The split favours companies supplying the largest cloud and AI spenders, such as chipmakers and advanced manufacturing, over hardware makers relying on broad corporate demand. That divergence is likely to persist in stock performance until enterprise orders show clearer growth. Memory prices are the swing factor: a levelling off would pressure hardware revenue, while continued strength would delay the reckoning. Upcoming results from the biggest cloud providers are the next major catalyst, since any sign of slowing AI budgets would undermine the bullish half of the picture.
The AI boom has not cracked. It has just revealed a split: the hyperscalers are spending as hard as ever, while corporate IT departments are still deciding.
Summary:
- TSMC beat forecasts and raised its outlook, saying AI demand is extremely robust and outrunning supply
- Barclays downgraded Dell, HPE and NetApp, arguing enterprise AI spending has not yet inflected
- Barclays says higher memory prices, not demand, are flattering hardware revenue
- The two views reflect different customer bases: hyperscalers versus ordinary businesses
- The divergence points to concentrated rather than broad AI demand
- Hyperscaler results, memory prices and enterprise orders are the markers to watch
Two of this week’s biggest technology stories appear to send opposite messages about artificial intelligence. Taken together, they suggest the AI boom is real but narrower than some valuations assume. Demand is concentrated among the largest cloud and AI companies rather than spreading evenly through corporate IT budgets.
The bullish signal came from Taiwan Semiconductor Manufacturing Co., the world’s largest contract chipmaker. Its third-quarter profit rose nearly 70%, it raised its full-year growth outlook, and Chief Executive C.C. Wei said AI demand remained extremely robust with no sign of a slowdown. High-performance computing, the segment that includes AI accelerators, now makes up more than two-thirds of TSMC’s revenue, and the company says supply still falls well short of demand.
The cautious signal came a day earlier from Barclays. In its downgrade of Dell, the bank argued that enterprise AI spending has yet to show a meaningful inflection. It said strong revenue at hardware makers owes much to higher memory chip prices being passed on to customers. Barclays applied the same reasoning to Hewlett Packard Enterprise and NetApp, and the selloff spread across AI hardware stocks.
These views are less contradictory than they look. TSMC’s customers are dominated by the chip designers supplying the largest cloud providers and AI developers, whose spending remains heavy. Dell and its peers depend more on ordinary businesses building their own AI capacity, which is where Barclays sees the gap between expectation and reality.
The distinction matters for investors. If AI demand is concentrated at the top, the companies supplying the biggest spenders are on firmer ground than those relying on broad corporate adoption. It also means part of the recent revenue growth in enterprise hardware reflects pricing rather than volume, which could reverse if memory prices level off.
There are reasons for caution on both sides. Concentrated demand depends on a small number of very large buyers continuing to spend, while enterprise adoption could still accelerate as businesses move from experiments to deployment. Neither trend is settled.
Several markers will help resolve the debate:
- Hyperscaler spending plans: upcoming quarterly results from the largest cloud providers will show whether their AI budgets are still rising.
- Memory prices: a levelling off would test how much hardware revenue growth is genuine demand.
- Enterprise orders: hardware makers’ own results will show whether AI server sales are broadening beyond a handful of large customers.
This article was written by Eamonn Sheridan at investinglive.com.