In accordance with Jefferies’ Christopher Wooden, international head of fairness technique, the size of spending by US hyperscalers has reached a degree the place it’s consuming an more and more giant share of their money flows, significantly on chips and reminiscence. Based mostly on the newest firm steering, capex as a share of working money circulation for the 4 main US hyperscalers has surged from 41% in 2023 to a projected 92% in 2026.
A good portion of that is being directed in the direction of reminiscence alone, which is estimated to account for about 30% of whole capex, implying roughly 28% of working money circulation being absorbed by reminiscence investments this 12 months, he mentioned in his Greed and Worry report.
This rising depth of funding brings into focus a extra elementary query: monetisation. A latest Jefferies report led by Edison Lee highlights that the challenges round AI enterprise fashions stay underestimated. The growing value of staying aggressive, pushed by larger compute, reminiscence, and energy necessities, means that sustainable profitability for pure AI mannequin gamers stays distant.
Wooden aligns with this view. His base case is that AI might in the end resemble a capital-intensive trade like airways, reasonably than the high-margin, winner-takes-all dynamics seen within the web period.
Even so, the present section of spending reveals little signal of slowing. Massive Tech firms proceed to push forward with aggressive capex plans. Microsoft expects to spend $190 billion this 12 months, together with about $25 billion attributed to larger element prices. Alphabet and Meta have each raised their 2026 capex steering to $180–190 billion and $125–145 billion, respectively, whereas Amazon has maintained its steering at $200 billion.
Amongst these, investor issues seem extra pronounced within the case of Meta, which lacks the identical direct cloud-driven advantages from AI spending as friends like Alphabet, Microsoft, and Amazon.For now, the “picks and shovels” commerce stays intact, supported by continued spending and restricted pushback from traders on returns.
Nevertheless, early indicators of pressure are starting to floor. A latest report famous that OpenAI has missed inside targets for each person development and revenues, together with a purpose of reaching 1 billion weekly energetic customers for ChatGPT by the tip of final 12 months. The corporate has additionally reportedly fallen wanting a number of month-to-month income targets in 2026, whereas going through elevated competitors.
Market share traits mirror this shift. Over the previous 12 months to March, Gemini’s share of net visitors within the generative AI market has risen sharply from 6% to 25.5%, whereas ChatGPT’s share has declined from 77.4% to 56.7%, in accordance with SimilarWeb information.
On the similar time, issues have been raised about financing buildings throughout the ecosystem, the place companions corresponding to Nvidia and Oracle present funding to OpenAI, which in flip makes use of that capital to buy compute from them.
Competitors can also be intensifying. Anthropic reported in early April that its annualised income run charge has exceeded $30 billion, up from round $9 billion on the finish of 2025, now surpassing OpenAI’s reported run charge of over $25 billion in February.
Taken collectively, the image that emerges is one in every of escalating funding, rising aggressive strain, and unresolved questions round returns. The spending cycle continues, however the pressure it locations on money flows and the uncertainty round monetisation have gotten more and more troublesome to disregard.
(Disclaimer: Suggestions, strategies, views and opinions given by the specialists are their very own. These don’t signify the views of The Financial Occasions)
