
The AI boom is changing the balance sheets behind the market’s growth story. Major technology companies are committing a larger share of their resources to investment while holding less cash relative to assets. For investors, the question is becoming how long internally generated funds can support the expansion—and what happens when external financing becomes more important.
The St. Louis Fed’s November 17 analysis, Hyperscalers’ Cash and Investment during the AI Boom, documents that transition. Alpha Trader News reads it as relevant to two audiences: markets assessing AI’s commercial promise and policymakers weighing the financial consequences of restrictive interest rates. That policy reading is our interpretation, rather than an announced Fed position.
Hyperscaler cash reserves are shrinking relative to assets
Researchers Julian Kozlowski, Juan M. Sánchez and Nicholas Sullivan examine 10 publicly traded technology companies: Alphabet, Amazon, Apple, Broadcom, Intel, IBM, Meta, Microsoft, Nvidia and Oracle. Their definition of hyperscalers includes semiconductor suppliers as well as cloud and data-center operators.
The study dates the AI era to 2018, following advances in deep learning. Its latest observations extend to early 2026; they are not a November balance-sheet snapshot.
| Measure | Earlier observation | Early 2026 |
|---|---|---|
| Cash and short-term investments / assets | About one-third in 2018 | About 16% |
| Capital expenditures / assets | About 4%–5% before 2018 | Nearly 13% |
| R&D expenditure / assets | About 7% before 2018 | About 9% |
| Combined capital and R&D expenditure / assets | 12.5% in early 2018 | 22% |
Source: St. Louis Fed, November 17, 2026. Figures are rounded and describe the study’s group, not every company individually. The cash ratio is weighted by total assets.
The cash ratio was already declining before ChatGPT’s November 2022 launch. It reached approximately 20% in late 2022 and 16% in early 2026. Other public companies moved from roughly 14% in 2018 to 11%, leaving a much smaller liquidity gap.
A lower cash-to-assets ratio does not necessarily mean cash fell in dollar terms. Asset growth can also reduce the ratio. The evidence establishes a smaller relative cash cushion; it does not establish that the largest technology companies are running out of money.
Physical investment now drives the acceleration
Capital spending rose to approximately 9% of assets in 2023, eased to 7% in 2024, then climbed toward 13% in early 2026. Other firms remained near 3% at the end of the period.
Research spending followed a different path: hyperscalers’ R&D-to-assets ratio peaked near 11% in 2023 before easing to 9%. It remains elevated, but capital expenditure is driving the latest expansion. These measures cover company spending, not exclusively AI projects, and a declining expenditure ratio does not prove a decline in expenditure dollars.
The economic implication is a growing commitment to physical capacity. Equipment and facilities must eventually generate sufficient customer cash flow to justify their cost. Strong adoption helps, but adoption alone cannot establish investment returns.
The first message: AI has commercial momentum, with a financing cost
Our reading of the St. Louis Fed’s research is constructive about AI’s economic potential while increasingly attentive to the funding burden. That should not be confused with institutional endorsement of particular companies or valuations.
The October FRED Blog analysis of AI adoption reports that more than half of businesses in Ramp’s measured population paid for an AI product or service by September 2026. Adoption increased across company sizes. This is evidence from Ramp’s sample, rather than a census of all U.S. businesses, and excludes free tools and some embedded AI.
Commercial use can therefore spread while financing conditions become more demanding. The market must assess both customer demand and the cash required to serve it.
A recurring theme across St. Louis Fed analyses
At least three publications develop the relationship between innovation, investment and liquidity:
- March 12: Financing R&D Spending connects innovation funding with corporate cash and warns of possible financing challenges as internal liquidity declines.
- May 18: Rising capital expenditures and declining cash holdings highlights the growing importance of external financing.
- November 17: Hyperscalers’ cash and investment documents the relative cash decline and acceleration in capital spending.
Experienced market participants may read the repetition as an increasingly important policy theme. These publications use different datasets and definitions, however; their numerical measures should not be spliced into one series. Recurrence also does not prove coordinated messaging or a forthcoming interest-rate decision.
The second message: another rate hike could intensify financial strain
ATN Analysis: As investment commitments expand and relative liquidity cushions diminish, an additional rate increase could place substantial pressure on businesses that depend on new borrowing or refinancing. The sensitivity may be greater among smaller developers and counterparties than among cash-generating technology leaders.
Higher financing costs can reduce project returns, delay revenue-producing capacity and make lenders less willing to fund expansion.
Systemic and market risk: thinner cash cushions and weaker customer credit
The risk to the financial system and markets becomes more significant when thinner corporate cash cushions coincide with trade financing extended to customers with weaker balance sheets. Suppliers can then face pressure on both sides: substantial investment commitments of their own and customer receivables that may be collected late or not at all.
Deferred payment terms, vendor financing and guarantees can support customer purchases, but they also create credit exposure. If financially weaker customers depend on continued borrowing to pay their suppliers, tighter credit can turn a customer funding problem into a supplier liquidity problem. Delayed collections or defaults can prompt suppliers to restrict credit and cut orders, transmitting pressure through the supply chain.
For markets, this can mean earnings disappointments, wider credit spreads, falling valuations and forced asset sales. Systemic risk arises if concentrated losses and shared financing exposures spread to banks, funds and other lenders while liquidity is already under pressure. The severity depends on leverage, guarantees, funding maturities and the capacity to absorb losses.
A lower cash ratio does not itself cause lending to weaker customers or establish that these exposures are widespread. This is ATN’s assessment of a potential amplification mechanism; the hyperscaler cash study does not measure customer credit quality or demonstrate systemic distress.
ATN’s argument is that further tightening could impose significant financial costs while doing little directly to resolve electricity, equipment or construction bottlenecks. Interest rates cannot manufacture additional chips or grid capacity. Delaying productive investment could also postpone future supply improvements.
That argument does not establish that a rate increase would have no effect on inflation: tighter policy can restrain aggregate demand. The policy trade-off is whether the incremental inflation benefit would justify the added financing pressure. The research does not estimate that balance or recommend an FOMC decision.
Financing strength differs across the AI supply chain
The accompanying St. Louis Fed Review paper distinguishes data-center builders, increasingly using debt alongside operating cash flow, from semiconductor producers whose cash generation has supported investment and stronger liquidity. A single AI label can conceal very different funding positions.
For markets, the useful questions are whether spending converts into collected revenue, how much investment operating cash flow covers, and when debt, leases and purchase commitments come due. Receivables growth, customer concentration and guarantees can reveal dependence that a cash balance alone misses.
These questions extend ATN’s analysis of AI financing and credit-squeeze risks. A productive technology can create economic value while some of its financiers suffer losses.
The central tension is now visible: AI adoption is expanding while the commitment of capital is becoming more demanding. Markets must distinguish durable cash generation from dependence on continuing credit. Policymakers must weigh inflation control against the consequences of tightening financing during a major investment cycle.