Fed hike tests Nvidia, OpenAI, Oracle financing circle

The Federal Reserve raised interest rates by a quarter point on September 16, 2026, to a range of 3.75% and 4%, marking the first increase since 2023 and introducing a new cost pressure on the interlocking capital commitments among major AI firms. The move comes as the Bureau of Economic Analysis estimates that AI-related data center and software investment accounted for roughly three-quarters of the 2.1% annualized US economic growth recorded in the first quarter of 2026.

This growth is heavily dependent on a financing structure where capital flows in a loop. Nvidia (NVDA) invested $30 billion in OpenAI, which uses funds to purchase compute power from Oracle (ORCL), while Oracle buys chips from Nvidia to meet that demand. Estimates place the total value of these interlocking commitments between $1 trillion and $1.5 trillion. Michael Burry, an investor known for his bearish stance, described the revenue streams in this circle as self-referential in his critique of Nvidia's Q2 FY'27 financials. Pablo Hernández de Cos, General Manager of the Bank for International Settlements, echoed these concerns in a speech earlier this month, noting that the capital expenditure of the largest firms is outpacing their cash flows, prompting a growing reliance on debt and private credit.

The financial strain is evident in OpenAI's reported figures. Some sources claim its Q2 2026 revenue was $6.7 billion, an 18% increase from the previous quarter, but its Q2 operating loss reached $12.3 billion, up 32% from $9.3 billion in Q1 2026. With its IPO pushed back to 2027, the company must raise capital faster than its losses grow to sustain the infrastructure buildout. Nvidia CFO Colette Kress addressed the circular financing claims during the company's Q2 FY'27 earnings call, stating that the support is justified by a major computing platform shift.

The scale of spending across the industry is substantial. For 2026, projected capital expenditure for Amazon (AMZN), Alphabet (GOOGL), Microsoft (MSFT), and Meta (META) totals $745 billion. This figure is comparable to the gross domestic product of Sweden or Belgium. Much of this spending is funded through debt and special purpose vehicles rather than free cash flow. The Institute of Electrical and Electronics Engineers notes that a single GB200 AI server rack consumes as much power as 100 US homes, creating physical infrastructure demands that utilities are meeting with long-term rate base investments.

Accounting practices also play a role in how this spending is reported. Several hyperscalers have extended the estimated useful lives of servers and networking equipment to five or six years, which spreads costs over more periods and makes quarterly profits appear healthier than they might under shorter depreciation schedules. While the BIS acknowledged real productivity improvements from AI adoption, it warned that history offers parallels to previous technological booms, such as the British railway mania of the 1840s and the dot-com surge of the late 1990s, where massive capital spending was not fully justified by returns.

The recent rate hike acts as a stress test for this model because it increases the cost of the debt funding these projects at a time when free cash flow at major spenders is projected to go negative. Investors are now watching for signals such as changes in depreciation schedules, additional funding needs from OpenAI and Anthropic, and actual enterprise usage data to determine if current capacity buildout will be matched by proven demand.