The next phase of the artificial intelligence investment cycle is likely to hinge on three factors: guidance from hyperscalers, the pace at which AI investments translate into revenues and the direction of global bond yields, according to a report by Dolat Capital.
The brokerage said the AI buildout is entering its “first meaningful macro test” as rising US and Japanese yields, fiscal concerns and a higher cost of capital begin to challenge an investment cycle that requires substantial funding.
“The next leg of the AI momentum will be determined by hyperscaler’s guidance, the pace of AI monetization and the direction of yields,” Dolat Capital said in its report, AI Buildout: The First Macro Test, dated September 16.
The current AI investment cycle is different from previous technology cycles, the brokerage said. Hyperscalers, which traditionally operated relatively asset light and returned cash to shareholders, are increasingly becoming large capital absorbers as they commit billions of dollars to data centres, computing infrastructure and related capacity.
According to Dolat Capital, these companies are increasingly relying not only on internal cash flows but also on debt and equity to fund the expansion. At the same time, heavy sovereign issuance, policy normalisation in Japan and competing infrastructure and defence capital expenditure are adding to pressure on the global cost of capital.
The yield environment has already become more challenging, with US and Japanese yields rising amid fiscal concerns and a recent crude oil shock. A more hawkish central bank stance could further raise the funding hurdle for AI infrastructure investments.
Monetisation remains the key question
While demand for AI continues to expand, Dolat Capital said the more important question for investors is whether the returns generated by the technology can keep pace with the capital being deployed.
The brokerage pointed to model commoditisation, falling token economics, improving model efficiency and rapid technological changes as factors that could make monetisation more difficult. With successive generations of AI models emerging quickly, companies may have a limited window to monetise each investment before the technology evolves further.
Recent comments from AI CEOs calling for a more measured pace of frontier model development add another layer of uncertainty, potentially extending the time required for monetisation even as infrastructure commitments remain elevated.
“The key risk is not demand for AI, but whether incremental investment continues to generate sufficient returns to sustain the current pace of spending,” Dolat Capital said.
This creates a potential mismatch for investors. Hyperscalers may continue to commit significant amounts of capital to ensure they remain competitive in AI, while the revenue payoff from those investments could take longer to emerge.
US yields could become increasingly important
Dolat Capital expects the trajectory of the US 10 year Treasury yield, Federal Reserve communication and the duration of the rate cycle to increasingly influence the AI investment cycle.
The brokerage highlighted the refinancing requirement for US Treasuries, estimating that around $8 trillion of Treasuries require refinancing. It also pointed to narrowing US Japan rate differentials following policy normalisation and a lower share of Treasuries being held by previously large holders such as China.
Importantly, the recent increase in yields appears to have a higher real rate component, suggesting that supply driven pressure on the longer end of the yield curve could prove more persistent.
For equity markets, this combination creates a more complicated backdrop. Higher yields and tighter liquidity can put pressure on valuations, while uncertainty around AI monetisation could make investors more sensitive to the scale and pace of hyperscaler capital expenditure.
Why hyperscaler guidance matters
The next round of results and management commentary from major hyperscalers could therefore assume greater importance. Investors will be watching not only the quantum of AI capex but also whether companies maintain their spending plans, how quickly capacity is being deployed and what management teams say about returns and monetisation.
Any moderation in hyperscaler capex could have implications beyond technology stocks because AI investment has become an increasingly important component of the global capital expenditure cycle, Dolat Capital said.
The brokerage said a combination of higher yields, tighter liquidity and uncertain AI monetisation creates a challenging backdrop for global equities.
For the AI trade, the focus is consequently shifting from simply measuring how much companies are spending on infrastructure to determining whether that spending can generate adequate returns and how the changing cost of capital affects the economics of the buildout.
