partnership with Oracle, plans to pay up to $22.4 billion to CoreWeave, and deploys billions of dollars in AMD chips.

While this “binding” structure appears to create synergy, it actually amplifies risks. If any link encounters problems—such as a company failing or facing a liquidity crisis—the negative impact can quickly spread like a virus, triggering a chain collapse across the industry.

This has already begun to show. After Oracle’s stock sharply declined, other related companies’ stocks also weakened. Even Broadcom, which exceeded earnings expectations, saw its stock come under significant pressure. This indicates the market is reassessing the hidden risks behind the current “clustering” of AI companies. Investors are realizing that the seemingly collaborative cooperation may be weaving a tangled web of intertwined risks and shared vulnerabilities.

Macro Echoes: What AI Venture Capital Cooling Means for the US Economy

In 2025, the resilience of the US economy is largely attributed to expanding AI-related fixed asset investments and the wealth effects they generate. Estimates show that in the first half of 2025 alone, AI contributed about 0.7 percentage points to the year-over-year growth of US real GDP, accounting for one-third of overall growth. In other words, without AI, the traditional US economy lacks real growth momentum, and overall performance is far less impressive.

Looking ahead to 2026, the situation may change. If the market continues to doubt the return efficiency of AI capital expenditures and financing conditions for related companies tighten further, a reasonable expectation is that growth in AI-related fixed asset investments will face significant slowdown. This risk cannot be offset by easing monetary policy, as the root problem isn’t high financing costs but fundamental uncertainty over investment returns. Additionally, tariffs implemented by the Trump administration have increased supply costs, further raising prices for AI capital goods. These supply-side constraints are not easily remedied by Fed rate cuts.

More importantly, the wealth effect may weaken. Moody’s research shows that nearly half of US consumer spending is contributed by the top 10% income group, which holds about 87% of US stocks. Over the past few years, they have enjoyed substantial capital market returns. If market adjustments lead to wealth shrinking, consumption will be pressured. Meanwhile, the US labor market shows signs of ongoing weakness, with uncertain employment prospects dampening consumer confidence and further constraining spending.

Historical experience suggests that in late-cycle phases, insufficient consumption becomes a prominent issue. The current “K-shaped” divergence in US consumption—where the wealthy maintain relatively stable spending while middle- and low-income groups are constrained—may be a warning sign. This warrants ongoing monitoring and vigilance.

If AI venture capital cooling becomes a reality, its chain effects could far exceed expectations—from micro-level corporate financing pressures, through industry chain risks, to macroeconomic growth slowdown. The entire system is interacting and amplifying each other. In this context, the evolution of AI investment risks is becoming a key variable in determining future economic trajectories.