Inflation concerns, high oil prices, and uncertainty regarding the duration of the conflict in Iran are pushing interest rates higher. The yield on the 10-year U.S. Treasury note is currently above 5%. When it comes to the large-scale construction of artificial intelligence (AI) data centers, the cost of capital is of critical importance.
This is due to the sheer scale of the investment. Estimates for spending on AI data centers, energy infrastructure, and chips suggest that total outlays could reach $10 trillion between now and 2030. Rising financing costs could significantly alter the economic rationale for these infrastructure investments; higher discount rates reduce the present value of future profits.
In the initial phase of a market sell-off, such distinctions may be overlooked. Both cash-generating businesses and companies reliant on external financing can lose value simultaneously.
While markets can update pricing instantly, it may take longer for expectations regarding financing costs, project returns, and earnings to stabilize. If corporate earnings continue to grow while these expectations recalibrate, the potential for long-term returns could be quite substantial.
This is precisely where we identify long-term opportunities.
Building AI Infrastructure Relies on Financing
High interest rates raise the bar for investments aimed at expanding data center capacity. Investors must evaluate both the demand for computing capacity and the financial resources required to provide it.
Established platforms—such as Meta, Alphabet, Microsoft, and Amazon— It is necessary to distinguish between companies like CoreWeave, Nebius, Cipher, and Iren—often categorized as "neo-cloud" (or "next-generation cloud") providers. These firms differ in their business models, capital structures, and risk profiles; therefore, they should not be viewed as interchangeable options when investing in the artificial intelligence sector.
The questions are straightforward: To what extent can existing operating cash flow support capacity expansion efforts? How much of this expansion relies on new debt or equity issuance? When do these debts mature? Furthermore, what happens if the construction process, energy procurement, or the realization of customer revenue takes longer than anticipated?
Credit markets are already signaling concern. In September, it was reported that the five-year credit default swaps (CDS) for CoreWeave had climbed above 800 basis points. Based on the underlying pricing assumptions, this implies a default probability of approximately 40%–45%.
Look for Bottlenecks and Inflection Points
Our focus is on companies that address critical constraints within the AI ecosystem—specifically those where demand extends beyond the next wave of new data center construction.
This entails identifying businesses that help customers extract greater utility from their existing infrastructure and that resolve the bottlenecks currently facing hardware (such as memory limitations, networking challenges, orchestration issues, etc.).
These companies are not immune to a slowdown in capital expenditures. The key distinction lies in how much of their growth depends on adding new capacity versus making existing capacity more efficient and productive. As the sell-off matures, we expect investors to become more selective. Some price declines will reflect genuine financing and execution risks, while others may create opportunities in businesses with resilient earnings prospects.
