Freedom analysts believe that the scale of investment in AI infrastructure is increasing companies’ reliance on external financing and raising the cost of forecasting mistakes on demand. According to the Bloomberg consensus cited in the report, the combined amount by which capital expenditures exceed operating cash flow for the companies under review will total $108 billion in 2026 and $210 billion in 2027. The experts maintain a long-term positive view on AI, but point to short-term risks related to the cost of capital and payback periods.
Why the rise in investment is becoming a risk
Companies are expanding computing capacity in anticipation of multi-year growth in demand for artificial intelligence. However, spending on infrastructure buildout and equipment may outpace proceeds from core operations.
Capital expenditures, or CAPEX, reflect investment in long-term assets. Operating cash flow shows how much cash the business’s current operations generate. If CAPEX exceeds that cash flow, the difference has to be covered by cash reserves, customer advances, borrowing, or issuing shares.
In Freedom analysts’ view, the key question is whether AI monetization will justify the scale of investment. Growth in the use of the technology by itself does not guarantee sufficient profit to finance new capacity.
Which companies’ spending exceeds cash flow
Differences among companies are significant. According to the Bloomberg consensus as of October 27, Microsoft’s (MSFT) CAPEX-to-operating-cash-flow ratio is 81% in the 2026 forecast and 87% in 2027. In both cases, capital expenditures remain below cash flow.
At Alphabet (GOOGL), the metric rises from 97% in 2026 to 117% in 2027. For Meta Platforms (META) it is 105% and 132%, and for Amazon (AMZN) — 112% and 118%, respectively.
For Oracle (ORCL), the ratio is estimated at 175% for 2026 and 113% for 2027. Dependence on additional sources of funds is even higher at CoreWeave (CRWV): 487% and 309%. At Nebius Group (NBIS), the figure reaches 559% and 371%.
Thus, for a number of companies, expected capital spending is several times higher than funds coming in from operating activities. At the same time, the authors of the report уточняют, that definitions of forecast CAPEX may differ across companies.
What the $210 billion gap means
The $108 billion for 2026 and $210 billion for 2027 cited in the report represent the sum of the amounts by which capital expenditures exceed operating cash flow. This metric is not the companies’ aggregate loss or the volume of mandatory new borrowing.
The gap can be closed in several ways. However, each source has its limitations: cash reserves get depleted, debt requires servicing, and issuing new shares increases the number of shares outstanding.
For investors, Freedom believes it is important to track not only cloud business growth rates and contract volumes, but also how companies finance infrastructure expansion.
The credit market is already repricing risks
The report notes higher costs of insuring the credit risk of Oracle and CoreWeave. As of October 26, the cost of five-year Oracle CDS reached 304 basis points, up 159 basis points since the start of the year. For CoreWeave, the metric was 1121 basis points, up 351 basis points. For comparison, the investment-grade bond index measure reached 63 basis points, up 13 basis points.
CDS is a credit risk insurance instrument. A rise in its cost reflects a higher risk assessment by market participants. The analysts conclude that new capital for Oracle and CoreWeave is becoming more expensive. Nebius is also dependent on investment, but has a lighter debt burden.
Future obligations are broader than balance-sheet debt
Another risk is tied to future contractual payments for procurement and leases. They need to be considered separately from debt obligations already reflected on the balance sheet. In the report, such future payments for Microsoft are estimated at $558 billion versus $129 billion of debt and leases on the balance sheet. For Alphabet — $902 billion versus $121 billion, for Meta — $810 billion versus $115 billion. Meta’s finance leases are not disclosed separately and are not included in the stated amount of balance-sheet obligations.
Why strong demand doesn’t remove the payback question
Freedom highlights several factors that could limit returns on infrastructure investment: underutilization of capacity, obsolescence of GPUs, declining rental prices, and rising energy and memory costs.
Additional pressure comes from competition among AI developers. According to the report, the average price per 1 million tokens fell by 54% from the peak in May 2026. To offset cheaper services, companies need to increase usage volumes.
At the same time, the authors do not deny the technology’s potential. Their main conclusion is that long-term growth in productivity and new markets is compatible with a short-term revaluation of certain stocks. Financing conditions, timelines for bringing capacity online, and converting contracts into cash receipts remain key metrics for validating the investment thesis.
This is not an individual investment recommendation.