Morgan Stanley says hyperscaler AI spending set to exceed dot-com peak
Investing.com -- Morgan Stanley expects hyperscaler AI investment to enter what it calls “a new era,” with capital-intensity metrics forecast to exceed levels last seen during the dot-com boom.
Analyst Todd Castagno warned that spending plans have accelerated sharply while revenue expectations have not kept pace.
The analyst told investors in a note that hyperscalers’ cash capex-to-sales ratios are set to surpass the roughly 32% recorded during the dot-com era, reaching 34%, 39%, and 37% in 2026-2028.
He added that widespread use of leases is further lifting the headline figures and could push capex-to-sales “as high as 38%, 44%, and 45% in 2026-28.”
Hyperscalers are expected to drive about 40% of total Russell 1000 cash capex over 2026-28, representing more than $2 trillion.
Broader AI-related investment could exceed half of all R1000 capex, underscoring what Morgan Stanley described as the “growing concentration of index-level investment within AI.”
The bank highlighted that capex revisions have been “unprecedented,” with 2026-27 spending estimates up $630 billion compared with six months ago. By contrast, “revenue revisions lag and FCF estimates trend lower,” reflecting the multiyear timeline required to monetise AI infrastructure.
Castagno believes the scale of spending has created significant modelling challenges, noting that 2026-28 revisions “reflect step-changes rather than incremental increases.”
The bank added that rising fixed-cost bases mean future earnings and free cash flow will become “more sensitive to changes in revenue expectations.”
Semiconductor suppliers remain the clearest beneficiaries, with Morgan Stanley citing 2026 sales revisions up roughly 60%.
