The current artificial intelligence (AI) capital expenditure (capex) cycle bears a striking resemblance to the shale boom of 2010–2015, during which the largest US producers spent themselves into negative free cash flow before the cycle ultimately corrected.
Aggregate trailing twelve month free cash flow among the largest hyperscalers has declined over 90% from its late 2024 peak.
Something seems oddly familiar about this AI capex cycle: early investment is funded by cash flows, returns look compelling, and the narrative of limitless demand justifies ever-larger commitments. Picks and shovels companies massively outperform. Then capital spending exceeds internal cash generation, companies turn to external financing, and the cycle enters a far more fragile phase. We have seen this before, and the data suggest we may be watching it unfold again.
The US shale boom provides a clear historical parallel. It was driven by a new technology that would transform US energy production and feed infinite consumer demand. Companies raced to acquire acreage and drill, and capex expanded accordingly, to the massive benefit of oil services companies. By late 2014, aggregate free cash flow had turned negative as spending far outstripped cash generation, and the group did not recover until years after the 2015 collapse in crude prices. The lesson was not that shale was a bad business, but that the pace of investment ultimately exceeded the industry's ability to fund its own growth despite the attractive underlying economics.
The AI hyperscaler complex is tracing a remarkably similar arc. Aggregate estimated free cash flow across the five largest hyperscalers peaked in late 2024, when capex was comfortably funded by enormous operating cash flows. Since then, estimated free cash flow has dropped over 90% in just eighteen months to the massive benefit of the “AI services companies.” The trajectory is steeper than the shale analog, and the numbers and returns are an order of magnitude larger.
The financing behavior confirms what the cash flow data implies. Since late 2024, these same companies have raised or announced nearly $500 billion in debt and equity financing. When companies with strong underlying businesses resort to simultaneous equity dilution and leveraged debt issuance to fund capex, we are no longer in the organic growth phase of the cycle.
None of this means artificial intelligence is a fad—just as the shale bust did not mean the US was demanding less oil. The underlying technology and demand are real. But capex cycles have their own logic, and when investment spending transitions from internally funded to externally financed, the risk profile changes materially. Whether the AI cycle follows the same path depends on whether the returns on these massive investments materialize before the financing costs catch up.
Important Disclosures & Definitions
Capital Expenditures (CAPEX/capex/CapEx): refers to investments in physical assets such as plant and machinery.
AAI001177